⚡ GPU LLM Inference Visualizer bandwidth vs VRAM

Select a GPU and model to see real-time token generation. Inspired by Lummox's benchmark: Bandwidth > VRAM — the spec sheet often lies by omission.

GPU

Model Size

Bandwidth (GB/s)

VRAM (GB)

Compute (TFLOPS FP16)

—

Tokens / s

—

Memory Used (GB)

—

Bandwidth Util

—

Bottleneck

—

↓ token pipeline · each dot = one token

GPU Comparison BW VRAM