Can I Run SmolLM2 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)?
Written by Jakub Rusinowski · Last updated November 20, 2024
Yes — comfortably
Yes, comfortably — SmolLM2 1.7B Instruct at Q8_0 needs about 4.2 GB of the 24 GB usable on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM), leaving ~19.8 GB spare and running at ~167.2 tok/s (estimated), with room for about 8,192 tokens of context.
Confidence: high · Recommended quantization: Q8_0 · Estimated speed: ~167.2 tok/s
See what else this hardware can run →
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
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 |
SmolLM2 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 | 5.8 GB | ✓ Yes | 8K | ~121 tok/s | 3.4 GB |
| Q8_0 | 4.2 GB | ✓ Yes | 8K | ~167.2 tok/s | 1.8 GB |
| Q6_K | 3.8 GB | ✓ Yes | 8K | ~185.5 tok/s | 1.4 GB |
| Q5_K_M | 3.6 GB | ✓ Yes | 8K | ~195.3 tok/s | 1.2 GB |
| Q4_K_M | 3.4 GB | ✓ Yes | 8K | ~205.6 tok/s | 1 GB |
| Q3_K_M | 3.1 GB | ✓ Yes | 8K | ~225.7 tok/s | 0.7 GB |
| Q2_K | 3 GB | ✓ Yes | 8K | ~238.4 tok/s | 0.6 GB |
Which SmolLM2 sizes fit
| Variant | Needs at Q4_K_M | Fits? | Est. speed |
|---|---|---|---|
| SmolLM2 1.7B Instruct | 3.4 GB | ✓ Fits | ~205.6 tok/s |
| SmolLM2 360M Instruct | 1.4 GB | ✓ Fits | ~357.6 tok/s |
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 computed from this model's published attention configuration.
FAQ
Can I run SmolLM2 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)?
Yes, comfortably — SmolLM2 1.7B Instruct at Q8_0 needs about 4.2 GB of the 24 GB usable on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM), leaving ~19.8 GB spare and running at ~167.2 tok/s (estimated), with room for about 8,192 tokens of context.
Which quantization of SmolLM2 should I use on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)?
Q8_0 — it needs about 4.2 GB of the 24 GB available, downloads as roughly 1.8 GB, and runs at an estimated 167.2 tokens/sec with up to 8K of context.
What limits SmolLM2 on RTX 3090 Desktop (24 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 3090 Desktop (24 GB VRAM, 64 GB RAM)
- SmolLM3 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)
- StarCoder 2 on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)
- Aya 3B (Tiny Aya) on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)
- VibeThinker on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)
- Yi 1.5 Family on RTX 3090 Desktop (24 GB VRAM, 64 GB RAM)
SmolLM2 on GPUs
- SmolLM2 on NVIDIA GeForce RTX 5060 Ti 8GB
- SmolLM2 on NVIDIA GeForce RTX 5060
- SmolLM2 on NVIDIA GeForce RTX 4060