GeekRadar / The comparison desk

COMPARE.

GPU

NVIDIA

NVIDIA A16 PCIe

Graphics processor

No matched benchmark results are available for this entry yet.

Processing & platform

Architecture: Ampere^*
Process node: 8 nm^*
GPUs per card: 4^*
Reported GPU frequency: 885 MHz^*
Graphics interface: PCIe 4.0 x8^*
Reported chip die area: 200 mm²^*
Chip/package TDP: 250 W^*
Transistors: 8,700 million^*
Graphics processor: GA107^*
Chip foundry: Samsung^*
Card or module format: Dual-slot^*
Declared GPU TDP: 250 W^*
Suggested system PSU: 600 W^*
External power connectors: 8-pin EPS^*
DirectX / feature level: 12 Ultimate (12_2)^*
OpenGL: 4.6^*
OpenCL: 3.0^*
Vulkan: 1.4^*
CUDA compute capability: 8.6^*
Shader model: 6.9^*
Reported video memory (allocation unspecified): 16 GB^*
Reported shader processors (allocation unspecified): 1,280^*
Reported memory bandwidth (allocation unspecified): 200.10 GB/s^*

Memory & storage

Video memory type: GDDR6^*

Power specifications

Thermal design power: 250 W^*

System configuration

Detailed specifications

Specifications 1 · 11 fields

Foundry: Samsung^*
Vendor: NVIDIA^*
FP16 GFLOPS: 8,678^* GFLOPS
FP32 GFLOPS: 8,678^* GFLOPS
FP64 GFLOPS: 271.2^* GFLOPS
Product: NVIDIA A16 PCIe^*
Type: GPU^*
Release Date: 2021-04-12^*
Process Size (nm): 8^* nm
TDP (W): 250^* W
Freq (MHz): 885^* MHz

Specifications 2 · 31 fields

proc transistors: 8,700 million^*
proc die size: 200^* mm²
Reported shader processors (allocation unspecified): 1,280^*
tensor cores: 40^*
rt cores: 10^*
sm count: 10^*
base clock: 1,312^*
boost clock: 1,755^*
memory clock: 1,563^*
Reported video memory (allocation unspecified): 16^*
name: NVIDIA A16 PCIe^*
ram type: GDDR6^*
bus width: 128^*
bus interface: PCIe 4.0 x8^*
Reported memory bandwidth (allocation unspecified): 200.1^* GB/s
fp16 performance: 4.493^*
fp32 performance: 4.493^*
fp64 performance: 0.1404^*
tdp: 250^* W
suggested psu: 600^* W
slot width: Dual-slot^*
power connectors: 8-pin EPS^*
api cuda: 8.6^*
is retail board: No^*
gpi value: 4.65^*
vendor: nvd^*
manufacturer: NVIDIA^*
arch name: Ampere^*
card release date: 2021-04-12^*
proc foundry: Samsung^*
proc process size: 8^* nm

Specifications 3 · 41 fields

CUDA compute capability

8.6^*

DirectX / feature level

12 Ultimate (12_2)^*

OpenCL

3.0^*

OpenGL

4.6^*

Shader model

6.9^*

Vulkan

1.4^*

Architecture

Ampere^*

Graphics interface

PCIe 4.0 x8^*

Memory bus (allocation unspecified)

128^* bit

Reported shader processors (allocation unspecified)

1,280^*

Reported theoretical FP16 peak

4.493^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP32 peak

4.493^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP64 peak

0.1404^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Declared L1 cache configuration

128 KB (per SM)^*

Declared GPU L2 cache (allocation unspecified)

2 MB^*

Reported display outputs

No outputs^*

Theoretical pixel fill rate

56.16 GPixel/s^*

External power connectors

8-pin EPS^*

Reported chip die area

200 mm²^*

Chip foundry

Samsung^*

Process node declaration

8 nm^*

Transistors

8,700 million^*

Graphics processor

GA107^*

GPUs per card

4^*

Reported video memory (allocation unspecified)

16^* GB

Rated memory bandwidth (allocation unspecified)

200.1 GB/s^*

Video memory type

GDDR6^*

Raster output units (allocation unspecified)

32^*

Ray tracing cores (allocation unspecified)

10^*

Reported shading units (allocation unspecified)

1,280^*

Card or module format

Dual-slot^*

Streaming multiprocessors (allocation unspecified)

10^*

Suggested system PSU

600 W^*

2:4 tensor sparsity

Yes^*

Declared GPU TDP

250 W^*

Tensor cores (allocation unspecified)

40^*

Theoretical texture fill rate

70.20 GTexel/s^*

Texture mapping units (allocation unspecified)

40^*

Reported shader processors (allocation unspecified)

1,280^*

Reported video memory (allocation unspecified)

16^* GB

Video memory type

GDDR6^*

* Provisional data