GeekRadar / The comparison desk

COMPARE.

Accelerators

NVIDIA

NVIDIA Rubin GPU

Accelerators

Process node: 3 nm^*

Processing & platform

Process node: 3 nm^*
Reported chip die area: 1,456 mm²^*
Transistors: 336,000 million^*
GPU L2 cache: 128 MB^*
Texture mapping units: 896^*
Raster output units: 24^*
Streaming multiprocessors: 224^*
Tensor cores: 896^*
Chip foundry: TSMC^*
Card or module format: SXM Module^*
Suggested system PSU: 2,700 W^*
OpenCL: 3.0^*
CUDA compute capability: 10.7^*
Compute processor: GR100^*
Reported accelerator cores: 28,672^*
Device memory: 288 GB^*
Device memory type: HBM4^*
Host interface: PCIe 6.0 x16^*
Declared accelerator TDP: 2,300 W^*

Memory & storage

Memory bus: 16,384 bit^*
Rated memory bandwidth: 22,100 GB/s^*

System configuration

Detailed specifications

Specifications · 29 fields

CUDA compute capability

10.7^*

OpenCL

3.0^*

Host interface

PCIe 6.0 x16^*

Memory bus

16,384^* bit

Reported accelerator cores

28,672^*

Reported theoretical FP16 peak

260.0^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP32 peak

130.0^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP64 peak

32.5^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Declared L1 cache configuration

256 KB (per SM)^*

Declared GPU L2 cache

128 MB^*

Reported display outputs

No outputs^*

Theoretical pixel fill rate

54.41 GPixel/s^*

Reported chip die area

1456 mm²^*

Chip foundry

TSMC^*

Process node declaration

3 nm^*

Transistors

336,000 million^*

Compute processor

GR100^*

Declared device memory

288^* GB

Rated memory bandwidth

22.1 TB/s^*

Device memory type

HBM4^*

Raster output units

24^*

Reported shading units

28,672^*

Card or module format

SXM Module^*

Streaming multiprocessors

224^*

Suggested system PSU

2700 W^*

Declared accelerator TDP

2300 W^*

Tensor cores

896^*

Theoretical texture fill rate

2,031.2 GTexel/s^*

Texture mapping units

896^*

* Provisional data