Home › GPUs & Macs › NVIDIA RTX 3080 vs NVIDIA RTX 3090
NVIDIA RTX 3080 vs NVIDIA RTX 3090 for local LLMs
Across 4 tests measured on both, the NVIDIA RTX 3090 is faster in 4 and generates a median 1.18× the tokens/s of the NVIDIA RTX 3080.
On Llama 2 7B (Q4_0, llama-bench) the NVIDIA RTX 3080 generates 146 tokens/s and the NVIDIA RTX 3090 172 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | RTX 3080 tok/s | RTX 3090 tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Llama 2 7B | Q4_0 llama-bench | 146 Vulkan | 172 Vulkan | RTX 3090 +18% | llama.cpp #10879 RTX 3080 · llama.cpp #10879 RTX 3090 |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 81.1 llamafile | 96.3 llamafile | RTX 3090 +19% | LocalScore RTX 3080 · LocalScore RTX 3090 |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 242 llamafile | 270 llamafile | RTX 3090 +12% | LocalScore RTX 3080 · LocalScore RTX 3090 |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 42.0 llamafile | 56.2 llamafile | RTX 3090 +34% | LocalScore RTX 3080 · LocalScore RTX 3090 |
Generation speed in tokens/s, best measured backend for each chip (shown under the number). Backends can differ between the two chips; check them before drawing conclusions. Measured results only, nothing estimated.
Specs
| NVIDIA RTX 3080 | NVIDIA RTX 3090 | |
|---|---|---|
| Memory | 10 GB | 24 GB |
| Memory bandwidth | 760 GB/s | 936 GB/s |
All results for the NVIDIA RTX 3080 · All results for the NVIDIA RTX 3090 · More comparisons