NVIDIAEstimated

B100

Blackwell · Blackwell · SXM · 2024

700 W Blackwell for air-cooled or retrofit deployments. Trades peak throughput for easier facility fit versus B200.

LLM trainingLLM inferenceMultimodalHPC
Precision fingerprint
6432t3216BF168i8
Memory

192 GB

HBM3e

Bandwidth

8 TB/s

peak

TDP

700 W

air or liquid

Max model

~80B

FP16, planning est.

Compute throughput

FP6430 TFLOPS
FP3230 TFLOPS
TF32875 TFLOPS
FP161.75 PFLOPS
BF161.75 PFLOPS
FP83.50 PFLOPS
INT83.50 PFLOPS

Platform & software

InterconnectNVLink 5 — 1.8 TB/s
PCIePCIe 6.0 x16
Coolingair or liquid
MIGSupported
PartitioningMIG
VirtualizationMIG, vGPU
FrameworksCUDA, TensorRT-LLM, Triton, NeMo, vLLM
Availabilitybare-metal
Known limitations
  • ·Power-capped variant of B200 — lower peak throughput
Related GPUs
NVIDIA

H100 SXM5

80 GB · 989 TFLOPS FP16

NVIDIA

H100 PCIe

80 GB · 756 TFLOPS FP16

NVIDIA

H100 NVL

94 GB · 989 TFLOPS FP16