GPU Wiki

Compare GPUs

Pick up to four GPUs and compare their specs row by row. Specs that mean different things across generations are kept apart.

Popular comparisons

Add a third or fourth GPU
Spec B300
Overview
Product name NVIDIA B300
Architecture Blackwell Ultra
Generation / family NVIDIA B-Series
GPU codename GB110
Launch date 2025-03-18
Market segment Data Center in NVIDIA B-Series
Board type / entry type Data Center GPU
GPU
Process node TSMC 5 nm
Transistor count 104 billion
GPU chip GB110
SM / multiprocessor count 148
CUDA Cores 18944
Tensor Cores 592
ROPs 24
TMUs 592
Clocks
Base / Core clock 1,665 MHz
Boost clock 2,032 MHz
Memory clock 2,000 MHz
Memory
Memory size 288 GB
Memory type HBM3e
Memory bus width 8,192-bit
Memory bandwidth 4,100 GB/s
Bus / Interface
Bus interface PCIe 5.0 x16
AGP / PCI / PCIe distinction PCIe
Power
TDP / TBP / board power 1,400 W
Recommended PSU 1,800 W
Slot width SXM Module
Features
CUDA support 10.3
Performance / Benchmarks
FP32 throughput 76.99 TFLOPS
FP16 throughput 1,231.8 TFLOPS
Muted cells indicate unavailable, not applicable, or not directly comparable specifications.

DLAA counts as native resolution; DLSS, FSR and XeSS upscaling never do. A ray tracing setting without a readable on or off value is read as on.