GPUDatacenterReleased Nov 8, 2022

NVIDIA A800 PCIe 40GB

40 GB HBM2 · 5120-bit · 6,912 cores

The NVIDIA A800 PCIe 40GB is a data center accelerator built on the same Ampere GA100 silicon as the A100 40GB, introduced in November 2022 as an export-compliant counterpart for the Chinese market with unchanged compute performance. It pairs 40 GB of HBM2 memory over a 5120-bit bus with 6,912 CUDA cores and 432 third-generation Tensor Cores, and two cards can be joined with an NVLink bridge restricted to 400 GB/s versus the 600 GB/s of the A100. Multi-Instance GPU splits the board into as many as seven isolated compute instances for workload isolation.

General

Manufacturer: NVIDIA
Model: A800 PCIe 40GB
Segment: Datacenter
GPU chip: GA100
Architecture: Ampere
Release date: Nov 8, 2022
Price: Search on Amazon

Memory

Memory size: 40 GB
Memory type: HBM2
Bus width: 5120-bit
Bandwidth: 1,555 GB/s
Memory clock: 1,215 MHz

Clocks & silicon

GPU clock: —
Boost clock: 1,410 MHz
Shader units: 6,912 CUDA Cores
Tensor / AI cores: 432
RT cores: —
Transistors: 54.2 billion
Technology (nm): 7 nm
TSMC N7
TDP / TBP: 250 W

Interface & APIs

Bus interface: PCIe 4.0 x16
SLI / multi-GPU: NVLink Bridge (2 GPUs, 400 GB/s)
DirectX: N/A (compute-only)
OpenGL: N/A (compute-only)
Vulkan: N/A (compute-only)
Other APIs: CUDA (compute capability 8.0), OpenCL 3.0

AI support

Precisions: FP8FP16BF16INT8INT4TF32FP64
AI accelerator: Third-generation Tensor Cores (432)
Peak AI TOPS: —
Notes: Third-generation Tensor Cores support FP64, TF32, BF16, FP16, INT8, INT4 and binary; FP8 is not available on Ampere and arrived with Hopper. Dense peaks are 19.5 TFLOPS FP64 Tensor Core, 156 TFLOPS TF32, 312 TFLOPS BF16 and FP16, and 624 TOPS INT8, each doubling with 2:4 structured sparsity. Multi-Instance GPU partitions the board into up to seven isolated instances.

All NVIDIA GPUs →

Datacenter gpus from NVIDIA announced around the same time as the NVIDIA A800 PCIe 40GB. Add any of them to your compare list to see the specs side by side.

gpu