GPU comparison
NVIDIA RTX 4090 (24GB) vs NVIDIA RTX 5090 (32GB) for local LLMs
How the two stack up for running open-source LLMs locally — memory, bandwidth, price, and how many of the 22 tracked models each one can run.
| NVIDIA RTX 4090 (24GB) | NVIDIA RTX 5090 (32GB) | |
|---|---|---|
| Memory | 24.0 GB | 32.0 GB |
| Bandwidth | 1008 GB/s | 1792 GB/s |
| Price (approx) | $1,999 | $3,299 |
| LLMs it runs | 16 of 22 | 16 of 22 |
| Best model it runs | Gemma 4 31B · 33–50 tok/s | Gemma 4 31B · 55–83 tok/s |
Both run the same number of models, but the NVIDIA RTX 5090 (32GB) has more memory (32.0 GB), so it can hold larger models at higher quality. The NVIDIA RTX 5090 (32GB) has more memory bandwidth (1792 GB/s), so it generates tokens faster at the same model and quant.