Qwen on NVIDIA Ampere

Can I run Qwen3 8B on an RTX 3090?

Yes. Qwen3 8B at Q4_K_M uses 6.7 GB of the 21.8 GB available on a RTX 3090, and runs at about 108 tokens per second. You can push the context to 115K.

Runs comfortably

6.7 GB of 21.8 GB · 31%
022 GB
Weights 4.7 GB
KV cache 1.1 GB
Runtime overhead 0.8 GB

Qwen3 8B at Q4_K_M leaves 15.1 GB spare on a RTX 3090. There is room to raise the context length or move up a quantisation level.

Generation108tok/s
Prompt processing1821tok/s
Max context115Ktokens
KV per 1K tokens0GB

Every quantisation of Qwen3 8B on a RTX 3090

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 15.3 GB 17.2 GB Runs comfortably 40K 36.7
INT8 / W8A8 8.50 7.8 GB 9.8 GB Runs comfortably 93K 68.8
Q8_0 (GGUF) 8.50 7.8 GB 9.8 GB Runs comfortably 93K 68.8
FP8 (E4M3) 8.00 7.6 GB 9.6 GB Runs comfortably 95K 70.4
Q6_K 6.56 6.3 GB 8.2 GB Runs comfortably 104K 84.4
AWQ 4-bit 4.25 5.8 GB 7.7 GB Runs comfortably 108K 90.9
GPTQ 4-bit 4.25 5.8 GB 7.7 GB Runs comfortably 108K 90.9
MXFP4 4.25 5.8 GB 7.7 GB Runs comfortably 108K 90.9
Q5_K_M 5.67 5.4 GB 7.4 GB Runs comfortably 110K 96.1
Q5_K_S 5.52 5.3 GB 7.2 GB Runs comfortably 111K 98.4
Q4_K_M 4.85 4.7 GB 6.7 GB Runs comfortably 115K 108
Q4_K_S 4.58 4.5 GB 6.4 GB Runs comfortably 117K 113
Q4_0 4.55 4.5 GB 6.4 GB Runs comfortably 117K 113
IQ4_XS 4.25 4.2 GB 6.2 GB Runs comfortably 119K 119
Q3_K_M 3.91 4.0 GB 5.9 GB Runs comfortably 121K 126
IQ3_M 3.70 3.8 GB 5.7 GB Runs comfortably 122K 131
IQ3_XXS 3.06 3.3 GB 5.2 GB Runs comfortably 126K 148
Q2_K 2.63 2.9 GB 4.9 GB Runs comfortably 128K 162
IQ2_XXS 2.06 2.5 GB 4.4 GB Runs comfortably 128K 186
IQ1_M 1.75 2.2 GB 4.2 GB Runs comfortably 128K 202

Also worth checking

Qwen3 8B on every GPU
Full hardware tableEverything that runs on a RTX 3090
Full model tableWhy the KV cache is the part that bites
Guide