Can I run Llama 3.2 3B on NVIDIA RTX 3090 24GB?
Yes: Llama 3.2 3B runs on NVIDIA RTX 3090 24GB with F16 at roughly 83 tokens/s. Estimated memory 7.6 GiB at 8,192 tokens of context.
With F16 you can push the context to about 108,288 tokens before memory runs out (F16 KV cache).
Memory by quantisation (8,192-token context)
| Quantisation | Memory | Verdict | Est. speed |
|---|---|---|---|
| F16Unquantized reference | 7.6 GiB | Fits | ~83 tok/s |
| Q8_0Near-lossless | 4.8 GiB | Fits | ~140 tok/s |
| Q6_KVery close to F16 | 4.1 GiB | Fits | ~170 tok/s |
| Q5_K_MGreat quality/size | 3.7 GiB | Fits | ~189 tok/s |
| Q4_K_MBest default for most | 3.4 GiB | Fits | ~211 tok/s |
| IQ4_XSSmaller 4-bit (imatrix) | 3.2 GiB | Fits | ~230 tok/s |
| Q3_K_MNoticeable loss | 3.1 GiB | Fits | ~243 tok/s |
| Q2_KLast resort; big loss | 2.9 GiB | Fits | ~266 tok/s |
NVIDIA RTX 3090 24GB offers about 24 GiB of memory (94% usable for a model) at 936 GB/s. Speed is a bandwidth-bound estimate of single-stream generation at 65% efficiency; real results vary by runtime, drivers and settings, so measure on your machine.
How this is calculated
Weights ≈ parameters × bits per weight ÷ 8. KV cache ≈ 2 × 28 layers × 8 KV heads × 128 head dim × 2 bytes per token. Decode speed ≈ efficiency × memory bandwidth ÷ bytes read per token. Learn the maths in VRAM math and memory bandwidth, or try the VRAM calculator.
Llama 3.2 3B on other hardware
Similar models on NVIDIA RTX 3090 24GB
Running slowly or out of memory? See CUDA out of memory and the quantisation lesson.