Hardware comparison
NVIDIA DGX Spark (GB10, 128GB) vs NVIDIA RTX 5080 (16GB) 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 DGX Spark (GB10, 128GB) | NVIDIA RTX 5080 (16GB) |
|---|---|---|
| Memory | 128.0 GB (better for this comparison) | 16.0 GB |
| Bandwidth | 273 GB/s | 960 GB/s (better for this comparison) |
| Price (approx) | $4,699Complete system | $1,399GPU only · host PC required, sold separately |
| LLMs it runs | 62 of 69 (better for this comparison) | 48 of 69 |
| Best model it runs | Gemma 4 31BQ2_K · estimated 13–20 tok/s | Gemma 4 31BQ2_K · estimated 48–71 tok/s |
These prices cover different purchases: a GPU card excludes its required host PC; a complete system includes the machine. No price winner is marked. Prices are approximate catalog estimates; check current configurations and prices before buying.
Both estimated speeds use the same model and quant shown above. Estimates use 4,096 context tokens, llama.cpp and one stream.
The NVIDIA DGX Spark (GB10, 128GB) runs 14 more of the tracked models (62 vs 48), thanks to its 128.0 GB of memory. The NVIDIA RTX 5080 (16GB) has more memory bandwidth (960 GB/s). Higher bandwidth can help generation speed, but model, quant, architecture and runtime also affect the estimates.