All AI inference chips

GB300 NVL72 vs B300

Spec-sheet and pricing comparison of NVIDIA GB300 NVL72 and NVIDIA B300 with links to continuously measured LLM inference benchmarks on identical workloads.

Spec-sheet comparison

GB300 NVL72B300GB300 NVL72 / B300
Memory per chip278 GB HBM3e268 GB HBM3e1.04x
Memory bandwidth8 TB/s8 TB/s1.0x
Dense FP8 compute5,000 TFLOP/s4,500 TFLOP/s1.11x
Dense FP4 compute15,000 TFLOP/s13,500 TFLOP/s1.11x
TDP1,400 W1,200 W1.17x
Hourly rate (neocloud tier)$2.79/hr$2.52/hr1.11x
Scale-up world size72 chips8 chips9.0x

Ratios are spec-sheet values; see the live compare pages for measured deltas.

Frequently asked questions

Which has more memory, GB300 NVL72 or B300?
GB300 NVL72 offers 278 GB HBM3e per chip versus 268 GB HBM3e on B300 (1.04x the capacity).
How do GB300 NVL72 and B300 prices compare?
At the neocloud tier the SemiAnalysis TCO model rates GB300 NVL72 at $2.79/hr versus $2.52/hr for B300. Hourly price alone is misleading; the per-dollar compare pages divide measured throughput by these rates.
Is GB300 NVL72 faster than B300 for LLM inference?
On paper GB300 NVL72 has 1.11x the dense FP8 compute of B300, but delivered tokens per second depend on the model, framework, precision and interactivity target. InferenceX measures both chips daily on identical workloads; see the live compare pages for current results.

See live benchmark results

Every number above is static hardware data. Delivered tokens per second, cost per million tokens and energy per token are measured continuously on the dashboard:

Go deeper with the SemiAnalysis models

InferenceX measures delivered inference performance. The SemiAnalysis institutional models cover the market behind these chips: who ships them, who buys them, and what they cost to own.