Can I Run InternLM 3 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)?

Written by Jakub Rusinowski · Last updated January 15, 2025

Yes, but it is tight

Yes, but it is tight — InternLM 3 20B Instruct at Q8_0 needs about 23.6 GB of the 24 GB usable on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM), leaving only ~0.4 GB before the runtime starts swapping. Expect ~29.7 tok/s (estimated), with room for about 8,192 tokens of context.

Confidence: low · Recommended quantization: Q8_0 · Estimated speed: ~29.7 tok/s

RTX 3090 Desktop (24 GB VRAM, 64 GB RAM) — what it gives a model

Usable memory for models	24 GB
Memory bandwidth	936 GB/s
Form factor	Desktop
Operating system	Windows or Linux
Memory upgradeable	Yes

InternLM 3 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM): memory by quantization

Quant	Memory needed	Fits 24 GB?	Max context	Est. speed	Download
F16	42.4 GB	✗ No	—	—	40 GB
Q8_0	23.6 GB	✓ Yes	8K	~29.7 tok/s	21.3 GB
Q6_K	18.8 GB	✓ Yes	32K	~37.4 tok/s	16.4 GB
Q5_K_M	16.6 GB	✓ Yes	32K	~42.4 tok/s	14.2 GB
Q4_K_M	14.5 GB	✓ Yes	32K	~48.6 tok/s	12.1 GB
Q3_K_M	10.9 GB	✓ Yes	32K	~64.5 tok/s	8.5 GB
Q2_K	9 GB	✓ Yes	32K	~78.5 tok/s	6.6 GB

Which InternLM 3 sizes fit

Variant	Needs at Q4_K_M	Fits?	Est. speed
InternLM 3 20B Instruct	14.5 GB	✓ Fits	~48.6 tok/s
InternLM 3 8B Instruct	6.5 GB	✓ Fits	~99 tok/s

What to watch out for

Only ~0.4 GB of headroom at Q8_0: a longer context or a second application can push this into swapping.
This model does not publish its full attention configuration, so the KV-cache share of these figures is inferred from its parameter count rather than computed exactly.

RTX 3090 desktop limitations

The cheapest route to 24 GB of VRAM, and the standard used-market recommendation for local LLMs.
Older architecture: no FP8 acceleration, and higher idle power than a current card.

Recommended setup

Ollama or llama.cpp (CUDA/ROCm) — vLLM if you need concurrent serving

How these numbers are calculated

Memory need = quantized weights + KV cache at 8,192 tokens (f16) + 0.8 GB runtime overhead.
24 GB of VRAM on the NVIDIA GeForce RTX 3090 at 936 GB/s.
64 GB of system RAM available for CPU offload when a model exceeds VRAM.
Throughput is a memory-bandwidth roofline estimate, not a measurement. It is labelled "estimated" everywhere it appears.
KV cache is inferred from the parameter count because this model does not publish its full attention configuration.

FAQ