Final week, Nvidia made an announcement that shook the trade as for the primary time ever, it swept apart its decades-old rivalry with AMD, choosing the EPYC server processor for its DGX A100 deep studying system and casting apart Intel’s Xeon.
In an announcement to CRN, Charlie Boyle, Vice President and Basic Supervisor of DGX Programs at Nvidia, defined the rationale behind the change.
“To maintain the GPUs in our system provided with information, we would have liked a quick CPU with as many cores and PCI lanes as attainable. The AMD CPUs we use have 64 cores every, a number of PCI lanes, and assist PCIe Gen4,” he mentioned.
Intel is predicted so as to add PCIe 4.zero to its characteristic listing when it launches the 10nm Ice Lake server chip later this yr however, for now, can solely sit and watch as AMD nibbles away at its market share. EPYC additionally helps eight-channel reminiscence, two greater than Intel’s Xeon Scalable processors.
The EPYC 7742 delivers extra cores (64 vs 56 with the Intel Xeon Platinum 9282) with considerably extra cache onboard (256MB vs 77MB), a decrease TDP (225W vs 400w) and a far cheaper price tag ($6,950 vs circa $25,000).
These marked enhancements are all due to AMD’s a lot finer 7nm manufacturing course of, which permits much more transistors to be packed collectively, optimising energy consumption and clock speeds.
Time will inform whether or not the transfer marks a everlasting thawing of the connection between Nvidia and AMD, or only a momentary truce.