En vivo
AI hardware

NVIDIA B200

FP16 throughput
2.3×10¹⁵ FLOP/s
Memory
168 GB
TDP
1.0 kW

NVIDIA B200 is an AI accelerator designed by NVIDIA (GPU), released in November 2024.

It delivers 2.3×10¹⁵ FLOP/s of tensor FP16/BF16 throughput, 4.5×10¹⁵ FLOP/s at FP8 and 9×10¹⁵ FLOP/s at FP4. It carries 168 GB of memory at 7 TB/s of bandwidth. Its thermal design power is 1.0 kW.

Full record
Manufacturer
NVIDIA
Type
GPU
Release date
Nov 15, 2024
Tensor FP16/BF16
2.3×10¹⁵ FLOP/s
FP8
4.5×10¹⁵ FLOP/s
FP4
9×10¹⁵ FLOP/s
FP32
6.2×10¹³ FLOP/s
FP64
3.1×10¹³ FLOP/s
INT8
4.5×10¹⁵ OP/s
Memory
168 GB
Memory bandwidth
7 TB
TDP
1.0 kW
Foundry
TSMC
More from NVIDIA
SourceEpoch AI, 'Machine Learning Hardware'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/machine-learning-hardware. Licensed under CC BY 4.0.
← All ai hardware