AMD will be announcing its next-generation HPC products during an event on the 8th of November. We expect the chipmaker to announce the Zen 3D-based Milan-X server processors and the Instinct MI250 accelerators. Milan-X will be the first lineup to feature 3D stacked cache (V-Cache) in the server and data center space. Intel’s Lakefield processors were the first to demonstrate the technology in the form of Foveros, but they were very limited in scope, die area, and availability. Milan-X will retain the Zen 3 core and the N7 process from TSMC, and as such, can be thought of as a special refresh or niche stack, much like the upcoming Sapphire Rapids-SP with on-die HBM memory.
|CPU Name||Cores/Threads||Base Clock||Boost Clock||L3 Cache (V-Cache + L3 Cache)||L2 Cache||TDP|
|AMD EPYC 7773X||64/128||2.2 GHz||3.5 GHz||512 + 256 MB||32 MB||280W|
|AMD EPYC 7573X||32/64||2.8 GHz||3.6 GHz||512 + 256 MB||32 MB||280W|
|AMD EPYC 7473X||24/48||2.8 GHz||3.7 GHz||512 + 256 MB||12 MB||240W|
|AMD EPYC 7373X||16/32||3.05 GHz||3.8 GHz||512 + 256 MB||8 MB||240W|
Looking at the specs, everything’s basically identical to the vanilla Milan parts, including the base and boost clocks, the TDP as well as the L2 cache (other than the crapton of L3 cache). This means that performance gains (as already indicated earlier) will vary from application to application, and won’t be much pronounced in every workload.
AMD to host Accelerated Data Center Premiere virtual event on November 8, 2021
AMD will host its Accelerated Data Center Premiere on November 8, 2021 at 11am ET, showcasing the company’s upcoming innovations with AMD EPYC processors and AMD Instinct accelerators.
The virtual event is scheduled to feature presentations by Dr. Lisa Su, AMD President and CEO, Forrest Norrod, Senior Vice President and General Manager, Data Center and Integrated Solutions Business Group, and Dan McNamara, Senior Vice President and General Manager of the Servers Business Unit.
The event will be accessible to the public at www.amd.com/en/events/data-center beginning at 11:00 AM EST. A replay will be available after the end of the live event.
The exact specifications of the MI2150X have also been shared. It’ll consist of a total of 110 CUs with a boost clock of 1.7GHz. This means that we’re likely looking at eight memory stacks, each featuring eight 2GB dies. This indicates a total bus width of 8,196-bits (1,024-bits x8 controllers), resulting in an overall bandwidth of 3.68 TB, roughly the same as the HBM variants of Sapphire Rapids-SP.
At the heart of the GPU core, there will be two 55 CU chiplets, resulting in an overall compute strength of 110 CU, with an impressive boost clock of 1.7GHz. Since Alderbaran can execute double-precision instructions (FP64) at native speeds, this will result in a double-precision throughput of 47.9 TFLOPs, an insane four times more than its predecessor, the MI100.
Even NVIDIA’s Ampere-based A100 Tensor core accelerator is capable of “only” 19.5 TFLOPs of FP64 compute. In terms of mixed-precision compute, we’re looking at 383 TFOPs of FP16 and BFLOAT16. In comparison, the MI100, topped out at “just” 184 and 92 TFLOPs in the two data types, respectively.