Last week, Nvidia made took a surprising move by announcing that it would be selecting the AMD processor for its upcoming DGX A100 deep learning system instead of Intel’s Xeon processor. This news has taken the industry by storm as NVIDIA has just swept aside its decades-old rivalry with AMD and decided to install the EPYC processor for its most powerful product.
Intel’s share in the market might go down compared to AMD
The vice president of NVIDIA’s DGX systems, Charlie Boyle, made a statement that they needed a fast CPU with as many cores and PCI lanes as possible to keep the GPUs in the system supplied with data. That’s why they used AMD CPUs with 64 cores each and lots of PCI lanes. Moreover, EPYC processor support PCIe Generation 4 and eight-channel memory which is two more than Intel’s Xeon Scalable processors.
Although, Intel is expected to add PCIe 4.0 to its features list during the launch of the 10nm Ice Lake server chip later this year. Yet, NVIDIA’s decision to use AMD Processor in the latest DGX system might nibble away Intel’s share in the market.
Why Nvidia chose AMD over Intel?
Although NVIDIA has begun a new relationship with AMD, time will tell for how much time they would work together. When compared to Intel’s Xeon Platinum 9282, the EPYC 7742 delivers more cores with significantly more cache onboard. Moreover, the EPYC processor has a lower TDP compared to the Xeon processor and a far lower price tag as well. Due to these improvements, the AMD processor is equipped with a much finer 7nm manufacturing process, which allows far more transitions to be packed together. Thus, it can optimize power consumption and clock speed as well.