Can the NVIDIA RTX 3070 run Llama 3.1 8B?
Yes
QuantQ5_K_MVery good
Memory7.4 GBof 8 GB
Speed~47tokens/s
Yes. The NVIDIA RTX 3070 can run Llama 3.1 8B fully in VRAM at Q5_K_M (very good), using about 7.4 GB of the 8 GB available, at roughly 47 tokens/second.
Try other context lengths in the calculator →Memory needed by quantization
At 8,192 tokens of context.
| Quant | Quality | Memory | Fits? | Speed |
|---|---|---|---|---|
| FP16 | Full quality | 17.7 GB | Too big | — |
| Q8_0 | Near-lossless | 10.2 GB | Too big | — |
| Q6_K | Excellent | 8.3 GB | Too big | — |
| Q5_K_M | Very good | 7.4 GB | Fits | ~47 tok/s |
| Q4_K_M | Good, most popular | 6.5 GB | Fits | ~55 tok/s |
| Q3_K_M | Noticeably degraded | 5.6 GB | Fits | ~68 tok/s |
| Q2_K | Heavily degraded | 4.7 GB | Fits | ~89 tok/s |
How to run Llama 3.1 8B on the NVIDIA RTX 3070
With Ollama installed, one command downloads and starts it:
ollama run llama3.1:8bOllama's default download is Q4_K_M (about 6.5 GB here). You have room for Q5_K_M, which is very good: pick that GGUF in LM Studio or from Hugging Face.
GPUs that run Llama 3.1 8B comfortably
- NVIDIA RTX 4060 (8 GB)
- NVIDIA RTX 3080 10GB
- NVIDIA RTX 3060 12GB
- NVIDIA RTX 4070 (12 GB)
- NVIDIA RTX 5070 (12 GB)
Models that run well on the NVIDIA RTX 3070
Run it in the cloud
Don't want to buy hardware? Run Llama 3.1 8B on a rented GPU today.
Rent a cloud GPU by the hour. No upfront cost, cancel any time.