NVIDIA
GPUDatacenterReleased Apr 12, 2021

NVIDIA A16

16 GB GDDR6 ECC · 128-bit · 1,280 cores

A data center card for virtual desktop infrastructure (VDI) and cloud workstation services that places four Ampere GA107 GPUs on a single passively cooled, dual-slot PCIe board. Each GPU has 1,280 CUDA cores, 40 third-generation Tensor Cores, 10 second-generation RT Cores and its own 16 GB of ECC GDDR6, for 64 GB per card. The card supports NVIDIA vPC, vApps, RTX vWS and AI Enterprise software, with four NVENC and eight NVDEC engines, including AV1 decode, to maximize user density.

General

Manufacturer
NVIDIA
Model
A16
Segment
Datacenter
GPU chip
GA107
Architecture
Ampere
Release date
Apr 12, 2021

Memory

Memory size
16 GB
Memory type
GDDR6 ECC
Bus width
128-bit
Bandwidth
200 GB/s
Memory clock
—

Clocks & silicon

GPU clock
885 MHz
Boost clock
1,695 MHz
Shader units
1,280 CUDA Cores
Tensor / AI cores
40
RT cores
10
Transistors
—
Technology (nm)
8 nm
Samsung 8N (8 nm)
TDP / TBP
250 W

Interface & APIs

Bus interface
PCIe 4.0 x16
SLI / multi-GPU
4 GPUs on one board (no NVLink)
DirectX
12 Ultimate (12_2)
OpenGL
4.6
Vulkan
1.4
Other APIs
CUDA (Compute Capability 8.6), OpenCL 3.0

AI support

Precisions
FP8FP16BF16INT8INT4TF32FP64
AI accelerator
3rd Gen Tensor Cores (4x 40) + 2nd Gen RT Cores (4x 10)
Peak AI TOPS
—
Notes
—
All NVIDIA GPUs →

Datacenter gpus from NVIDIA announced around the same time as the NVIDIA A16. Add any of them to your compare list to see the specs side by side.

NVIDIAgpu

NVIDIA A10G

24 GB GDDR6 · 384-bit · 10,240 cores

Released
Nov 11, 2021
NVIDIAgpu

NVIDIA A100X

80 GB HBM2e · 5120-bit · 6,912 cores

Released
Nov 9, 2021
NVIDIAgpu

NVIDIA A2

16 GB GDDR6 · 128-bit · 1,280 cores

Released
Nov 9, 2021
NVIDIAgpu

NVIDIA A10

24 GB GDDR6 · 384-bit · 9,216 cores

Released
Apr 12, 2021
NVIDIAgpu

NVIDIA A30

24 GB HBM2 · 3072-bit · 3,584 cores

Released
Apr 12, 2021
NVIDIAgpu

NVIDIA A40

48 GB GDDR6 ECC · 384-bit · 10,752 cores

Released
Oct 5, 2020