Is a used RTX 3090 still worth it for AI?
Not at today's prices: from $1,475 ($61.46 per GB of VRAM) it's number 17 of 52 PCIe cards with 24 GB or more by price per GB. Cheaper per GB now: Tesla P40 ($11.25/GB), Instinct MI50 32 GB ($19.80/GB), Radeon Pro W6800 ($27.50/GB).
For running models, VRAM decides what fits and memory bandwidth decides how fast it writes; the RTX 3090 has both, CUDA support from every AI tool, and NVLink on most cards. Against it: 350 W, and no FP8 (newer cards run some workloads faster). Two of them hold a 70B model at Q4.
24 GB and more, cheapest per GB
| Card | VRAM | Lowest | Per GB | Bandwidth | Software |
|---|---|---|---|---|---|
| Tesla P40 | 24 GB | $270.00 | $11.25/GB | 346 GB/s | CUDA |
| Instinct MI50 32 GB | 32 GB | $633.68 | $19.80/GB | 1,024 GB/s | — |
| Radeon Pro W6800 | 32 GB | $879.99 | $27.50/GB | 512 GB/s | ROCm |
| Tesla V100 PCIe 32 GB | 32 GB | $975.00 | $30.47/GB | 900 GB/s | CUDA |
| Tesla V100S PCIe | 32 GB | $1,099 | $34.34/GB | 1,134 GB/s | CUDA |
| A16 | 64 GB | $2,200 | $34.37/GB | 800 GB/s | CUDA |
| Arc Pro B60 24 GB | 24 GB | $849.00 | $35.38/GB | 456 GB/s | — |
| Radeon RX 7900 XTX | 24 GB | $900.00 | $37.50/GB | 960 GB/s | ROCm |
| Titan RTX | 24 GB | $910.00 | $37.92/GB | 672 GB/s | CUDA |
| Instinct MI100 | 32 GB | $1,240 | $38.75/GB | 1,229 GB/s | ROCm |
| Quadro RTX 6000 | 24 GB | $979.98 | $40.83/GB | 672 GB/s | CUDA |
| Quadro RTX 8000 | 48 GB | $2,000 | $41.67/GB | 672 GB/s | CUDA |
Live eBay prices, shipping included. Every GPU, the best GPUs for local AI.
Prices and picks update every time the site does (live listings checked within the last few hours).