NVIDIA GRID A100A Full Specs

Summary

6,912 Shaders 900MHz

32GB HBM2e1.87TB/s

Peak·FP32·FP16 12.44 TFLOPS

Form Factor vGPU

TDP 400W

Highlights

Peak OPS

Peak OPS FP32 FP16 Tensor FP16-16 FP16-16 Tensor Sparse Tensor FP16-32 FP16-32 Tensor Sparse FP64 Tensor FP64 BF16 Tensor BF16 BF16 Tensor Sparse Tensor TF32 Tensor INT8 Tensor INT4 Memory Bandwidth

GRID A100A1.59 POPSINT4 Tensor Sparse

x1

Top Score3.35 PFLOPS FP4 Tensor SparseDesktop1xGeForce RTX 5090

x2.10

Theoretical Performance

Clock Speed900MHz UC·Boost·OC

Peak OPS 1.59 POPSINT4 Tensor Sparse

Tensor FP16-16 199.1 TFLOPS FP16-16 Tensor Sparse 398.1 TFLOPS Tensor FP16-32 199.1 TFLOPS FP16-32 Tensor Sparse 398.1 TFLOPS

BF16 24.88 TFLOPS Tensor BF16 199.1 TFLOPS BF16 Tensor Sparse 398.1 TFLOPS

Tensor TF32 99.53 TFLOPS

FP32 12.44 TFLOPS

FP64 6.22 TFLOPS Tensor FP64 12.44 TFLOPS

Tensor INT4 796.3 TOPS

Tensor INT8 398.1 TOPS

Pixel Rate 172.8 GPixel/s

Texture Rate 388.8 GTexel/s

Core Config

Shaders 6,912 Shaders

TMUs 432 TMUs

ROPs 192 ROPs

Tensor Cores 432 T-Cores

SMs 108 SMs

Clocks

Boost Clock 900MHz

Cache

L1 64KB/SM Tex

L2 Cache 40MB shared

Memory

32GB HBM2e

Memory Bus 6144-bit

Memory Speed 2.4GT/s

Memory Bandwidth 1.87TB/s

ECC No

Power & Thermals

TDP 400W

Video Output

Adaptive Sync G-Sync FreeSync

DSC Not Supported

API Support

OpenCL 3 Vulkan 1.2

CUDA 8 PureVideo HD VP10 VDPAU Feature Set J

Decoder

Decoder 5x NVDEC 4

Codec MPEG-1 MPEG-2 MPEG-4 VC-1 VP8 VP9 AVC (H.264) HEVC (H.265)

Card

Form Factor vGPU

PCIe PCIe 4.0 x16

Cooling Passive

Multi-GPU Multi-GPU Support Supported Multi-GPU Type NVLink

Chip

Manufacturer NVIDIA

Chip Designer NVIDIA

Architecture Ampere

Family GRID

Branding

Codename NV170

Market Segment Server

Release Date May 14, 2020

Fabrication

Foundry TSMC

Fabrication Node 7N

Die Size 826mm²

Transistor Count 54.2B

Transistor Density 65.62 MTr/mm²

Images