NVIDIA Taps AMD’s EPYC Enterprise CPUs for DGX A100 AI Accelerator

NVIDIA has gone down an alternate route with the introduction of its DGX A100 AI accelerator when put into comparison with previous variants of DGX systems. With the DGX A100, NVIDIA has opted to utilize AMD’s EPYC line of CPUs, a change from previous models using Intel’s Xeon Platinum.

NVIDIA DGX A100 – Ampere’s Electrifying Launch

Last month, we received news that NVIDIA was gearing up to launch the long-anticipated and much-rumored Ampere GPU microarchitecture which has now taken form as the GA100. In the DGX A100’s case, the A100 GPUs will use the cut-down variant of the GA100.

AMD’s Record Q1 Growth Fueled By 7nm Sales Yet Qualcomm Led In IC Revenue

A100 – Specifications

  • 6,912 CUDA Cores
  • 108 SMs (Streaming Multiprocessors)
  • 422 Tensor Cores
  • 826mm Die Area
  • TSMC 7nm CoWoS (Chip-on-Wafer-on-Substrate)
  • 40 GB HBM2
  • PCIe 4.0

DGX A100 Specifications

  • 8X NVIDIA A100
    • 320GB Total VRAM
    • 3rd Generation NVLink
  • 2X AMD EPYC 7742
    • 128 Cores
    • 256 Threads
    • 128 PCIe 4.0 Lanes
  • 1 TB DDR4-3200 System Memory
  • 15 TB PCIe 4.0 NVMe SSD
  • 5 PFLOPS Compute

AMD & NVIDIA Partnership – AMD to Supply EPYC CPUs for DGX A100

To many individuals, a partnership between AMD and NVIDIA may seem counterintuitive, but for both manufacturers, such a partnership is highly beneficial. AMD’s EPYC lineup of CPUs boasts quite the feature set, and in that case, has attracted NVIDIA’s attention. Previously, NVIDIA’s DGX systems had been powered by Intel’s Xeon Platinum CPUs, but NVIDIA has migrated to AMD’s enterprise platform for the DGX A100.

 

Only 2nd Gen AMD EPYC processors can provide up to 64 cores and 128 lanes of PCIe 4 interconnectivity in a single x86 data center processor, and we’re excited to see how the power of the NVIDIA DGX A100 system enables the I/O bandwidth to be effectively doubled.

Raghu Nambiar – Corporate Vice President. Data Center Ecosystems & Application Engineering

The 2nd Gen AMD EPYC processors used in DGX A100 provide high performance and support for PCIe Gen4. NVIDIA has put those features to work to create the world’s most powerful AI system while maintaining compatibility with the GPU-optimized software stack used across the entire DGX family.

Charlie Boyle, Vice President & General Manager, DGX Systems – NVIDIA

NVIDIA’s DGX A100 will be powered by two AMD EPYC 7742 CPUs for a total of 128 cores and 256 threads, along with 1 TB DDR4-3200 system memory and will offer PCIe 4.0, a feature exclusive to AMD’s EPYC CPUs, with a total of 128 PCIe 4.0 lanes.



[ad_2]