Can I Run Nemotron 70B on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)?

Superseded model. Nemotron 70B has been superseded by Nemotron 3 Super. This page is kept for reference; the newer family is a better starting point. View Nemotron 3 Super →

Written by Jakub Rusinowski · Last updated October 15, 2024

Yes

Yes — Nemotron 70B Instruct at Q2_K needs about 26.7 GB of the 32 GB usable on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM) (~5.3 GB spare), at ~48.7 tok/s (estimated), with room for about 16,384 tokens of context.

Confidence: high · Recommended quantization: Q2_K · Estimated speed: ~48.7 tok/s

RTX 5090 Desktop (32 GB VRAM, 64 GB RAM) — what it gives a model

Usable memory for models32 GB
Memory bandwidth1792 GB/s
Form factorDesktop
Operating systemWindows or Linux
Memory upgradeableYes

Nemotron 70B on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM): memory by quantization

QuantMemory neededFits 32 GB?Max contextEst. speedDownload
F16144.7 GB✗ No141.2 GB
Q8_078.5 GB✗ No75 GB
Q6_K61.4 GB✗ No57.9 GB
Q5_K_M53.5 GB✗ No50 GB
Q4_K_M46.1 GB✗ No42.6 GB
Q3_K_M33.6 GB✗ No30.1 GB
Q2_K26.7 GB✓ Yes16K~48.7 tok/s23.2 GB

What to watch out for

RTX 5090 desktop limitations

Recommended setup

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

How these numbers are calculated

FAQ

Can I run Nemotron 70B on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)?

Yes — Nemotron 70B Instruct at Q2_K needs about 26.7 GB of the 32 GB usable on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM) (~5.3 GB spare), at ~48.7 tok/s (estimated), with room for about 16,384 tokens of context.

Which quantization of Nemotron 70B should I use on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)?

Q2_K — it needs about 26.7 GB of the 32 GB available, downloads as roughly 23.2 GB, and runs at an estimated 48.7 tokens/sec with up to 16K of context.

What limits Nemotron 70B on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)?

Nothing binding — the model fits with headroom and generates at a usable speed on this hardware.

Which runtime should I use?

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

Other Computers

Other Models on RTX 5090 Desktop (32 GB VRAM, 64 GB RAM)

Nemotron 70B on GPUs

What This Model Is Good At

Model & Tools

← Can I Run It? | Nemotron 70B model page | Check your hardware