GB300
NVIDIA GB300 Full Specs
40,960 Shaders 2.04GHz |
576GB HBM3e16TB/s |
·· 166.7 TFLOPS |
Form Factor Superchip |
TDP 2800W |
GB30040 PFLOPSMXFP4 Tensor Sparse (2x 20 PFLOPS)Clock Speed ·· |
Peak OPS 40 PFLOPSMXFP4 Tensor Sparse |
Tensor FP4 30 PFLOPSFP4 Tensor Sparse 40 PFLOPSMXFP4 Tensor 30 PFLOPSMXFP4 Tensor Sparse 40 PFLOPSNVFP4 Tensor 30 PFLOPSNVFP4 Tensor Sparse 40 PFLOPS |
FP6 Tensor 10 PFLOPSFP6 Tensor Sparse 20 PFLOPSMXFP6 Tensor 10 PFLOPSMXFP6 Tensor Sparse 20 PFLOPS |
Tensor FP8-16 10 PFLOPSFP8-16 Tensor Sparse 20 PFLOPSTensor FP8-32 10 PFLOPSFP8-32 Tensor Sparse 20 PFLOPSMXFP8 Tensor 10 PFLOPSMXFP8 Tensor Sparse 20 PFLOPS |
FP16 166.7 TFLOPSTensor FP16-16 5 PFLOPSFP16-16 Tensor Sparse 10 PFLOPSTensor FP16-32 5 PFLOPSFP16-32 Tensor Sparse 10 PFLOPS |
BF16 166.7 TFLOPSTensor BF16 5 PFLOPSBF16 Tensor Sparse 10 PFLOPS |
Tensor TF32 2.5 PFLOPS |
FP32 166.7 TFLOPS |
FP64 2.6 TFLOPS |
Tensor INT8 312.5 TOPS |
INT32 83.33 TOPS |
Pixel Rate 65.1 GPixel/s |
Texture Rate 1.3 TTexel/s |
Shaders 40,960 Shaders |
TMUs 1,280 TMUs |
ROPs 64 ROPs |
Tensor Cores 1,280 T-Cores |
SMs 320 SMs |
Boost Clock 2.04GHz |
Tensor Clock 1.91GHz |
L2 Cache 126MB shared |
576GB HBM3e |
Memory Bus 8192-bit |
Memory Speed 7.8GT/s |
Memory Bandwidth 16TB/s |
ECC Yes |
TDP 2800W |
CUDA 10.3 |
Decoder 7x NVDEC 6 |
Codec MPEG-1 MPEG-2 MPEG-4 VC-1 VP8 VP9 AVC (H.264) HEVC (H.265) AV1 |
Form Factor Superchip |
Cooling Passive |
Multi-GPU Multi-GPU Type NVLink |
Manufacturer |
Chip Designer |
Architecture |
Family |
Branding  |
Market Segment Server |
Release Date Mar 18, 2025 |
Foundry TSMC |
Fabrication Node 4NP |
Die Size 3240mm² |
Transistor Count 416B |
Transistor Density 128.4 MTr/mm² |