Can I Run Nemotron Cascade 2 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)?

Written by Jakub Rusinowski · Last updated March 15, 2026

Yes, but it is tight

Yes, but it is tight — Nemotron Cascade 2 30B at Q5_K_M needs about 23.8 GB of the 24 GB usable on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM), leaving only ~0.2 GB before the runtime starts swapping. Expect ~29.6 tok/s (estimated), with room for about 8,192 tokens of context.

Confidence: low · Recommended quantization: Q5_K_M · Estimated speed: ~29.6 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

Nemotron Cascade 2 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	62.6 GB	✗ No	—	—	60 GB
Q8_0	34.5 GB	✗ No	—	—	31.9 GB
Q6_K	27.2 GB	✗ No	—	—	24.6 GB
Q5_K_M	23.8 GB	✓ Yes	8K	~29.6 tok/s	21.3 GB
Q4_K_M	20.7 GB	✓ Yes	16K	~34.1 tok/s	18.1 GB
Q3_K_M	15.4 GB	✓ Yes	32K	~46 tok/s	12.8 GB
Q2_K	12.4 GB	✓ Yes	32K	~56.9 tok/s	9.9 GB

Which Nemotron Cascade 2 sizes fit

Variant	Needs at Q4_K_M	Fits?	Est. speed
Nemotron Cascade 2 70B	45.4 GB	✗ Too large	—
Nemotron Cascade 2 30B	20.7 GB	✓ Fits	~34.1 tok/s

What to watch out for

Only ~0.2 GB of headroom at Q5_K_M: a longer context or a second application can push this into swapping.
1 larger variant of Nemotron Cascade 2 does not fit and would need CPU offload or different hardware.
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