Following on from Intel’s “Plus” series CPU refresh, AMD in turn introduces a new flagship in the guise of the 9950X3D2.
The AMD Ryzen X3D series first arrived with the 5800X3D back in 2022, where the increase in low level cache notably helped that chip secure the lead in the gaming market at the time. A chip still so well regarded in gaming circles that we’ve seen it’s recent re-release as part of the Ryzen 10th anniversary celebration. It’s benefits in audio terms were far more muted however, with the extra heat from the cache implementation requiring the chips to be slightly down-clocked in comparison to their none X3D equivalent. This lead to results where the chips often under-performed the cheaper none X3D variants, except in a limited number of scenarios.
Over the generations the design has evolved with the cache location being physically relocated to allow for more efficient cooling layout inside of the chip. One question has remained however, as when the cache has been implemented up until now, it has been done so on only one of the chips CCD blocks. The *800 chips and below (i.e. currently the 9800X3D) are based around a single core design, whereas if we look at the larger chips in the range *900 or *950 (i.e. currently the 9900X3D & 9950X3D) they are built around a twin CCD block arrangement. With the regular 9950X3D only having one CCD with the extra cache attached, it opens the design up to inter-block latency, the sort of internal lag which can cause problems for tighter ASIO buffer settings when recording. In theory, spreading the available cache across both CCD blocks, to optimize the shortest path, should allow for the best performance to be achieved from the chip design.
The AMD Ryzen 9950X3D2 is the first model to arrive with a dual cache arrangement and the chip we’re taking a look at today.
With a slightly reduced 100Mhz max clock speed on the 9950X3D2 when compared with the two other 9950X variants, the DSP test with its focus on raw processing performance sees it slot it right behind the two existing chips. With the DSP testing focusing on the CPU performance itself, as we’ve seen previously the memory related benefits of the extended cache will tend to have a limited impact upon this particular test.
The Kontakt focused VI test, as with most typical sound libraries will benefit from both processor and memory performance. Historically with the X3D series the improved cache performance has translated into performance gains seen at tightest of ASIO settings, a pattern that shows up here once more. Whilst it shows a modest gain over the older 9950X3D model on the ASIO 64 buffer, it continues to be outshone by the single core block based 9800X3D CPU, showing that the internal block latency still favours a simplified layout. The 9800X3D however drops off sharply once you move above the 64 buffer setting and the performance of the 9950X3D2 chip really does get to show it’s strength as it races away with the best result at this level.
However, once we move above this ASIO setting towards 256 and above, the 9950X3D2 loses traction and the Intel chips leap ahead to quite a notable degree. It perhaps makes the AMD chips more suitable for low latency situations for library playback and performance, but less so where working in the box with large libraries where typically the session is run with a higher buffer setting to maximize available performance.
The DAWBench BUS test sees the 9950X3D2 come in with the strongest X3D based result overall on the 64 buffer, although with the single CCD based AMD 9600X & 9700X chips still managing to keep ahead. It falls behind slightly as we step up through the 128 and 256 buffer before seeing the AMD results all levelling out across the ASIO 512 buffer.
Since the launch of the 5800X3D there has been a question of what could we achieved with local cache across both CCD blocks on the larger chips and AMD has typically been keen to insist that the gains were modest at best. Seeing the results here and that have been generally reported around the 9950X3D2 that certainly seems to have been the case. Is it a poor chip? No, its a most definitely a new flagship for AMD, the only problem is that they were already trailing Intel’s 285K CPU in most of the results and the recent 270K arrival gives similar levels of performance at a very attractive price point.
At the time of writing the cost of a 9950X3D2 is about 2.5 times the cost of a Intel 270K, without offering a convincing argument in performance terms. Had it been released a few months earlier and at a leaner price point this could have been an interesting chip, but Intel’s Ultra series “Plus” editions” look to have stolen it’s thunder in 2026.