Can the NVIDIA RTX 3070 run Gemma 2 9B?

Yes, with heavy quantization
QuantQ3_K_MNoticeably degraded
Memory7.9 GBof 8 GB
Speed~60tokens/s

Only just. Gemma 2 9B fits on the NVIDIA RTX 3070 at Q3_K_M, which costs noticeable quality. Expect about 60 tokens/second. A card with more memory is a better experience.

Try other context lengths in the calculator →
Run it in the cloud

Want full quality? Run Gemma 2 9B on a bigger rented GPU.

Rent a cloud GPU by the hour. No upfront cost, cancel any time.

RunPodOn-demand GPUs with ready-made templates for Ollama and vLLM.Rent a GPU on RunPod → Vast.aiA marketplace of rentable GPUs, from RTX cards to H100s.Rent a GPU on Vast.ai →

Memory needed by quantization

At 8,192 tokens of context.

QuantQualityMemoryFits?Speed
FP16Full quality21.9 GBToo big—
Q8_0Near-lossless13.2 GBToo big—
Q6_KExcellent11 GBToo big—
Q5_K_MVery good10 GBToo big—
Q4_K_MGood, most popular9 GBToo big—
Q3_K_MNoticeably degraded7.9 GBFits~60 tok/s
Q2_KHeavily degraded6.9 GBFits~78 tok/s

How to run Gemma 2 9B on the NVIDIA RTX 3070

Ollama's default gemma2:9b download is Q4_K_M, which needs about 9 GB, more than the 8 GB available. Download the Q3_K_M GGUF instead (LM Studio lists quants per model), or try a smaller context length.

GPUs that run Gemma 2 9B comfortably

Check NVIDIA RTX 3080 10GB prices →

Models that run well on the NVIDIA RTX 3070