Hardware comparison
NVIDIA RTX 4090 (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 69 tracked models each one can run.
Scroll sideways to compare both machines. Keyboard: focus the table, then use the arrow keys. Metric names stay visible.
| Hardware | NVIDIA RTX 4090 (24GB) | NVIDIA RTX 6000 Ada (48GB) |
|---|---|---|
| Memory | 24.0 GB | 48.0 GB (better for this comparison) |
| Bandwidth | 1008 GB/s (better for this comparison) | 960 GB/s |
| Price (approx) | $2,999 (better for this comparison)GPU only · host PC required, sold separately | $6,800GPU only · host PC required, sold separately |
| LLMs it runs | 51 of 69 | 57 of 69 (better for this comparison) |
| Best model it runs | Gemma 4 31BQ3_K_M · estimated 41–62 tok/s | Gemma 4 31BQ8_0 · estimated 20–30 tok/s |
Both prices cover GPU cards only. A host PC is required and its cost is excluded. Prices are approximate catalog estimates; check current configurations and prices before buying.
The recommended quants differ. These estimated speeds are not a like-for-like comparison. Estimates use 4,096 context tokens, llama.cpp and one stream.
The NVIDIA RTX 6000 Ada (48GB) runs 6 more of the tracked models (57 vs 51), thanks to its 48.0 GB of memory. The NVIDIA RTX 4090 (24GB) has more memory bandwidth (1008 GB/s). Higher bandwidth can help generation speed, but model, quant, architecture and runtime also affect the estimates.