GPU comparison
NVIDIA RTX 3090 (24GB) vs NVIDIA RTX 5070 (12GB) for local LLMs
How the two stack up for running open-source LLMs locally — memory, bandwidth, price, and how many of the 34 tracked models each one can run.
| NVIDIA RTX 3090 (24GB) | NVIDIA RTX 5070 (12GB) | |
|---|---|---|
| Memory | 24.0 GB | 12.0 GB |
| Bandwidth | 936 GB/s | 672 GB/s |
| Price (approx) | $1,099 | $629 |
| LLMs it runs | 25 of 34 | 13 of 34 |
| Best model it runs | Gemma 4 31B · 39–58 tok/s | Devstral Small 2 24B · 45–68 tok/s |
The NVIDIA RTX 3090 (24GB) runs 12 more of the tracked models (25 vs 13), thanks to its 24.0 GB of memory. The NVIDIA RTX 3090 (24GB) has more memory bandwidth (936 GB/s), so it generates tokens faster at the same model and quant.