GPUDatacenterReleased Nov 11, 2021

NVIDIA A10G

24 GB GDDR6 · 384-bit · 10,240 cores

A variant of the Ampere A10 built for Amazon Web Services and used in EC2 G5 instances for graphics-intensive and machine learning workloads such as remote workstations, game streaming, rendering and inference. It enables more of the GA102 GPU than the A10, with 80 RT Cores, and comes as a dual-slot 300 W board. Introduced with the EC2 G5 family in November 2021.

General

Manufacturer: NVIDIA
Model: A10G
Segment: Datacenter
GPU chip: GA102
Architecture: Ampere
Release date: Nov 11, 2021
Price: Search on Amazon

Memory

Memory size: 24 GB
Memory type: GDDR6
Bus width: 384-bit
Bandwidth: 600 GB/s
Memory clock: 1,563 MHz

Clocks & silicon

GPU clock: —
Boost clock: 1,710 MHz
Shader units: 10,240 CUDA Cores
Tensor / AI cores: 320
RT cores: 80
Transistors: 28.3 billion
Technology (nm): 8 nm
Samsung 8N (8 nm)
TDP / TBP: 300 W

Interface & APIs

Bus interface: PCIe 4.0 x16
SLI / multi-GPU: None (no NVLink)
DirectX: 12 Ultimate (12_2)
OpenGL: 4.6
Vulkan: 1.3
Other APIs: CUDA (compute capability 8.6), OpenCL 3.0

AI support

Precisions: FP8FP16BF16INT8INT4TF32FP64
AI accelerator: 3rd Gen Tensor Cores (320) + 2nd Gen RT Cores (80)
Peak AI TOPS: 280 TOPS
Notes: Third-generation Tensor Cores; FP8 is not supported on Ampere. Peak throughput with structured sparsity: INT4 560 TOPS, INT8 280 TOPS, FP16 and BFLOAT16 140 TFLOPS, TF32 70 TFLOPS; dense rates are half. FP32 is 35 TFLOPS. One encoder and two decoders, including AV1 decode.