Llama on NVIDIA Ada

Can I run Llama 3.1 8B on an RTX 4080 Super?

Yes. Llama 3.1 8B at Q4_K_M uses 6.4 GB of the 14.2 GB available on a RTX 4080 Super, and runs at about 84.9 tokens per second. You can push the context to 70K.

Runs comfortably

6.4 GB of 14.2 GB · 45%
014 GB
Weights 4.6 GB
KV cache 1.0 GB
Runtime overhead 0.8 GB

Llama 3.1 8B at Q4_K_M leaves 7.8 GB spare on a RTX 4080 Super. There is room to raise the context length or move up a quantisation level.

Generation84.9tok/s
Prompt processing2720tok/s
Max context70Ktokens
KV per 1K tokens0GB

Every quantisation of Llama 3.1 8B on a RTX 4080 Super

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 15.0 GB 16.8 GB Won't fit — 9.58
INT8 / W8A8 8.50 7.7 GB 9.5 GB Runs comfortably 46K 53.2
Q8_0 (GGUF) 8.50 7.7 GB 9.5 GB Runs comfortably 46K 53.2
FP8 (E4M3) 8.00 7.5 GB 9.3 GB Runs comfortably 48K 54.7
Q6_K 6.56 6.1 GB 8.0 GB Runs comfortably 58K 65.7
AWQ 4-bit 4.25 5.4 GB 7.2 GB Runs comfortably 64K 73.6
GPTQ 4-bit 4.25 5.4 GB 7.2 GB Runs comfortably 64K 73.6
MXFP4 4.25 5.4 GB 7.2 GB Runs comfortably 64K 73.6
Q5_K_M 5.67 5.3 GB 7.1 GB Runs comfortably 65K 75.0
Q5_K_S 5.52 5.2 GB 7.0 GB Runs comfortably 66K 76.8
Q4_K_M 4.85 4.6 GB 6.4 GB Runs comfortably 70K 84.9
Q4_K_S 4.58 4.4 GB 6.2 GB Runs comfortably 72K 88.6
Q4_0 4.55 4.4 GB 6.2 GB Runs comfortably 72K 89.1
IQ4_XS 4.25 4.1 GB 5.9 GB Runs comfortably 74K 93.7
Q3_K_M 3.91 3.8 GB 5.7 GB Runs comfortably 77K 99.5
IQ3_M 3.70 3.7 GB 5.5 GB Runs comfortably 78K 104
IQ3_XXS 3.06 3.2 GB 5.0 GB Runs comfortably 82K 118
Q2_K 2.63 2.8 GB 4.6 GB Runs comfortably 85K 130
IQ2_XXS 2.06 2.3 GB 4.2 GB Runs comfortably 89K 151
IQ1_M 1.75 2.1 GB 3.9 GB Runs comfortably 91K 165

Also worth checking

Llama 3.1 8B on every GPU
Full hardware tableEverything that runs on a RTX 4080 Super
Full model tableWhy the KV cache is the part that bites
Guide