NVIDIA
Nvidia has finally revealed its Grace processor, which was originally revealed in April 2021.
Thus, Grace CPU is a 72-core processor based on Arm v9.0 architecture. It contains 117MB of L3 cache and 68 PCIe 5.0 lanes and will be manufactured at TSMC's facility using 4nm process technology.
The processor is destined for the data center, and it's also worth remembering that Nvidia will also have a Grace CPU Superchip - in fact, two Grace processors combined.
Grace is also the first in the world to support LPDDR5x ECC memory with a total bandwidth of 1 TB/s. You can also highlight the C2C NVLINK interface with a transfer rate of 900 GB/s and twice the performance per watt compared to today's leading CPUs.
In terms of performance, Grace scores 370 points on Specrate_int_base and Grace CPU Superchip scores 740 points. For comparison, a pair of Epyc 7763 (128 cores in total) scores 861 points. At the same time, the Grace CPU Superchip consumes around 500 watts, while the two aforementioned Epyc processors require around 560 watts.
It's also worth noting that Grace is a highly specialized CPU for learning NLP models. That is, comparing it directly with Epyc or Xeon is not very correct.
Source: Nvidia
No comments:
Post a Comment