Home › GPUs & Macs › NVIDIA RTX 3090 vs NVIDIA RTX 3090 Ti
NVIDIA RTX 3090 vs NVIDIA RTX 3090 Ti for local LLMs
Across 5 tests measured on both, the NVIDIA RTX 3090 Ti is faster in 5 and generates a median 1.14× the tokens/s of the NVIDIA RTX 3090.
On Qwen3.8 27B (Q4_K_M, llama-server) the NVIDIA RTX 3090 generates 36.6 tokens/s and the NVIDIA RTX 3090 Ti 41.6 tokens/s (measured).
Same model, same quant, same test
| Model | Quant / test | RTX 3090 tok/s | RTX 3090 Ti tok/s | Faster | Sources and run details |
|---|---|---|---|---|---|
| Qwen3.8 27B | Q4_K_M llama-server | 36.6 CUDA | 41.6 CUDA | RTX 3090 Ti +14% | qwen38-mtp RTX 3090 · qwen38-mtp RTX 3090 Ti |
| Llama 2 7B | Q4_0 llama-bench | 172 Vulkan | 172 CUDA | even | llama.cpp #10879 RTX 3090 · llama.cpp #15013 RTX 3090 Ti |
| Llama 3.1 8B Instruct | Q4_K_M LocalScore | 96.3 llamafile | 110 llamafile | RTX 3090 Ti +14% | LocalScore RTX 3090 · LocalScore RTX 3090 Ti |
| Llama 3.2 1B Instruct | Q4_K_M LocalScore | 270 llamafile | 354 llamafile | RTX 3090 Ti +31% | LocalScore RTX 3090 · LocalScore RTX 3090 Ti |
| Qwen2.5 14B Instruct | Q4_K_M LocalScore | 56.2 llamafile | 64.2 llamafile | RTX 3090 Ti +14% | LocalScore RTX 3090 · LocalScore RTX 3090 Ti |
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 3090 | NVIDIA RTX 3090 Ti | |
|---|---|---|
| Memory | 24 GB | 24 GB |
| Memory bandwidth | 936 GB/s | 1008 GB/s |
All results for the NVIDIA RTX 3090 · All results for the NVIDIA RTX 3090 Ti · More comparisons