GeekRadar / The comparison desk

COMPARE.

Accelerators

NVIDIA

NVIDIA B300 SXM6 AC

Accelerators

Process node: 5 nm^*

Processing & platform

Process node: 5 nm^*
Reported chip die area: 1,628 mm²^*
Transistors: 208,000 million^*
GPU L2 cache: 126 MB^*
Texture mapping units: 592^*
Raster output units: 24^*
Streaming multiprocessors: 148^*
Tensor cores: 592^*
Chip foundry: TSMC^*
Card or module format: SXM Module^*
Suggested system PSU: 1,500 W^*
OpenCL: 3.0^*
CUDA compute capability: 10.3^*
Compute processor: GB110^*
Reported accelerator cores: 18,944^*
Device memory: 288 GB^*
Device memory type: HBM3e^*
Host interface: PCIe 6.0 x16^*
Declared accelerator TDP: 1,100 W^*

Memory & storage

Memory bus: 8,192 bit^*
Rated memory bandwidth: 8,190.00 GB/s^*

System configuration

Detailed specifications

Specifications · 29 fields

CUDA compute capability

10.3^*

OpenCL

3.0^*

Host interface

PCIe 6.0 x16^*

Memory bus

8,192^* bit

Reported accelerator cores

18,944^*

Reported theoretical FP16 peak

76.99^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP32 peak

76.99^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Reported theoretical FP64 peak

1.2029^* TFLOPS

Scope and conditions

Arithmetic density: Not stated^*

Declared L1 cache configuration

256 KB (per SM)^*

Declared GPU L2 cache

126 MB^*

Reported display outputs

No outputs^*

Theoretical pixel fill rate

48.77 GPixel/s^*

Reported chip die area

1628 mm²^*

Chip foundry

TSMC^*

Process node declaration

5 nm^*

Transistors

208,000 million^*

Compute processor

GB110^*

Declared device memory

288^* GB

Rated memory bandwidth

8.19 TB/s^*

Device memory type

HBM3e^*

Raster output units

24^*

Reported shading units

18,944^*

Card or module format

SXM Module^*

Streaming multiprocessors

148^*

Suggested system PSU

1500 W^*

Declared accelerator TDP

1100 W^*

Tensor cores

592^*

Theoretical texture fill rate

1,202.9 GTexel/s^*

Texture mapping units

592^*

* Provisional data