A new processor was introduced half a year ago Nvidia Grace, which uses the ARMv9 architecture. This can be combined with an Nvidia Hopper GPU or with a second Grace chip, which then means the creation of a processor Grace Superchip. In this case, we reach 144 cores (2×72 cores), each of which has 1 MB of L2 cache memory. The entire superchip has 234 MB of L3 cache (2×117 MB). The chips should communicate at 900 GB/s using the NVLink-C2C connection. LPDDR5x memory with ECC with an access speed of 500 GB/s is used, the capacity can be up to 960 GB. The entire superchip, including memory, consumes 500 W.
In a recent blog post, Nvidia published a performance comparison of this superchip with two AMD EPYC 7763 processors, each with 64 cores and 280W TDP. In total, it was a configuration of 128 cores and 256 threads.
According to Nvidia, Grace is supposed to offer up to 2.5x more performance than AMD (which sounds suspiciously good considering the differences in core and thread counts). It is of course the classic “up to”, so in other tests we have lower results, for example 1.8× or 1.5×. According to Nvidia, the bigger difference should be in efficiency, where 2x to 3.5x better results were achieved in the four published tests.
Source: Svět hardware by www.svethardware.cz.
*The article has been translated based on the content of Svět hardware by www.svethardware.cz. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!
*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.
*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!