GeekRadar / The comparison desk

COMPARE.

GPU

NVIDIA

NVIDIA P104-101

Graphics processor

Video memory: 4 GB^*
Memory bus: 256 bit^*
Rated memory bandwidth: 256.30 GB/s^*

No matched benchmark results are available for this entry yet.

Processing & platform

Architecture: Pascal^*
Process node: 16 nm^*
GPU cores: 2,560^*
Reported GPU frequency: 1,506 MHz^*
Graphics interface: PCIe 1.0 x4^*
Reported chip die area: 314 mm²^*
Transistors: 7,200 million^*
GPU L2 cache: 2 MB^*
Graphics processor: GP104^*
Texture mapping units: 160^*
Raster output units: 64^*
Streaming multiprocessors: 20^*
Chip foundry: TSMC^*
Card or module format: Dual-slot^*
Suggested system PSU: 200 W^*
External power connectors: 1x 8-pin^*
DirectX / feature level: 12 (12_1)^*
OpenGL: 4.6^*
OpenCL: 3.0^*
Vulkan: 1.4^*
CUDA compute capability: 6.1^*
Shader model: 6.8^*

Memory & storage

Memory options: 4^*
Video memory: 4 GB^*
Video memory type: GDDR5^*
Memory bus: 256 bit^*
Rated memory bandwidth: 256.30 GB/s^*

System configuration

Detailed specifications

Specifications 1 · 12 fields

Foundry: TSMC^*
Vendor: NVIDIA^*
FP16 GFLOPS: 134.6^* GFLOPS
FP32 GFLOPS: 8,617^* GFLOPS
FP64 GFLOPS: 269.3^* GFLOPS
Product: NVIDIA P104-101^*
Type: GPU^*
Release Date: 2018-01-05^*
Process Size (nm): 16^* nm
Die Size (mm^2): 314^* mm²
Transistors (million): 7,200^*
Freq (MHz): 1,506^* MHz

Specifications 2 · 28 fields

proc transistors: 7,200 million^*
proc die size: 314^* mm²
cores: 2,560^*
sm count: 20^*
base clock: 1,506^*
boost clock: 1,683^*
memory clock: 2,002^*
ram: 4^*
name: NVIDIA P104-101^*
ram type: GDDR5^*
bus width: 256^*
bus interface: PCIe 1.0 x4^*
ram bandwidth: 256.3^* GB/s
fp16 performance: 0.1346^*
fp32 performance: 8.617^*
fp64 performance: 0.2693^*
suggested psu: 200^* W
slot width: Dual-slot^*
power connectors: 1x 8-pin^*
api cuda: 6.1^*
is retail board: No^*
gpi value: 5.16^*
vendor: nvd^*
manufacturer: NVIDIA^*
arch name: Pascal^*
card release date: 2018-01-05^*
proc foundry: TSMC^*
proc process size: 16^* nm

Specifications 3 · 37 fields

CUDA compute capability

6.1^*

DirectX / feature level

12 (12_1)^*

OpenCL

3.0^*

OpenGL

4.6^*

Shader model

6.8^*

Vulkan

1.4^*

Architecture

Pascal^*

Graphics interface

PCIe 1.0 x4^*

Memory bus

256^* bit

Reported shader processors

2,560^*

Reported theoretical FP16 peak

0.1346^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP32 peak

8.617^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP64 peak

0.2693^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Declared L1 cache configuration

48 KB (per SM)^*

Declared GPU L2 cache

2 MB^*

Reported display outputs

No outputs^*

Theoretical pixel fill rate

107.7 GPixel/s^*

External power connectors

1x 8-pin^*

Reported chip die area

314 mm²^*

Chip foundry

TSMC^*

Process node declaration

16 nm^*

Transistors

7,200 million^*

Graphics processor

GP104^*

Reported video memory

4^* GB

Rated memory bandwidth

256.3 GB/s^*

Video memory type

GDDR5^*

Raster output units

64^*

Reported shading units

2,560^*

Card or module format

Dual-slot^*

Streaming multiprocessors

20^*

Suggested system PSU

200 W^*

2:4 tensor sparsity

No^*

Theoretical texture fill rate

269.3 GTexel/s^*

Texture mapping units

160^*

GPU cores

2,560^*

Video memory

4^* GB

Video memory type

GDDR5^*

* Provisional data