GPU comparison
NVIDIA RTX 3090 (24GB) vs NVIDIA RTX 6000 Ada (48GB) for local LLMs
How the two stack up for running open-source LLMs locally — memory, bandwidth, price, and how many of the 18 tracked models each one can run.
| NVIDIA RTX 3090 (24GB) | NVIDIA RTX 6000 Ada (48GB) | |
|---|---|---|
| Memory | 24.0 GB | 48.0 GB |
| Bandwidth | 936 GB/s | 960 GB/s |
| Price (approx) | $800 | $6,800 |
| LLMs it runs | 15 of 18 | 16 of 18 |
| Best model it runs | Gemma 4 31B · 31–47 tok/s | Gemma 4 31B · 21–31 tok/s |
The NVIDIA RTX 6000 Ada (48GB) runs 1 more of the tracked models (16 vs 15), thanks to its 48.0 GB of memory. The NVIDIA RTX 6000 Ada (48GB) has more memory bandwidth (960 GB/s), so it generates tokens faster at the same model and quant.