Today we take another look at the Ryzen chipset and discuss further optimization’s. Memory is always a question that comes up and historically it hasn’t really had an impact for audio, where the bottleneck in performance often ends up being elsewhere in the setup.
Even with the previous generations on Ryzen where the optimal memory advised were around the 2666MHz (first generation) – 3200MHz (second generation) clock speeds and in our own testing moving up from 2666MHz to 3200MHz on either generation didn’t get us any favourable results in audio benchmarking, although it did help for video rendering workloads.
As such, I went with the previous suggested the best memory when testing around launch and AMD has publically outlined that the optimum speed is now 3733MHz with a CAS16 timing as this puts the memory on a perfect 1:1 ratio with the internal Infinity Fabric bus arrangement.
At this point 3733MHz RAM is still not overly common, even more, uncommon is the super low CAS 16 kits. I’ve currently got a 3733MHz pack being shipped to me (although only CAS17) for further testing when it arrives, although I’ll keep that for when I do a full retest in the coming week.
The results I have today is more of a comparison to show some basic gains and at a slightly cheaper price point. Above 3600MHz memory carries a sizable price premium and some of you may be wondering what gains can be achieved at what price points.
To do this testing I’ve got results generated using the 3200 RAM used in the previous testing, 3600 RAM with CAS18 which are the standard packs we use here and then I’ve run the same 3600MHz RAM clocked up to 3733MHz, which in real-terms ended up being around 3725MHz running in Windows.
Stock CPU 3200MHz RAM
Stock CPU 3600MHz RAM
Stock CPU 3725MHz RAM
The DAWBench DSP test gave us some small gains on the 64 buffer and then became much more apparent at larger buffer sizes, where we’re talking closer to 8% at the 512 buffer.
Stock CPU 3200 RAM
CPU Over load Point
Stock CPU 3600 RAM
CPU Over load Point
Stock CPU 3725 RAM
CPU Over load Point
What we can see here is similar small gains moving from 3200MHz to 3600MHz, with it being fairly marginal overall moving up at this level.
Clocking the RAM up towards it’s advised 3733MHz clocks in this instance produced us more notable gains with excess to 10% being seen at most buffer sizes. I’ll also note that that between the 3600MHz and 3725MHz results the memory hole started to disappear as the CPU overload point moved upwards. I suspect and remain hopeful when we see perfectly matched 3733MHz RAM with CAS 16 timings as they’ve advised, that we’ll finally see that performance hole disappear for good.
Given that 3600MHz RAM is only about 10% more costly than 3200MHz then that’s a no brainer of an upgrade, but the jump above that to 3733MHz can easily cost twice as much again depending on the quantity and size of RAM sticks that you need.
I’d expect memory costs to continue to drop over the coming months as no doubt many firms will now be ramping up 3733MHz production over the coming months. Our own provider was also on the back foot, having already killed off their 3733MHz supplies due to a lack of customer interest before the AMD launch, it’s only now that they are rapidly bringing back old lines and looking to flesh out their ranges to support the popular new platform.
In regards to overclocking the advice that AMD put forward early on appears to be very true with faster memory installed. In initial testing, I overclocked the systems and ran 3200MHz memory and saw some solid gains. With the faster memory, we see the same if not better gains and we can also run the CPU cooler at stock.
I did note that I had both an overclocked chip up and running with 3600MHz RAM and the memory performance hole pretty much disappeared completely, but the system wasn’t stable under heavy loads an there is no way you would want to run that in a production environment.
Indeed, it seems that overclocking is more or less impossible when taking the memory over 3200MHz at this time, although given the performance boost we see with the faster RAM this isn’t a complaint. This might even improve in the future as the BIOSes get optimized and better high-speed memory continues to arrive, but it’s very much something to be aware of if buying a machine at this point in the lifecycle.
One thing that the results have left me wondering, especially with the closing of the gap as we approach the 3733MHz optimum is has this always been the case. 3733MHz didn’t exist when Ryzen generation 1 arrived and I’m not even sure if it was a widely available product when Ryzen 2 launched. Even now it carries a rather hefty cost premium and I have to ponder is this simply a case of the memory market catching up to the Ryzen chipset.. has Ryzen so far simply been ahead of its time?
The last bit of testing I’m going to carry out over the coming week is to retest with the information that we’ve picked up since the first look. It’ll now be running stock clocks with the 3733MHz RAM that is shipping to us now and it’ll be running a none hybrid test version that of a freshly expanded test setup.
The AMD Ryzen 3000 series has been well anticipated and in fact the last time there was this much buzz around a release, it was quite possibly the first Ryzen generation a couple of years back. At that time we saw the platform pull AMD back into the limelight and whilst the results were mixed across many usage scenarios, it was clear the platform certainly had the potential to live up to.
In the interim we’ve of course had the 2000 series, which built upon the gains we’d already seen and continued to close the gap even further. AMD, of course, has continued tweaking the platform throughout this period, acknowledging some of the internal latency issues we also saw in the first round of testing and generally showing positive improvements along the way.
I’m coming to this with a short delay after launch due to a shortage of hardware over the first week. The mainstream reviewers looked to have got their hands on them in the week prior to launch, so there is already a lot of coverage out there regarding the more common applications and the hardware has performed well. The upside of this is that I only managed to put one of the chips across the bench before the launch day AGESA BIOS update surfaced which after applying I also saw a small improvement to performance and so ensured that all testing was done with this in place.
I’m going to be putting the 3600, 3700X & 3900X through their paces here and as normal I’ll be looking to max turbo clock them where I can. This ensures we don’t have a slowest core scenario tipping things into overloading earlier than necessary, but it does mean that on a stock setup you should allow for some variance.
The 3600 for a none X series chip did well, allowing us to take it to a steady 4.2GHz on all of it’s 6 dual threaded cores. The 3700X allowed us to max out all of 8 of its cores at 4.4GHz, and the 3900X managed 4.3GHz whilst sitting around 70 degrees even under maximum load. I managed a 4.4Ghz but not without seeing a huge increase in temperatures alongside it.
Whilst the promotional headlines have been focused on the 7nm die shrink, it should be noted that the entire architecture has received an overhaul in the process with AMD noting a sizable 15% increase in IPC performance. Other notable improvements include further tuning to the internal memory latencies and a sizable increase to the L3 cache, both of which should be beneficial to our performance scores.
For the Ryzen testing, I’m using the X570 Asus Tuf board which was the first of the new range to land in the office. It has been fully updated prior to testing with the latest BIOS and running 3200MHz Corsair LPX memory
With news over the past 12 months of security concerns and various performance affecting patches that have since followed, I’ve set up a new test bench where the Windows 10 build being used is the current 1903 with all drivers being freshly installed. Also given all these changes I’ve benched a number of the Intel chips in this round of testing, with both of the Z390 and X299 boards being fully updated Asus Prime boards.
On top of that reinstall and due to exceeding the benchmarking overhead in the last round with the largest available chips, I’ve made a few modifications to the standard DAWBench tests this time as I suspect that I run the risk of easily surpassing the tests in their default forms.
With the DAWBench DSP test, the SGA1566 plugin now has all instances set to the high-performance setting, running at 24/48 and this gives us plenty of headroom for our needs.
The Kontakt based DAWBench VI test, on the other hand, I would expect to quite quickly outperform with the CPU’s we have here.
I’ve attempted to soak up some of the available performance we have on offer by applying two instances of the SGA1566 plugin in high-performance mode to each of the 16 hidden sine tracks, which on my 9900K testbed took up about 50% of the available performance. This should give us a reasonable baseline to start from and still have the ability to check for any performance affected by latency.
Since the Native KA6 interface has had a generational jump and the older model is now discontinued, I’ve now gone and retired the old testing interface. I’ve now switched it up to an RME Babyface on this round of testing and shall be sticking with that going forward.
Another change is that both sets of tests are being reported using Reaper this time. I completed the testing initially using both Reaper and Cubase, but upon looking over the results I saw an irregularity that sent me back to retest again using Reaper for both sets, more of which I shall cover in the results section.
Having made all those changes, please be aware that these results and prior results are in no way comparable. I’ve changed the following and have retested all CPU’s listed in the results.
OS version. BIOS versions. Reaper version. Cubase removed. Different audio interface. Modified DAWBench versions.
The one benchmark that can still be used to loosely compare is CPU-Z and that is where we shall start.
I’ve used the inbuilt 9900K metric as a baseline comparison. I’ll note that the result they have recorded is within 10% of my last round of testing, so it seems quite fitting to use it here.
So first up DAWBench wise is the classic DSP test, running the SGA1566 variant as covered up top.
This test sees us stacking up plugin instances on a thread by thread by thread basis until the whole CPU hits a breaking point. An impressive result from the sub £200 AMD and the top end 3900X is trading blows with the £1000+ Intel chips. This test is our raw performance test and as hoped the results are impressive.
The second chart we have is the DAWBench VI Kontakt test, which I imagine is going to be the interesting one for most readers following on from prior writes ups.
And interesting it certainly was. The cross core latency we’ve seen in earlier models has gone on the 3600 and we were hitting 95% on the 64 buffer on the 3700X with 100% leveraged on the 128 buffer, both of which are certainly welcome sights.
The 3900X with its new dual chiplet design was the only model to not come away with a completely clean sheet, although given we have an extra die section to deal with, this might not prove to be a huge surprise. I would expect it to mature in much the same fashion as the other chips below it in the range have done over future iterations and no doubt looking forward to them fine-tuning this new design further.
With the first round of testing as noted up top, I used Cubase 10 initially for the DAWBench VI testing and everything looked great right up until the final test. On the 3900X I saw a 20% performance drop at 64 buffer with 90% of the CPU being used at maximum load and 128 gave me 85% – 95% with the full CPU being leveraged at the 256 buffer and above mirroring the low buffer latency we’ve seen in previous generations.
It’s at this point where I wondered if we’d see any difference with a sequencer switch out, it didn’t help in the previous testing, but C10 had a few major changes under the hood over earlier versions and I’m keen to see if any of those have impacted here. I rebuilt the new test in Reaper and took another look at it with Reaper offering differing results. I still saw a performance hole at the 64 and 128 buffers, but this time it was more like 85% (64 buffer) and 95% (128 buffer) of the CPU being leveraged before it started to top out.
So, interesting to note variance drift between sequencers and how efficiently Kontakt appears to be running within them. I did upon seeing this completely re-bench in Reaper and those are the results presented. but do be aware of the sequencer variance that appears to be in play.
It’s also worth noting that some sequencer may not be able to address the full 32 threads efficiently, even if it can currently see them. I can foresee a lot of optimization being required by various DAW coders in order to ensure that their software can still keep up with the new hardware that is currently emerging.
So, overall thoughts are one of being largely impressed at each given price point. I don’t think I’d drop as low as the 3600 personally, but the 3700X has a strong claim as a superb all-rounder at the entry-level and both of these chips seem to have largely shaken any concerns that remain about internal latency handling.
The 3900X has the noted performance latency still, although it seems to vary between applications and we don’t see that occurring with either the Reaper or Cubase test on the Intel side. I wouldn’t normally be happy with seeing anything drop out at 70% or 80% load but there is certainly an argument that it still offers reasonable value as even then it exceeds the 9900K which is currently sat around the same price.
Certainly, anyone working above a 128 buffer has little to no concern there as it appears to recover in full by the 256 buffer.
So there we have it, a great first outing for AMD’s 7nm design. I’ve seen comments aplenty about the lack of overclocking capabilities and yes we’ve come up short of the all-core clock that I was aiming for in two of the tests, but I do kind of expect that from any first-generation chip after a die shrink. I’ve certainly no doubt that we’ll have refinements over the next couple of years that will successfully extract every last bit of performance from the Zen 2 platform.
My only reservation at this time is compatibility with third-party hardware and mainly interfaces. We saw some compatibility issues with Ryzen 1 & 2 with some PCIe sound cards and some USB based interfaces. ASMedia have a bit of a poor rep on Intel board where they’ve provided their third party USB3 solutions as audio devices don’t tend to play too well on them. We saw similar incidents with the implementation they packaged for the Ryzen board on generation one and thankfully it was less common on Ryzen gen 2.
Ryzen Gen 3 has an AMD designed USB implementation but built around an ASMedia package and at this point, I’ve little idea how it will hold up with all the device we have available. I was testing using a Babyface Pro this time around, so that’s validated, but I would certainly check with user groups for your key devices for any compatibility issues prior to buying.
Looking forward, unsurprisingly Intel’s next refresh details have started to leak across various sites. The Cometlake refresh has a 10 core chip and various price reductions being dangled via those leaks, which obviously look to challenge this Ryzen release when they arrive.
Whilst some people might already be rolling their eyes at this leak timing, those who remember back to the last time we were entrenched in some good ole CPU wars, they’ll remember that this is pretty much business as usual and I can see price wars on the horizon as AMD snatches more and more market share.
But that’s all still to come in the future. Right now, for the time being, the third iteration AMD Ryzen series is easily their most compelling offering yet.
Coffee Lake has been with us now for just over a year and it’s been a rather turbulent period for Intel. AMD’s continued gains over the last 12 – 18 months has marked a change in the marketplace and the first generation Coffee Lake launch perhaps felt a little rushed last time around, especially as Intel was attempted to respond to the opening volley in the now ongoing CPU wars.
This time around I find myself looking over the selection of chips in front of me and the key question on my mind right now is one of “have they managed to extract the platforms potential this time around?”
So, I’ve got 3 different models here all new to the Intel mid-range:
1. The new flagship in the form of the 8 core + Hyper-threading i9 9900K running at 3.6 with a turbo clock of 5GHz out of the (oddly) shaped box.
Chip is being run at all core 5GHz
2. The i7 9700K featuring 8 cores but no Hyper-Threading. The chip is clocked to 3.6GHz and 4.9GHz out of it’s rather more normally shaped box.
Chip is being run at all core 4.9GHz
3. Lastly the 9600K in another boring box. 6 cores, no Hyper-Threading and 3.7GHz with 4.6GHz on the turbo.
Chip is being run at all core 4.6GHz
So, we see some firsts here and some repositioning in the range. The i9’s go mainstream and in this case, we’re seeing a few notable key differences there. The big one is that it’s the first time we’ve seen Intel put out an 8 core mainstream chip. Given we only got our first mainstream 6 core back on the last range refresh, it’s good to see them again being pushed into cramming more value onto the die this time around.
The i9’s are also promising us solder under the heat-spreader this time around, rather than the paste found in models elsewhere in the range, so this should in theory help with overclocking for those wishing to push them a bit more.
The i7 & i5 models this time around are limited to 8 cores and 6 cores respectively with no hyper-threading. Whilst it helps to differentiate between the respective ranges, it is going to come as a bit of a shock to anyone used to the current i5/i7 naming convention. On first thought, we wondered it this meant that we could expect the new 8 core with no HT to be outperformed by the older 6 core + HT models or not, although this could very well come down to specific workloads.
Hyper-Threading by its very nature is based around stealing unused clock cycles to get more work done, so if your workload is already thrashing the CPU, then having Hyper-Threading isn’t really going to have much of an impact. In previous testing I’ve tended to note anywhere between 20% and 60% gains with it turned on depending upon the software in use, so it could be argued that having an extra 2 real cores, could equate to somewhere in the region of 4 or even more lost Hyper-Threads (once again, workload permitting) and we’ve also got to consider clock and IPC gains here, so playing off the 9600K & 9500K’s against their predecessors are going to be certainly interesting.
So lets get down to it.
All the standard tests to start with and nothing unusual going on so far. Whilst they are all clocked fairly close together as far as the cores go, you can note differing amounts of L3 cache on each of the chips, which is no doubt going to help a little in both the single and multi-core benchmarks.
So on with the DAWBench SGA DSP Test and we can see the 3 new chips in Yellow above. Starting with the 9600K the obvious comparison here is against its predecessor and frankly, it’s a little underwhelming with a somewhere between a 1% – 10% increase depending upon the buffer in play and scaling upwards as the buffer size is increased.
The 9700K is next and we get to compare its new design configuration of 8 true cores and no Hyper-threading, which also appears to come off poorly here when compared against the older 8700K with the results showing up a 20% – 40% drop off against Intel’s own previous generation class leader.
The loss of Hyper-threading here really looks to have impacted the testing on the new generation at least under the DAWBench classic test. I do get the thought process here with the chip design itself, as the largest new segment in recent years that seems to have captured the marketing teams imagination has been the rise in content creation users who are live streaming. True cores for that sort of content generation is far more beneficial, especially gamers who wish to live stream at the same time, so I fully understand this design choice, in fact it could be argued that this style of chip would be preferential for anyone working live but for anyone looking for raw performance in the studio it’s all a bit disappointing so far.
The flagship here, however, is no longer the i7 model, but rather the i9 9900K and it’s at least here where things are making rather more sense. It’s the first time that we’ve seen an 8 core in Intel’s mid-range line up and looking at the result above, it looks to have settled itself just above the 7820X from the Intel Enthusiast range (X299) and to be fair, on paper at least it makes perfect sense that it would replace that chip.
It’s the same core count, a few generations newer and clocked higher, so it was always going to be a contender, what it does mean, however, is that once again we see one of Intel’s mid-range chips start to cannibalize their own enthusiast class of chips. In fact, we’ve now reached the point where the lower end i7 enthusiast class has had a dearth of releases over the last 15 months and largely been killed off, wherein the same period AMD has successfully taken a sizable bite out of that part of the market space too and we see them continue to take advantage of Intel’s lack of new competing models.
Indeed, in the chart here sat above it, we see the large core count AMD’s as well as the older generation i9’s outlining exactly what this test is good at, which is small files being spread efficiently over the all the available processing space and honestly, the results here once again don’t really give us any surprises as to how and where the chips are being positioned in the range.
Switching over to the DAWBench VI Kontakt based test we see a more interesting picture as the higher single core clocks appear to give us a welcome boost here. In the one thing, it does really outline for us here is that the Kontakt handling looks to benefit from IPC figures all around.
Having the dedicated cores looks to help when working at tighter ASIO buffer settings on both the 9600K and 9700K, although we can see that this benefit disappears on the 9700K once we slacken that setting off to around the 256 buffer. It appears at this point that the Hyper-threading on the older 8700K finally gets a bit of room to breath and flex it’s stuff once you open up the buffer far enough and this in itself is interesting information.
Thinking about this from a live point of view where you’re aiming for the tightest RTL score and quite likely to be making use of Rompler style libraries, this does outline that going with these new chips that feature all real cores might well pay off for you in this situation. However, if you’re working in the studio, the loss performance at the larger buffer settings, at least in comparison with the older generation might once again prove a little perplexing.
Taking a look at the i9 9900K by comparison and it starts to make more sense again, with it doing rather a good job at once more making the older 7820X chip irrelevant. There is less challenge up this end of the chart from the red team largely due to the lack of solid benchmarks obtained in the last round which you can catch up on if you hit the link.
What this means is that the options here do seem to be becoming even more divided. It’s been pointed out that the higher latency jobs that the Zen chips were excelling at are applicable to all sorts of media editors still and with each additional chip it becomes ever more clear that these continue to remain very scenario dependent, and that Kontakts way of working tends to favour highly clocked cores and larger IPC figures over the workload being spread out over more numerous but slower cores.
Before I round up I just want to throw out a couple of additional charts. I didn’t get a chance to do it with all of them, but I did record the i9 9900K at both stock and at the all core overclock, largely so you can see the difference it can make by setting it to the all core turbo.
Depending on the test and buffer size it’s up to around 8% in these benchmarks, although this can grow as you use more complex chains of processing in your projects. A chip is only really as strong as it’s the weakest core, as once you max out any given core you begin to run the risk of audio artefacts creeping in.
I mention this specifically with the i9 9900K as a lot of premium boards have been shipping with 5GHz profiles now for a few years and it’s rather easy to hit the results I’m showing above with a halfway decent cooler solution. Above that, you’ll probably want to move to a water cooler solutions with 5.2GHz looking to be the target for anyone wanting to really drive it.
I’ll also note that the i7 9700K was running comfortably just below 80 degrees by the time I all core turbo’d it, whereas the i5 9600K was sitting nicely around the 60 degrees mark even with Prime 95 absolutely thrashing it, so I reckon for anyone wanting true cores only, you might have quite a chunk of headroom there to play with if you want to tinker with it.
So, overall, what are my final thoughts?
The i5 9600K and i7 9700K both feel like a step backwards for our part of the market to a degree. Sure, they have some strengths and I’ll come back to the example of low latency machines for live use again being a prospective user base, but their value proposition in comparison to other chips already out there is where it really falls over in the studio.
Having a sideways move in the overall performance is a little disappointing but we’re seeing an initial street price on the i7 9600K of around £350 against the i5 8600K historical showing of around £250. Similarly the 8700K was around £350 for most of its lifecycle and the 9700K sits at £499 at launch, so we’re seeing price increases with each of those ranges, although I suspect as supply catches up with the initial demand we may find some price realignment over the coming months and I wouldn’t be all that surprised to see the new chips reflect older price points once the market stabilises. This is a fairly common occurrence with any new chip release, but admittedly it leaves me feeling a rather underwhelmed given all I’ve discussed already from a performance point of view.
The i9 9900K, on the other hand, replaces the 7820X which spent most of its lifecycle between £400 – £500 in the UK and the i9 9900K has landed at £599. Assuming it’s going to drift over the coming months we’re still essentially looking at £100 mark up over the older model.
The DAWBench classic test here shows us mixed gains depending upon the workload and it’s up against the AMD’s which manage to still outperform it within this test. By contrast, the DAWBench VI test flips it with it outperforming the chips on the chart and keeping in mind the Threadripper results previously.
So, does even the i9 9900K make sense? Well, yes, it’s the one that really does here. With the change to the Z390 platform, we see a cost saving over the older X299 platform complete with a more advanced feature set. With the cost differences between boards often totalling and surpassing the £100 amount, the overall cost of going with an i9 9900K over an i7 7820X looks to come out in the i9’s favour and that’s before considering the performance gains it offers.
The additional good news here is that the other previous sticking point with the Z390 platform for some users is it’s restricted memory capabilities, as the four slots could only handle a maximum of 64GB. We’ve seen an announcement recently however that they are going to start offering double stacked DIMM’s over the coming months to support this platform, so hopefully, it shouldn’t be all that long until these boards can handle 128GB as well.
Overall this feels like Intel’s real response to AMD’s advances last year although given the swift execution and release of the second generation Zen chips, perhaps they are still a tad on the backfoot here. It’s kinda where Coffee Lake should have been last time around and it’s of course good to see more power in the mid-range. It does leave me questioning where exactly it’s going to leave the enthusiast class, as anything less than an i9 on that platform is going to prove to be poor value at this point and given the age of that platform I really can’t help but hope that the next Intel enthusiast platform can’t be all that far off now.
It feels like this is the repostioning that Intel needed to happen to put it’s own range back into some context, but it may not prove to be the change that everyone was looking for, at least in our small corner of the market.
At the very least here the i9 9900K emerges as a rather strong contender for us audio users and I suspect any other i9 based refresh over the coming months is going to make this all make a whole load more sense when the dust settles. But with AMD already promising updates to its own platform and announced tweaks for their memory balancing promised over the next few weeks Intel may have to work even harder over the coming months.
I’m the first to admit that I’m a little late to the table with this write-up. The original 2990WX sample arrived whilst I was on leave and was quickly placed into a video rig and sent out for review, meaning I’ve had to locate another one at a later date. Along with that, I’m honestly a little overwhelmed with how much interest this £1700 workstation grade CPU has generated with the public in recent weeks, as I really didn’t expect this level of interest in a chip at this sort of price point.
I’ve also approached this with a little trepidation due to earlier testing. As someone noted over the GS forum, the 2990WX might not prove all that interesting for audio due to the design layout of the cores and the limitations we’ve seen previously with memory addressing inside of studio-based systems. They were certainly right there, as the first generation failed to blow me away and there remains a number of reservations I have with the under-laying design of this technology that potentially could be amplified by this new release. During the initial testing of the 2990WX this time around, the 1950X replacement also arrived with us too in the shape of the 2950X and given some of the results of the 2990WX I thought throwing it into the mix might prove a handy comparison.
Why bring all this up at all? Well, because everything I discussed back then is still completely relevant. In fact, I’m going to go as far as to suggest that anyone doesn’t understand what I’m referring to at this point should head over to last years 1950X coverage and bring themselves up to speed before venturing forward any further.
Back again? Up to speed?
Then I shall begin.
The 2990WX is the new flagship within the AMD consumer range and features a 32 core / 64 thread design. It has a base clock of 3GHz with a max twin core turbo of 4.2GHz and an advised power draw of 250W TDP.
I won’t split hairs. It’s a beast… something I’m sure most people reading this are well aware of given the past week or so’s publicity.
In fact, for offline rendering, I could close the article right there. If you’re a video editor on this page and don’t happen to care about audio (hello… you might be lost, but welcome regardless) then you should feel secure in picking up one of these right now if you have the resources and the need for more power in your workshop.
But as was proven with the release of the 1950X, the requirements for a smooth running audio PC for a lot of users are largely pinned on how great it is for real-time rendering, which is a whole different ballgame.
In the 1950X article I linked up top, I went into a great deal of detail in regards to where performance holes existed. I found that low latency response was sluggish and resulted in a loss of performance overhead that left it not in an ideal place for audio orientated systems. I had a theory that NUMA load optimization for offline workloads was leaving the whole setup in a not ideal situation for real-time workloads like ASIO based audio handling.
In the weeks following that article, we saw AMD release BIOS updates and application tweaks to try and resolve the NUMA addressing latency I had discussed in the original article, largely to no avail as far as the average audio user was concerned. In AMD’s defence, they were optimizing it further for tasks that didn’t include the sort of demands that real-time audio places upon it, so whilst I understand the improvements were successful in the markets they were designed to help, few of those happened to be audio-centric.
At the time it was just a theory, but my conclusion was largely one being that if this is as integral to the design as I thought it might be, then it would take a whole architecture redesign to reduce the latency that was occurring to levels that would keep us rather demanding pro users happy.
The 2990WX we see here today is not the architecture change we would require for that to happen as where the 1950X has 2 dies in one chip, the 2990WX is now running a 4 die configuration which has the potential to amplify any previous design choices. If I was right about hard NUMA being the root of the lag in the first generation then on paper it looks like we can expect this to only get worse this time around due to the extra data paths and potential extra distance the internal data routing might have to cope with.
The 2950X, by comparison, is an update to the older 1950X and maintains 2 functional dies, with tweaks to the chip’s performance. Given the similar architecture, I would expect this to perform similar to the older chip, although make gains from the process refinements and tweaks enacted within this newer model. I’ll note that the all core overclocking is improved this time around and a stable 4GHz was quick and easy to achieve.
OK, so let’s run through the standard benchmarking and see what’s going on.
As normal I’ve locked it off at an all core turbos on both of the chips. As with a lot of these higher core count chips, I’ve not managed to hit a stable all core max turbo clock, which would have been 4.2GHz, rather settling for 3.8GHz on the 2990WX and 4GHz on the 2950X both of which perform fine with aircooling.
I’ve spoken to our video team about this and they managed to hit a stable 4.1GHz on the 2990WX using a Corsair H100, so it looks like you can eak out a bit more performance if noise is less of a consideration in your environment.
If you’re not aware from previous coverage why I do this, if you’re running a turbo with a large spread between the max and minimum clock speeds then the problem with real-time audio is that when 1 core falls over, they all fall over. So, whilst you might have 2 cores running at 4.2GHz the moment one of the cores still running at 3.2GHz fails to keep up then the whole lot will come tumbling down with it. Locking cores off will give you a smoother operating experience overall and I’m always keen to find a stable level of performance like this when doing this sort of testing.
I don’t always remember to run this benchmark, although this time I’ve made the effort as Geekbench doesn’t appear to support this many cores at this point. Handily enough, I did at least run this over the 1950X last time which returned results of 428 on the single core and 9209 on the multi-core at the time.
Given that the 2990WX looks to be pulling twice the performance and physically has twice the number of cores, it looks to all be scaling rather well at this point. The 2950X, on the other hand, sees around a 10%-15% gain on the single and multi-core scores over the previous generation.
Moving onwards and the first test result here is the SGA DAWBench DSP test.
This initial test is very promising, as was the older 1950X testing. Raw performance wise we’re talking about it by the bucket, I really can’t stress that enough with both chips performing well in what is essentially a very CPU-centric test.
At the lowest buffer we see it being exceeded by the older chip, so what is going on there? Well, we’re seeing a repeat in the pattern that was exhibited by the 1950X where there is an impact to performance at tighter buffers, and it does appear that at the very tightest buffer setting that we’re seeing some additional inefficiency caused by the additional dies, although this does resolve itself when we move up a buffer setting.
Last time we scaled up from 70% load being accessible at a 64 buffer and this time, I imagine due to the extra dies being used we see the lowest setting corrupting around the 65% load level and then scaling up by 10% every time we double the buffer.
As a note when I pulled that 512 buffer result this time around and it returned 529 instances.
The 2950X, by comparison, returned me a load handling around the 85% on a 64 buffer, rising to 95% at a 256. An improvement on the first look we took a look at the original 1950X chip, although I’ll note I was also seeing this improved handling when I did the 1950X retest a few months ago using the newer SGA1156 charts that has replaced the classic DSP test, so this might be down to the change in benchmarks over the last year, or it could also be down to the BIOS level changes they’ve made since original generation launch.
So far, so reasonable. A lot of users, even those with the most demanding of latency requirements can get away with a 128 buffer on the better audio interfaces and the performance levels seen at a 128 buffer, at least in this test are easily the highest single chip results that I’ve seen so far.
In fact, knowing we’re losing 40% of the overhead on the 2990WX is really frustrating when you understand the sort of performance that we could be seeing otherwise. But even with that in mind, if you wanted to go all out and grab the most powerful option that you can, then wouldn’t this still make sense?
Well, that test is pure CPU performance and in the 1950X testing, the irregularities started to really manifest themselves in the DAWBench Kontakt test where it started to depend equally on the memory addressing side of things.
Normally I would insert a chart here to show how that testing panned out.
But I can’t.
It started off pretty well. I fired it up with a 64 buffer and started adding load to the project. I made it up to around 70% CPU load on the first attempt before the whole project collapsed on me and started to overload. I slackened it off by muting the tracks and took it back down to around 35% load where it stabilised, but from this point onwards I couldn’t take it above 35% without it overloading, not until I restarted the project.
I then tried again at each buffer setting up to 512 and it repeated the pattern each time.
I proceeded to talk this one through with Vin the creator of the various DAWBench suites and a number of other ideas were kicked about, some of which I’ve dived further into.
One line of thought was that as I was still using Cubase and the last 8.5 build specifically, precisely for the reason that C9 has a load balance problem for high core count CPU’s that is currently being worked upon. The older C8.5 build is noted as not having the same issue manifest due to a difference in the engine and during testing this time Windows itself was showing a fairly balanced loads mapped across all of the cores whilst I was looking at performance meter, but even so, historically, exceeding 32 cores has always been questionable inside many of the DAW clients.
So, to counter this concern, I went and ran the same tests under Reaper and saw much the same result. I could push projects to maybe 65%-70% and then it would distort the audio as the chip overloaded and this wouldn’t resolve itself until the sequencer was closed and reloaded.
So what is going on there? If I was to speculate, then the NUMA memory addressing is designed to allocate the nearest RAM channel to it’s nearest physical core and not to use other RAM channels until on core’s local channel is full.
I suspect with knowing that, that the outcome here is that it’s maintaining the optimal handling up until that 70% level and then once it figures out that the RAM channel is overloaded it starts allocating data on the fly as it sees fit. The reallocation of that data to one of the other 3 dies would result in it being buffered and then allocated to the secondary memory location and would result in additional latency when the data is recalled in a later buffer cycle which would result in audio being lost when the buffer cycle completes before it can be recalled.
In short, we’re seeing the same outcome as the first generation 1950X but amplified by the additional resources that now need to be managed.
This way of working is the whole point of hard NUMA addressing and indeed is the optimal design for most workstation workloads where multiple chips (or die clusters in this case) need to be managed. It’s a superb way for dealing with optimization for many workloads from database servers through to off-line render farms, but for anything requiring super-tight real-time memory allocation handling it remains a poor way of doing things.
As I’ve said previously, this is nothing new for anyone who deals with multi-CPU workstations where NUMA management has been a topic of interest to designers for decades now. There has always been a performance hit for dealing with multiple CPU’s in our type of workflow and it’s largely why I’ve always shy’d away from multiple chip Xeon based systems as they too exhibit this to a certain extent.
Much like the first generation 1950X with it’s 2 dies, we see similar memory addressing latency when we use 2 seperate Xeons and this has always been the case. I would never use 4 of those together in a cluster for this sort of work simply due to that latency and so the overall outcome with 4 dies being used in this fashion isn’t all that surprising.
I also tried retesting with SMT turned off, so it could only access the 32 physical cores in order to rule out a multi-threading problem. The CPU usage didn’t quite double at each buffer instead settling around the 70% total usage mark but the total amount of usable tracks remained the same and once again going over this lead to the audio collapsing quite rapidly.
So, much like the first generation the handling of VST instruments and especially those which are memory heavy look like they may not be the best sort of workload for this arrangement. This ultimately remains a shame, especially as one of the other great concerns from last time which was heat has been addressed by quite some degree. Running the 2990WX even with an overclock didn’t really see it get much above 70 degrees and that was on air. Given that the advised TDP here is 250W at stock, rising quickly when overclocked even to the point of doubling the power draw, the temperatures for a core count this huge is rather impressive. I think there is a lot to pay attention too here by Intel in regards to thermals and the news that the forthcoming i9’s are finally going to be soldered again, makes a whole load of sense given what we’ve seen here with the AMD solutions. If anything it’s just a shame it took the competition pulling this out of the hat before they took notice of all the requests for it to be brought back by their own customers over recent years.
Still, that’s the great thing about a competitive marketplace and very much what we like to see. Going forward I don’t really see these performance quirks changing within the Threadripper range, much the same way that I never expect it to change within the Xeon ecosystem. Both chip ranges are designed for certain tasks and optimized in certain ways, which ultimately makes them largely unsuitable for low latency audio work, no matter how much they exceed in other segments.
There is some argument here for users who may not require ultra-tight real-time performance. It’s been brought to my attention in the past that users like mastering guys could have a lot of scope for using the performance available here and if they are doing video production work too, well, that only strengthens the argument.
On paper that all makes sense and although I haven’t tested along those lines specifically, the results seem to indicate that even the trickiest of loads for these CPU’s seem to stabilise at 512 and above with 80%+ of the CPU being accessed, even in the worst case scenario. I have to wonder how it would stand up in mixed media scenarios although I would hope that ultimately in any situation where you render it offline that you should be able to leverage the maximum overhead from these chips.
I suspect the other upshot of this testing might be one of revisiting the total CPU core count that each DAW package can access these days. Last time I did a group test was about half a decade ago and certainly, all the packages look to have up’d their game since then. Even so, I doubt anyone working on a sequencer engine even 3 years ago would have envisioned a core count such as the one offered by the 2950X here, let along the monstrous core count found in the 2990WX.
AMD’s Zen core IPC gains this generation as we’ve already seen with Ryzen refresh earlier in the year were around the 12% mark and it looks to have translated faithfully into Threadripper series with the 2950X model. One of AMD’s big shouting points at launch was regarding just how scalable the Zen package was simply by upping the die count and that’s clear by the raw performance offered by the 2990WX, they really have proven just how effective this platform can be when dealing with workloads it’s designed for.
One day I just hope they manage to find a way of making it applicable to the more demanding of us studio users too.
It’s been a while now since we sat down and took a good look at any of the mobile processor releases. It’s a market segment that has been crawling along slowly in recent years with minor incremental upgrades and having checked out the last couple of mobile flagship chips, it was obvious that with each generation we were seeing those refinements focused more on improved power handling rather than trying to extract every drop of performance.
Admittedly in the shape of last years 7700HQ they perhaps got closer to the equivalent desktop model than any generation previously managed to achieve in previous years. Whilst welcome, this was really more a symptom of stagnating desktop speeds, rather than any miraculous explosion in mobile power. Whilst the chip itself was a great performer, the fact that it got there by eaking a few percent generation, upon generation… well, by the time we got there, it was all ultimately a little underwhelming.
But now, thanks to AMD’s continued push in the current desktop CPU war, we’ve seen Coffee Lake emerge from the blue camp and now we’re going to get hands-on with the mobile equivalent.
The i7 8750H we have here today is a 6 core with hyperthreading, running with a base clock of 2.20GHz and a max single core turbo frequency of 4.10GHz and leads the way when it comes to mobile i7’s.
Just as a side note before we kick this off, there is another chip above this, in the form of the i9 8950HK which is also 6 cores + hyperthreading but with another 500MHz on the clock. I mention this as Apple has just announced it’s going into the flagship Macbook later in the year, we do have them due to land with us in PC laptops as well in a month or two, so I will be benchmarking that when it arrives with us too.
Already in the very first screenshot above, we’ve inadvertently tipped a nod to what’s going to be the crux of this write-up. The clock speeds are somewhat wide-ranging, to say the least. On paper, there is almost 2GHz worth of clock between the base and turbo clocks. Keeping in mind that it’s single core turbo only up to the 4.1GHz and suddenly you find yourself asking about what the rest of the cores will be doing at that point.
Quickly throwing CPUid on and running it returns us a result of 3890GHz, which if it had been all cores would have been rather impressive for a mobile chip. In this instance, however, I wasn’t doing anything other than sitting on the desktop when this snapshot was taken. The score you see is the highest core score and it’s hyper-thread was showing as matching it.
The rest of the cores, however, well, they were largely unused and sat around the baseline 2.0GHz – 2.6GHz level. What we really want to know of course is what sort of average speed we can expect from all the cores being kicked up to 100% load.
Any longer term followers of these pieces will already be well aware that my preference for testing involves doing an all core overclock or in more basic terms, I tend to favour locking all the cores to the single core max turbo speed.
Yes, it’s an overclock, but it’s one that the chips are kind of rated to. Admittedly, it’s not rated to quite the level we’re working at here, but hey… that’s why we favour some chunky aftermarket cooling in those systems to make everything alright.
Except, when dealing with laptops we can’t go strapping a large chunk of copper to it, in fact, a lot of the tweaks we would wish to make on a desktop system, simply don’t exist in laptop land. Often with laptops, it’s a case of a unit either working out of the box or with a few basic tweaks or otherwise due to drivers or hardware choices it’ll never really be suitable for the sort of real-time processing required for working with audio.
I grabbed a copy of AIDA64 and gave it a quick run, at least enough to force the CPU to load up all the cores and simulate a heavy workload and how those cores would respond to such a load.
What we see here is all the cores being pushed, with the highest speed core running about 3000MHz in the screenshot. Monitoring it in real-time it was bouncing around 3000 – 3200MHz range. Similarly, at the lower end, we see a core sat around 2600MHz and this would bounce up to around 2800MHz at times.
So, where’s our 4.1GHz turbo? Well, that single core turbo only really achieves such lofty heights if the rest of the cores are sat around doing nothing. In the interest of load balancing and heat management should more than a couple of cores need to be turbo’d then all of them will shift to a safer average.
You see on desktops with chips that have a range of a 3.8Ghz to 4.3Ghz sitting mostly around the 4GHz level and is why I tend to notch them all up to 4.3GHz in that sort of situation. It ensures no sudden ramping up and down and ensures we get some nice stable but optimized performance out of a setup without taking any major risks.
With these laptops, we don’t get those sort of options, nor I suspect would heat permit us to be quite so aggressive with the settings. Whilst the headline here of 6 cores is fairly unprecedented within a consumer level laptop, and certainly, on a fairly mainstream chipset, it’s a little bit smoke and mirrors with how it’s presented if you don’t fully understand how the turbo presents itself.
The potential issues it presents to us are in the form of the ASIO buffer. With whole channels being assigned to each given thread, we ideally want the performance level across all cores to be as equal as possible. For audio systems the overall performance can often be limited by how powerful the weakest core is, this is something we need to keep in mind heading into this results roundup.
With the DAWBench DSP test, we’re using the SGA1566 variant running under Reaper for this generation of testing and we see the 8750H performing around the level of an entry-level desktop i5 chip. In comparison to previous generations, this isn’t overly surprising as historically the mobile i7 CPU of any given generation tends to sit around the level of the leading i5 desktop solution in the performance stakes.
Running the DAWBench Vi test we see similar results here too, with the chip coming in just behind the i5 8400 once again. It’s a reasonable showing and in reality, we’re probably looking at maybe a 25% gain over the last generation flagship mobile chip.
Given that we’ve seen 3 or 4 generations now where 10% gains year on year has been the standard then normally we’d be pretty happy about seeing a jump of 25% coming out of single refresh and indeed it’s certainly a far better value option than the model it replaced.
However, we saw a jump of 40% on the desktop last year and frankly all we’re doing here is shoehorning in another couple of cores, rather than bringing in a whole new platform. It looks like they’ve played it cautiously by not pushing the chip too much and the temperatures do seem a little on the safe side even under stress testing.
To be fair to them, this is pretty much what the average user wants from a laptop chip, giving us quick bursts to deal with any sudden intensive activity, but otherwise, aggressive power-saving to ensure a long battery life when on the move.
Which of course, is pretty much the opposite of what most of us power users want, as we tend to be looking for a high-performance desktop replacement solution. It’s clear there is a bit of headroom here which will no doubt be leveraged over the next couple of range refreshes, it’s just a little bit frustrating that we can’t extract a bit more of it right now ourselves.
With all that said I suspect that after seeing the CPU war kick expectations up a notch as it did last year, that I may have headed into this with slightly higher expectations than normal this time around.
Overall, the final result here is a solid release with above average generational gains that I’m sure will be more than appreciated by anyone who is in the market for a new model this year.
Looking back over the rather hectic first few months of 2018 in the PC industry, it’s clear that a lot has changed since the last CPU benchmark session late last year. In the space of 6 months, we’ve seen security concerns and the resulting software patches swing windows performance back and forth as they’ve arrived with us thick and fast. I’ve largely been trying to wait it out and see how the dust settles in the interim, but with the release of new hardware, it’s time to get back into it.
My last bench was based on a build of windows frozen in late 2016 and associated drivers have gone through a number of revisions during the time since, so with the launch of Ryzen 2 it’s very much the time for an all-new software bench to be set up.
Cubase has moved from 8.0 to 9.5 and Reaper too has advanced a number of builds to 5.79 at the point of testing being initiated. This time around we also see the introduction of the newer SGA build of the DSP test, replacing the older DAWBench DSP test and the latest build of the DAWBench Vi test too.
Before getting underway please note that the new results are in no way comparable to the older charts, other than looking at the rough performance curve differences between certain chips which do appear to be in line with prior results. They are certainly not directly value comparable with all the bench changes that have taken place and it’s always key to keep the playing field as level as possible when doing these comparisons.
This time around I’ve tried to run each chip at its turbo frequency across all cores once again. Modern chips will tend to be rated with both a stock clock and a turbo clock, although what isn’t always clear is that the max turbo rating is often only over 1 or 2 cores by default.
Historically it’s been relatively easy to run most CPUs with those cores being pushed and locked off at the turbo max. However, in the event of a platform being pushed too hard, then this isn’t always viable. For instance, I saw this in testing some of the higher end i9’s, where I would choose to all core at 4.1GHz, rather than leave it at stock and let it 2 core to 4.2GHz with a far lower average leaving me open to possible audio interruptions due to clocking.
It’s also the case here with the 2700X where the overclock would hang the machine if trying to push everything to the 4.2GHz rated turbo speed. Instead, I tried to clock it up both manually and using the AMD tool, both of which topped out around 4.1GHz. After speaking to my gaming team and realising this is fairly common (a number of other reviews have picked up on it as well) I ended up using the utility to set everything up with the slightly lower all core turbo at 4.1GHz and testing there.
The 2700X here slots in behind the 8700K which leads by just short of 20% extra overhead at the tightest buffer setting, and both chips look to scale upwards in a similar pattern as you increase the buffer setting. The 8700K seems to be the most suitable comparison here as the price point (at time of writing in the UK) is around £30 more or about 10% more than the cost of the 2700X at launch.
The story of the performance curve scaling looks to repeat when we come to examine the 2600X and by comparison the 8600K from Intel. However, this time around the results are reversed with the Intel chip lagging behind the AMD model by about 5% across the buffer settings whilst the AMD costs around £25 less which makes it roughly 12% cheaper at launch.
So a strong showing for the DSP test, where we’re mostly throwing a load of small VST plugs at the CPU. The other test we run here is the DAWBench Vi test, based on stacking up Kontakt instances which allows us to test the memory response through sample loading along the CPU as we see with the DSP test.
With the Gen1 Ryzens, we saw them perform worse here overall, we suspect down to the memory response and performance. AMD saw similar performance issues across various segments with certain core software ranging from gaming to video processing and the was a lot of noise and multiple attempts to improve this over the life cycle of the chip. One suggestion we saw pay off to some extent in other segments (once again, video and gaming made notable gains) was to move over to using faster memory speeds.
We didn’t see any improvement here for audio applications, although in this instance all testing (both Intel and AMD) has been carried out with 3200MHz RAM, in the interest of trying to maximize the performance where we can as well as keeping things level in that regard.
The headline figure this time around suggests a rough 10% improvement to the IPC (instruction per clock) scores, which of course is promising, although notably, this is where AMD was lagging behind Intel even after bringing Ryzen to the market. In the interim we’ve seen the Coffee Lake launch, which also improved Intel’s IPC scores meaning that whilst AMD has been catching up rapidly of late, Intel does seem to remain intent on clawing back the lead on each successive launch.
So looking it over this time, both the 2700X and 2600X look to fall behind their Intel comparable chips. The 2600X is roughly 20% lower than the 8600K this time although it’s moving up to the 2700X that proves more interesting, if only because it helps to outline what’s occurred between the two generation releases.
The older 1800X stood up well against the old 7700K edition at its launch, and indeed that extra 10% IPC boost we see this time may well have given it a solid lead over the Intel, if not for the Coffee Lake release in the interim in the shape of 8700K which pulls off a convincing lead at this price point currently. Indeed, not only does the 8700K show gains over the previous 7700K chip, but it also overtakes the more expensive although admittedly older, entry-level 6 core 7800X on the Intel’s own enthusiast platform.
The 2700X is comparable to the 7800X at a far keener price point, although as noted the 7800X more or at least exists as a bit of an oddity by this point, even within it’s own range, so whilst this might have been a more impressive comparison 12 months ago, now it feels like they may have landed it just a few months too late to make serious waves.
Speaking from an audio point of view, the chips are good, but not exactly groundbreaking. If you also work in another segment where the AMD’s are known to have strengths, then the good news here is that they offer reasonable bang per buck for audio and hold their ground well as far as giving you performance at those price points.
But once again, they don’t appear to be breaking any performance to cost records overall at least for the audio market. They’ve got solid gains, but then again so has Intel last time around and this is often how it goes with CPU’s when we have the firms battling it out for market share. Not that this is a bad thing, certainly it benefits the end user, whichever your choice of platform.
As a closing note, I saw in my early generation 1 testing a number of interfaces fail to enumerate on the AMD boards. I reported this to a few manufacturers and interestingly the device that first showed up problems on the X370 boards the first time around (in this instance a UAD Twin USB), is behaving superbly on the X470 platform.
Whilst this is a sample size of approximately “1” unit in a range, it does point towards a reconsidering of the USB subsystem this time around, which can only be a positive. Anyone who was perhaps considering this the Ryzen 1 platform, but found themselves out of luck with interface compatibility, might well fare far better this time around. Obviously, if the were problems known before then please do check with the manufacturers your considering for the latest compatibility notes in each instance.
Looking forward there is a rumoured 2800X flagship Ryzen which is already well discussed but as yet no release date on the horizon. The has been already been discussion, rumours and even some testing and validation leaks out in the wild that suggest that Intel might be sitting on an 8 core Coffee Lake. It would certainly make sense for them to be keeping such a chip in the wings waiting on them seeing the public reaction to these new AMD chips. Similarly, it might turn out that the 2800X will be held back as an answer for those rumoured Intel models should they suddenly appear on the market in the near future.
To wrap it up, essentially we’re in peak rumour season and I’ve no doubt we’ll continue to see a pattern of one-upmanship for the foreseeable future which continues to be a very positive thing indeed. If you need to buy a system today, then the charts should help guide you, although if you’re not in rush right now, I’m sure the will be some interesting hardware to also consider coming over the year ahead.
Today we have a few more models from the Intel i9 range on the desk in the shape of the 14 core 7940X and the 7960X. I was hopeful that the 18 core would be joining them as well this time around, but currently, another team here have their hands on it so it may prove to be a few weeks more until I get a chance to sit down and test that one.
Now I’m not too disappointed about this as for me and possibly the more regular readers of my musings, the 16 core we have on the desk today already is threatening to be the upper ceiling for effective audio use.
The reason for this is that I’ve yet to knowingly come across a sequencer that can address more than 32 threads effectively for audio handling under ASIO. These chips offer 28 and 32 threads respectively as they are hyper-threaded, so unless something has changed at a software level that I’ve missed (and please contact me if so), then I suspect at this time the 16 core chip may well be well placed to max the current generation of sequencers.
Of course, when I get a moment and access to the larger chip, I’ll give it a proper look over to examine this in more depth, but for the time being on with the show!
Both chips this time around are advising a 165W TDP figure, which is up from the 140W TDP quoted back on the 7920X we looked at a month or two back. The TDP figure itself is supposed to be an estimate of the power usage under regular workloads, rather than peak performance under load. This helps to explain how a 14 core and 16 core chip can both share the same TDP rating, as the 14 core has a higher base clock than the 16 core to compensate. So in this instance, it appears that they have to some degree picked the TDP and worked backward to establish the highest performing, clocks at that given power profile point.
Once the system itself starts to push the turbo, or when you start to overclock the chip the power draw will start to rise quite rapidly. In this instance, I’m working with my normal air cooler of choice for this sort of system in the shape of the BeQuiet Dark Rock Pro 3 which is rated at 250W TDP. Water-loop coolers or air coolers with more aggressive fan profiles will be able to take this further, but as is always a concern for studio users we have to consider the balancing of noise and performance too.
Much like the 7920X, we looked at previously, the chips are both rated to a 4.2GHz max two core turbo, with staggered clocks running slower on the other cores. I took a shot at running all cores at 4.2GHz but like the 7920X before it we could only hit that on a couple of cores before heat throttling would pull them back again.
Just like the 7920X again however if we pull both of these chips back by 100MHz per core (in this instance both to 4.1GHz) they prove to be stable over hours of stress testing and certainly within the temp limits we like to see here, so with that in mind we’re going to test at this point as it’s certainly achievable as an everyday setting.
As always first up is the CPUid chip info page and benchmarks along with the Geekbench results.
Intel i9 7940X @ 4.1GHz
Intel i9 7960X @ 4.1GHz
Both chips are clocked to the same level and the per-core score here reflects that. The multi-core score, of course, offers a leap from one chip to the other as you’d expect from throwing a few more cores into the equation.
The DAWBench classic and newer DSP test with Kontakt follow this and once again as there isn’t a whole lot I can add to this.
The added cores give us improvements across both of these chips as we’ve already seen in the more general purpose tests. The 7960X does appear to offer a slightly better performance curve at the higher buffer rates, which I suspect could be attributed to the increase in the cache but otherwise, it all scales pretty much as we’d expect.
Given the 7940X maintains the roughly £100 per extra core figure (when compared to the 7920X) at current pricing that Intel was aiming for at launch, it does seem to offer a similar sort of value proposition as the smaller i9’s just in this case more is more. The 7960X raises this to roughly £125 per core extra over the 7940X at current pricing, so a bit of cost creep there but certainly not as pricey as we’ve sometimes seen over the years on the higher end chips in the range.
The main concern initially was certainly regarding heat, but it looks like the continued refinement of the silicon since we saw the first i9 batches a few months ago has given them time to get ahead of this and ensure that the chips do well out of the box given adequate cooling.
With the launch of the CoffeeLake’s in the midrange, some of the value of the lower end enthusiast chips appear to have quickly become questionable, but the i9 range above it continues to offer performance levels henceforth unseen by Intel. The’s a lot of performance here, although the price matches it accordingly and we often find ourselves at this time where more midrange level systems are good enough for the majority of users.
However, for the power user with more exhaustive requirements who find that they can still manage to leverage every last drop of power from any system they get their hands on, I’m sure there will plenty here to peak your interest.
I’ll be honest, as far as this chipset naming scheme goes it feels that we might be starting to run out of sensible candidates. The Englishman in me wants to eschew this platform completely and hold out for the inevitable lake of Tea that is no doubt on the way. But alas the benchmarking has bean done and it’s too latte to skip over it now.
*Ahem* sorry, I think it’s almost out of my system now.
Right, where was I?
Time To Wake Up and Smell The….
Coffee Lake has been a blip on the horizon for quite a while now, and the promise of more cores in the middle and lower end CPU brackets whilst inevitable has no doubt taken a bit longer than some of us might have expected.
Is it a knee-jerk reaction to the AMD’s popular releases earlier this year? I suspect the platform itself isn’t, as it takes a lot more than 6 months to put together a new chipset and CPU range but certainly it feels like this new hardware selection might be hitting the shelf a little earlier than perhaps was originally planned.
Currently its clear that we’ve had a few generations now where the CPU’s haven’t really made any major gains other than silicon refinement and our clock speeds haven’t exceeded 5GHz from the Intel factory (of course, the more ambitious overclockers may have had other ideas), the obvious next move for offering more power in the range would be to stack up more cores much like the server-based bredrin in the Xeon range.
What is undeniable is that it certainly appears even to the casual observer that the competitor’s recent resurgence has forced Intel’s hand somewhat and very possibly accelerated the release schedule of the models being discussed here.
I say this as the introduction of the new range and i7 8700K specifically that we’re looking at today highlights some interesting oddities in the current lineup that could be in danger of making some of the more recent enthusiast chips look a little bit redundant.
This platform as a whole isn’t just about an i7 refresh though, rather we’re seeing upgrades to the mainstream i7’s, the i5’s and the i3’s which we’ll get on the bench over the coming weeks.
The i7’s have gained 2 additional physical cores and still have the hyperthreading meaning 12 logical cores total.
The i5’s have 6 cores and no hyperthreading.
The i3’s have 4 cores and no hyperthreading.
Positioning wise Intel’s own suggestions have focused towards the i5’s being pushed for gaming and streaming with up to 4 real physical cores being preferred for games and then a couple extra to handle the OS and streaming. The i3’s keep their traditional entry-level home office and media center sort of positioning that we’ve come to expect over the years and then that gives us the 6 core i7’s sat at the top of the pile of the more mainstream chip options.
Intel traditionally has always found itself a little lost when trying to market 6 cores or more. They know how to do it with servers where the software will lap up the parallelization capabilities of such CPUs with ease. But when it comes to the general public just how many regular users have had the need to leverage all those cores or indeed run software that can do it effectively?
It’s why in recent years there has been a marked move towards pushing these sorts of chips to content creators and offering the ability to provide the resources that those sort of users tend to benefit from. It’s the audio and video producers, editors, writers and artists that tend to benefit from these sorts of advances.
In short, very likely you dear reader.
Ok, so let’s take a look at some data.
At base clock rates the chip itself is sold as a 6 core with Hyper-threading and runs with a clock speed of 3.7GHz and a max turbo of 4.7GHz. For testing, I’ve locked off all the cores to the turbo max and tested with a Dark Rock 3 after testing various models before starting. With the cooler in hand, it was bouncing around 75 degrees after a few hours torture testing which is great. I did try running it around the 5GHz mark, which was easy to do and perfectly stable, although with the setup I had it was on the tipping point of overheating. If you updated it to a water cooling loop I reckon you’ll have this running fine around the 5GHz and indeed I did for some of the testing period with no real issues, although I did notice that the voltages and heat start to creep up rapidly past the 4.7GHz point.
The Geekbench 4 results show us some interesting and even slightly unexpected results. With the previous generation 7700K being clocked to 4.5GHz when I benched it and the 8700K being run at 4.7GHz I was expecting to see gains on the single core score as well as the increase in the multicore score. It’s only a few percent lower and I did retest a couple of times and found that this was repeatable and I had the results confirmed by another colleague.
The multicore score, on the other hand, shows the gains that this chip is all about with it not only exceeding the previous generation as you would expect with more cores being available. The gains here, in fact, highlight something I was already thinking about earlier in the year when the enthusiast i7’s got a refresh, in that this chip looks to not only match the 7800X found in the top end range but somewhat exceeding its capabilities at a lower overall price point.
In the testing above both the DAWBench DSP and the DAWBench vi tests continue to reflect this too, effectively raising questions as to the point of that entry-level 7800X in the enthusiast range.
The is almost price parity between the 7800X and 8700K at launch although the X299 boards tend to come in around £50 to £100 or more than the boards we’re seeing in the Z370 range. You do of course get extra memory slots in the X299 range, but then you can still mount 64GB on the mid-range board which for a lot of users is likely to be enough for the lifecycle of any new machine.
You also get an onboard GPU solution with the 8700K and if anything has been proven over the recent Intel generations, its that those onboard GPU solutions they offer are pretty good in the studio these days, perhaps also offering additional value to any new system build.
Grinding Out A Conclusion
I’m sure pricing from both sides will be competitive over the coming months as they aim to steal market share from each other. So with that in mind, it’s handy to keep these metrics in mind, along with the current market pricing at your time of purchase in order to make your own informed choice. I will say that at this point Intel has done well to reposition themselves after AMD’s strongest year in a very long time, although really their biggest achievement here looks to have been cannibalizing part of their own range in the process.
That, of course, is by no means is a complaint as when pricing is smashed like this then the biggest winner out there is the buying public and that truly is a marvelous thing. Comparing the 8700K to the 7700K on Geekbench alone shows us a 50% improvement in performance overheads for a tiny bit more than the previous generation cost, which frankly is the sort of generation on generation improvement that we would all like to be seeing every couple of years, rather than the 10% extra every generation we’ve been seeing of late.
Whether you choose to go with an Intel or an AMD for your next upgrade, we’ve seen that the performance gains for your money are likely to be pretty great this time around on both platforms. If your current system is more than 3 or 4 years old then it’s even more likely that the will be a pretty strong upgrade path open to you when you do finally choose to take that jump. With hints of Ryzen 2 being on its way next year from AMD and the likelihood that Intel would never leave any new release unchallenged, we could be in for an interesting 2018 too!
Back in June this year we took a look at the first i9 CPU model with the launch of the i9 7900X. Intel has since followed on from that with the rest of the i9 chips receiving a paper launch back in late August and with the promise of those CPU’s making it into the publics hands shortly afterward. Since then we’ve seen the first stock start to arrive with us here in Scan and we’ve now had a chance to sit down and test the first of this extended i9 range in the shape of the i9 7920X.
The CPU itself is 12 cores along with hyper-threading, offering us a total of 24 logical cores to play with. The base clock of the chip is 2.9GHz and a max turbo frequency of 4.30GHz with a reported 140W TDP which is much in line with the rest of the chips below it in the enthusiast range. Running at that base clock speed the chip is 400MHz slower per core than the 10 core edition 7900X. So if you add up all the available cores running at those clock speeds (12 X 2900 vs 10 X 3300) and compare the two chips on paper, then the looks to be less than 2GHz total available overhead separating them but still in the 7920X’s favor.
So looking at it that way, why would you pay the premium £200 for the 12 core? Well interestingly both CPU’s claim to be able to turbo to the same max clock rating of 4.3GHz, although it should be noted that turbo is designed to factor in power usage and heat generation too, so if your cooling isn’t up to the job then you shouldn’t expect it to be hitting such heady heights constantly and whilst I’m concerned that I may be sounding like a broken record by this point, as with all the high-end CPU releases this year you should be taking care with your cooling selection in order to ensure you get the maximum amount of performance from your chip.
Of course, the last thing we want to see is the power states throttling the chip in use and hampering our testing, so as always we’ve ensured decent cooling but aimed to keep the noise levels reasonable where we can. Normally we’d look to tweak it up to max turbo and lock it off, whilst keeping those temperatures in check and ensuring the system will be able to deliver a constant performance return for your needs.
However, in this case, I’ve not taken it quite all the way to the turbo max, choosing to keep it held back slightly at 4.2GHz across all cores. I was finding that the CPU would only ever bounce of 4.3GHz when left to work under its own optimized settings and on the sort of air cooling we tend to favour it wouldn’t quite maintain the 4.3GHz that was achieved with the 7900X in the last round of testing without occasionally throttling back. It will, however, do it on an AIO water loop cooler, although you’re adding another higher speed fan in that scenario and I didn’t feel the tradeoff was worth it personally, but certainly worth considering for anyone lucky to have a separate machine and control room where a bit more noise would go unnoticed.
Just as a note at this point, if you run it at stock and let it work its own turbo settings then you can expect an idle temperature around 40 degrees and under heavy load it still should be keeping it under 80 degrees on average which is acceptable and certainly better than we suspected around the time of the 7900X launch. However, I was seeing the P-states raising and dropping the core clock speeds in order to keep its power usage down and upon running Geekbench and comparing the results that my 4.2GHz on all cores setting gave us an additional 2000 points (around 7% increase) over the turbo to 4.3GHz default setting found in the stock configuration. My own temps idled in the 40’s and maxed around 85 degrees whilst running the torture tests for an afternoon, so for a few degrees more you can ensure that you get more constant performance from the setup.
Also worth noting is that we’ve had our CAD workstations up to around 4.5GHz and higher in a number of instances although in those instances we’re talking about a full water loop and a number of extra fans to maintain stability under that sort of workload, which wouldn’t be ideal for users working in close proximity to a highly sensitive mic.
Ok, so first up the CPUz information for the chip at hand, as well it’s Geelbench results.
More importantly for this comparison is the Geekbench 4 results and to be frank it’s all pretty much where we’d expect it to be in this one.
The single core score is down compared with the 7900X, but we’d expect this given the 4.2GHz clocking of the chip against the 4.3GHz 7900X. The multicore score is similarly up, but then we have a few more cores so all in all pretty much as expected here.
On with the DAWBench tests and again, no real surprises here. I’d peg it at being around an average of 10% or so increase over the 7900X which given we’re just stacking more cores on the same chip design really shouldn’t surprise us at all. It’s a solid solution and certainly the highest benching we’ve seen so far barring the models due to land above it. Bang per buck it’s £1020 price tag when compared to the £900 for the 10 core edition it seems to perform well on the Intel price curve and it looks like the wider market situation has curbed some of the price points we might have otherwise seen these chips hit.
And that’s the crux of it right now. Depending on your application and needs the are solutions from both sides that might fit you well. I’m not going to delve too far into discussing the value of the offerings that are currently available as prices do seem to be in flux to some degree with this generation. Initially, when it was listed we were discussing an estimated price of £100 per core and now we seem to be around £90 per core at the time of writing which seems to be a positive result for anyone wishing to pick one up.
Of course, the benchmarks should always be kept in mind along with that current pricing and it remains great to see continued healthy competition and I suspect with the further chips still to come this year, we may still see some additional movement before the market truly starts to settle after what really has been a release packed 12 months.
Intels i9 announcement this year felt like it pretty much came out of nowhere, and whilst everyone was expecting Intel to refresh its enthusiast range, I suspect few people anticipated quite the spread of chips that have been announced over the recent months.
So here we are looking at the first entry to Intel’s new high-end range. I’ve split this first look into 2 parts, with this section devoted to the i9 7900X and some discussion of the lower end models as the full range is explained. I’ll follow up in the near future with a forthcoming post to cover the i7’s coming in below this model, just as soon as we have the chance to grab some chips and run those through the test bench too.
There has been a sizable amount of press about this chip already as it was the first one to make it out into the wild along with the 4 core Kabylake X chips that have also appeared on this refresh, although those are likely to be of far less interest to those of us looking to build new studio solutions.
A tale of two microarchitectures.
Kabylake X and Skylake X have both launched at the same time and certainly raised eyebrows in confusion from a number of quarters. Intels own tick/tock cycle of advancement and process refinement has gone askew in recent years, where the “high-end desktop” ( HEDT chips) models just as the midrange CPU’s at the start of this year have gained a third generation at the same 14nm manufacturing process level in the shape of Kabylake.
Kabylake with the mid-range release kept the same 14nm design as the Skylake series before it and eaked out some more minor gains through platform refinement. In fact, some of the biggest changes to be found were in the improved onboard GPU found inside of it rather than the raw CPU performance itself, which as always is one of the key things missing in the HEDT edition. All this means that whilst we have a release where it’s technically two different chip ranges, the isn’t a whole lot left to differentiate between them. IN fact given how the new chip ranges continue to steam ahead in the mid-range, this looks like an attempt to help bring the high-end options back up to parity with the current mid-range again quickly which I think will ultimately help make things less confusing in future versions, even if right now it has managed to confuse things within the range quite a bit.
Kabylake X itself has taken a sizable amount of flak prior to launch and certainly appears to raise a lot of questions on an initial glance. The whole selling point of the HEDT chip up until this point has been largely more cores and more raw performance, so an announcement of what is essentially a mid-range i5/i7 grade 4 core CPU solution appearing on this chipset was somewhat of a surprise to a lot of people.
As with the other models on this chipset range, the 4 cores are being marketed as enthusiast solutions, although in this instance we see them looking to capture a gaming enthusiast segment. The have been some early reports of high overclocks being seen, but so far these look to be largely cherry-picked as the gains seen in early competition benchmarking have been hard to achieve with the early retail models currently appearing.
Whilst ultimately not really of much interest in the audio & video worlds where the software can leverage far more cores than the average game, potentially the is a solid opportunity here for that gaming market that they appear to be going after if they can refine these chips for overclocking over the coming months. However early specification and production choices have been head-scratchingly odd so far, although we’ll come back to this a bit later.
Touch the Sky(lake).
So at the other end of the spectrum from those Kabylake X chips is the new current flagship for the time being in the shape of the Skylake 7900X. 10 physical cores with hyper-threading give us a total of 20 logical cores to play with here. This is the first chip announced from the i9 range and larger 12,14,16,18 core editions are all penciled in over the coming year or so, however, details are scarce on them at this time.
At first glance, it’s a little confusing as to why they would even make this chip the first of its class when the rest of the range isn’t fully unveiled at this point. Looking through the rest of range specifications alongside it, then it becomes clear that they look to be reserving the i9’s for CPU’s that can handle a full 44+ PCIe lane configuration. These lanes are used for offering bandwidth to the connected cards and high-speed storage devices and needless to say this has proven a fairly controversial move as well.
The 7900X offers up the full complement of those 44 lanes although the 7820X and 7800X chips that we’ll be looking at in forthcoming coverage both arrive with 28 lanes in place. For most audio users this is unlikely to make any real difference, with the key usage for all those lanes often being for GPU usage where X16 cards are the standard and anyone wanting to fit more than one is going to appreciate more lanes for the bandwidth. With the previous generation we even tended to advise going with the entry level 6800K for audio over the 6850K above it, which cost 50% more but offered very little of benefit in the performance stakes but did ramp up the number of available PCIe lanes, choosing instead to reserve this for anyone running multiple GPU’s in the system like users with heavy video editing requirements.
Summer of 79(00X)
So what’s new?
Much like AMD and their infinity fabric design which was implemented to improve cross-core communication within the chip itself, Intel’s arrived with its own “Mesh” technology.
Functioning much like AMD’s design, it removes the ring based communication path between cores and RAM and implements a multi-point mesh design, brought in to enable shorter paths between them. In my previous Ryzen coverage I noted some poor performance scaling at lower buffer settings which seemed to smooth itself out once you went over a 192 buffer setting. In the run-up to this, I’ve retested a number of CPU’s and boards on the AMD side and it does appear that even after a number of tweaks and improvements at the BIOS level the scaling is still the same. On the plus side, as it’s proven to be a known constant and always manifests, in the same manner, I feel a lot more comfortable working with them now we are fully aware of this.
In Intels case I had some apprehension going in that given it is the companies first attempt at this in a consumer grade solution and that perhaps we’d be seeing the same sort of performance limitations that we saw on the AMD’s, but so far at least with the 7900X the internal chip latency has been superb. Even running at a 64 buffer we’ve been seeing 100% CPU load prior to the audio breaking up in playback, making this one of the most efficient chips I think I’ve possibly had on the desk.
So certainly a plus point there as the load capability seems to scale perfectly across the various buffer settings tested.
RAW performance wise I’ve run it through both CPU-Z and Geekbench again.
The multi-core result in Geekbench looks modest, although it’s worth noting the single core gains going on here compared to the previous generation 10 core the 6950X. On the basic DAWBench 4 test this doesn’t really show us up any great gains, rather it returns the sort of minor bump in performance that we’d kind of expect.
However whilst more cores can help spread the load, a lot of firms have always driven home the importance of raw clock speeds as well and once we start to look at more complex chains this becomes a little clearer. A VSTi channel with effects or additional processing on it needs to be sent to the CPU as a whole chain as it proves rather inefficient to chop up a channel signal chain for parallel processing.
A good single core score can mean slipping in just enough time to be able to squeeze in another full channel and effects chain and once you multiply that by the number of cores here, it’s easy to see how the combination of both a large number of cores and a high single core score can really translate into a higher total track count and is something we see manifest in the Kontakt based DAWBench VI test.
In this instance the performance gains over the previous generation seems quite sizable and whilst there is no doubt gains have been had from a change in architecture and that high-efficiency CPU usage we’ve already seen it should be noted here that this is close to a 20% increase in clock speed in play here too.
When we test we aim to do so around the all core turbo level. Modern Intel CPU’s have two turbo ratings, one is the “all core” level to which we can auto boost all the cores if the temperatures are safe and the other is the “Turbo 3.0” mode where it boosts a single core or it did in previous generations, but now we see it boosting the two strongest cores where the system permits.
The 7900X has a 4.5GHz 2 core turbo ability of 4.5GHz but we’ve chosen to lock it off at the all core turbo point in the testing. Running at stock clock levels we saw it boost the two cores correctly a number of times, but even under stress testing the 2 core maximum couldn’t be hit constantly without overheating on the low noise cooling solution we are using. The best we managed was a constant 4.45GHz at a temperature we were happy with, so we dialed it back to all core turbo clock speed of 4.3GHz across all cores and locked it in place for the testing, with it behaving well around this level.
It’s not uncommon for a first few batches of silicon on any new chip range to run a bit hot and normally this tends to get better as the generation gets refined. It’s the first time we’ve seen these sorts of temperatures on a chip range however and the is a strong argument to be made for going with either one of the top 2 or 3 air coolers on the market currently or defaulting to a water loop based cooling setup for any machine considering this chip. In a tower case this shouldn’t prove a problem but for rack systems, I suspect the 7900X might prove to be off-limits for the time being.
I’d fully expect the i7’s that are going to come in below it to be more reasonable and we should know about that in the next update, but it does raise some questions regarding the chips higher up in the i9 range that are due with us over the next 12 months. The has already been some debate about Intel choosing to go with thermal paste between the chip and the heatsink, rather than the more effective soldering method, although early tests by users de-lidding their chips hasn’t returned much more than 10 degrees worth of improvement, which is a fairly small gain for such a drastic step. We can only hope they figure out an improved way of improving the chips thermal handling with the impending i9’s or simply return to the older soldered method, otherwise, it could be quite some time until we see the no doubt hotter 12+ core editions making it to market.
In isolation, it looks fine from a performance point of view and gives the average sort of generation on generation gains that we would expect from an Intel range refresh, maybe pumped up a little as they’ve chosen to release them to market with raised base clocks. This leaves little room for overclocking, but it does give the buyer who simply wants the fastest model they can get out of the box and run it at stock.
The problem is that this isn’t in isolation and whilst we’ve gotten used to Intel’s 10% year on year gains over recent generations, there has to be many a user who longs for the sort of gains we saw when the X58 generation arrived or even when AMD dropped the Athlon 64 range on us all those years ago.
Ryzen made that sort of gain upon release, although they were so far behind that it didn’t do much more than breaking them even. This refresh puts Intel in a stronger place performance wise and it has to be noted that this chip has been incoming for a while. Certainly since long before Ryzen reignited the CPU war and it feels like they may have simply squeezed it a bit harder than normal to make it look more competitive.
This isn’t a game changer response to AMD. I doubt we’ll be seeing that for a year or two at this point and it will give AMD continued opportunities to apply pressure. What it has done however is what a lot of us hoped for initially and that it is forcing Intel to re-examine its pricing structure to some degree.
What we have here is a 10 core CPU for a third cheaper than the last generation 10 core CPU they released. Coming in around the £900 it rebalances the performance to price ratio to quite some degree and will no doubt once more help make the “i” series CPU’s attractive to more than a few users again, after a number of months of it being very much up for debate in various usage segments.
So will the impending AMD Threadripper upset this again?
I guess we’re going to find out soon enough over the coming months, but one thing for sure is that we’re finally seeing some competition here again, firstly on pure pricing but surely this should be a safe bet for kick-starting some CPU advancements again. This feels kinda like the Prescott VS Athlon 64 days and the upshot of that era was some huge gains in performance and solid improvements being made generation upon generation.
The cost and overall performance here keeps the 7900X in the running despite its obvious issues, and that raw grunt on offer makes it a very valid choice where the performance is required. The only real fly in the ointment is the heat and noise requirements most audio systems have, although hopefully as the silicon yields improve and refine this will mature into a cooler solution than it is now. It’s certainly going to be interesting to see how this pans out as the bigger models start making it to market over the coming year or so and of course with the smaller i7 brethren over the coming days.