AMD’s upcoming 3D V-Cache has been shown to improve bandwidth with minimal increase in latency

Highly anticipated: Speculation around AMD’s new 3D V-Cache technology has swirled since Dr. Lisa Su gave us a preview of Computex 2021. Since then, AMD and tech enthusiasts have remained cautiously optimistic about claims that the new chip stacking approach can generate substantial performance. gains with minimal impact on latency, responsiveness and overall functionality. A recent test of an EPYC processor with V-Cache gives an early indication that AMD’s performance-boosting claims might just be true.

Nobody knew quite what to expect when AMD announced its 3D V-Cache technology at Computex last summer. While some enthusiasts viewed the substantial increase in cache as an exciting development, other members of the community felt upset that the new offerings would not offer substantial increases in clock speed, improvements in energy consumption, etc. Last Friday, the Chips and Cheese tech news outlet published results from their initial tests with one of AMD’s new Milan-X processors with 3D V-Cache, the server-oriented EPYC 7V73X. And so far, things look promising.

According to the site’s summary, AMD has managed to significantly increase the cache size of a CPU (768MB) compared to the previous Milan CPU family (256MB). Chips and Cheese’s tests report impressive performance from the stacked processor and much larger L3 cache without causing significant increases in cache and memory latency. Initial testing shows that the latency penalty keeps the increase between three and four cycles.

If these preliminary results hold true for AMD’s upcoming AM4 and AM5 releases, such as the Ryzen 7 5800X3D, the chipmaker will no doubt continue to explore the possibilities and benefits associated with 3D chip stacking.

AMD’s current 3D stacking technology binds a single V-Cache chiplet to a processor’s existing base complex array (CCD) and cache. As the technology matures, it may be possible for future architectures to further extend their L3 cache capabilities by using additional chiplets.

We’ll have to wait and see what the future holds, but if the EPYC-based results are any indication of what’s possible, then AMD could deliver another massive performance boost with its next series of processors.