Persevering with our protection of AMD’s 2022 Monetary Analyst day, we’ve got the matter of AMD’s forthcoming RDNA 3 GPU structure and the Navi 3X GPUs that might be constructed upon it. Up till now, AMD has been pretty quiet about what to anticipate with RDNA 3, however as RDNA 2 approaches its second birthday and the primary RDNA 3 merchandise are slated to launch this 12 months, AMD is providing a number of the first vital particulars on the GPU structure.
Before everything, let’s discuss efficiency. The Navi 3X household, to be constructed on a 5nm course of (TSMC’s, little doubt) is focusing on a greater-than 50% performance-per-watt uplift versus RDNA 2. It is a vital and comparable uplift as to AMD noticed shifting from RDNA (1) to RDNA 2. And whereas such a declare from AMD would have appeared ostentatious two years in the past, RDNA 2 has given AMD’s GPU groups a big quantity of renewed credibility.
Fortunately for AMD, in contrast to the 1-to-2 transition, they don’t need to discover a strategy to provide you with a 50% uplift based mostly on structure and DVFS optimizations alone. The 5nm course of signifies that Navi 3X is getting a full node’s enchancment from the TSMC N7/N6 based mostly Navi 2X GPU household. In consequence, AMD will see a big effectivity enchancment from that alone.
However with that mentioned, as of late a single node bounce by itself can’t ship a 50% perf-per-watt enchancment (RIP Dennard scaling). So there are a number of structure enhancements deliberate for RDNA 3. This consists of the following era of AMD’s on-die Infinity Cache, and what AMD is terming an optimized graphics pipeline. In response to the corporate, the GPU compute unit (CU) can also be being rearchitected, although to what diploma stays to be seen.
However the largest information of all on this entrance is that, confirming a 12 months’s price of rumors and a number of other patent purposes, AMD might be utilizing chiplets with RDNA 3. To what diploma, AMD isn’t saying, however the implication is that not less than one GPU tier (as we all know it) is shifting from a monolithic GPU to a chiplet-style design, utilizing a number of smaller chips.
Chiplets are in some respects the holy grail of GPU building, as a result of they provide GPU designers choices for scaling up GPUs previous in the present day’s die dimension (reticle) and yield limits. That mentioned, it’s additionally a holy grail as a result of the immense quantity of knowledge that should be handed between completely different components of a GPU (on the order of terabytes per second) may be very onerous to do – and really essential to do if you’d like a multi-chip GPU to have the ability to current itself as a single system. We’ve seen Apple sort out the duty by basically bridging two M1 SoCs collectively, however it’s by no means been finished with a high-performance GPU earlier than.
Notably, AMD calls this an “superior” chiplet design. That moniker tends to get thrown round when a chip is being packaged utilizing some form of superior, high-density interconnect resembling EMIB, which differentiates it from easier designs resembling Zen 2/3 chiplets, which merely route their indicators by the natural packaging with none enhanced applied sciences. So whereas we’re eagerly awaiting additional particulars of what AMD is doing right here, it wouldn’t in any respect be stunning to seek out out that AMD is utilizing a type of Native Si Interconnect (LSI) know-how (such because the Elevated Fanout Bridge used for the MI200 household of accelerators) to straight and intently bridge two RNDA 3 chiplets.
At this level, AMD isn’t going into any extra particulars on the structure or Navi 3X GPUs. At present is a teaser and roadmap replace for the analyst market, not an announcement of what we will solely assume would be the Radeon RX 7000 household of video playing cards. None the much less, with the primary RDNA 3 merchandise slated to launch later this 12 months, a extra formal announcement can’t be too distant. So we’re trying ahead to listening to extra about what stands to be a significant shake-up within the nature of GPU design and fabrication.
