Database
AI Hardware Benchmarks
Standardized AI workload benchmarks across every modern GPU. Filter, sort, compare.
| GPU | VRAM | TDP | Flux 1024 | SDXL | L3 70B Q4 | Score | MSRP |
|---|---|---|---|---|---|---|---|
| RTX 5090 NVIDIA | 32 GB | 575W | 4.80 it/s | 18.20 it/s | 42 t/s | 100 | $1999 |
| RTX 4090 NVIDIA | 24 GB | 450W | 3.20 it/s | 14.60 it/s | 28 t/s | 86 | $1599 |
| RTX 4080 Super NVIDIA | 16 GB | 320W | 2.10 it/s | 10.80 it/s | 19 t/s | 72 | $999 |
| RTX 4070 Ti Super NVIDIA | 16 GB | 285W | 1.80 it/s | 8.90 it/s | 14 t/s | 62 | $799 |
| RTX 3090 NVIDIA | 24 GB | 350W | 1.60 it/s | 7.40 it/s | 16 t/s | 58 | $799 |
| RX 7900 XTX AMD | 24 GB | 355W | 1.20 it/s | 5.80 it/s | 11 t/s | 49 | $899 |
All benchmarks run on standardized methodology · Flux.1 dev FP16 1024² 25 steps · SDXL Base 1024² 20 steps · Llama 3 70B Q4_K_M, 2k context