Llama on NVIDIA Ada

Can I run Llama 3.2 1B on an L40S?

Yes. Llama 3.2 1B at Q4_K_M uses 1.8 GB of the 44.3 GB available on a L40S, and runs at about 579 tokens per second. You can push the context to 128K.

Runs comfortably

1.8 GB of 44.3 GB · 4%
044 GB
Weights 0.7 GB
KV cache 0.3 GB
Runtime overhead 0.8 GB

Llama 3.2 1B at Q4_K_M leaves 42.6 GB spare on a L40S. There is room to raise the context length or move up a quantisation level.

Generation579tok/s
Prompt processing30653tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Llama 3.2 1B on a L40S

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 2.3 GB 3.3 GB Runs comfortably 128K 209
INT8 / W8A8 8.50 1.2 GB 2.2 GB Runs comfortably 128K 386
Q8_0 (GGUF) 8.50 1.2 GB 2.2 GB Runs comfortably 128K 386
FP8 (E4M3) 8.00 1.2 GB 2.2 GB Runs comfortably 128K 390
AWQ 4-bit 4.25 1.0 GB 2.0 GB Runs comfortably 128K 452
GPTQ 4-bit 4.25 1.0 GB 2.0 GB Runs comfortably 128K 452
MXFP4 4.25 1.0 GB 2.0 GB Runs comfortably 128K 452
Q6_K 6.56 0.9 GB 2.0 GB Runs comfortably 128K 462
Q5_K_M 5.67 0.8 GB 1.9 GB Runs comfortably 128K 522
Q5_K_S 5.52 0.8 GB 1.8 GB Runs comfortably 128K 533
Q4_K_M 4.85 0.7 GB 1.8 GB Runs comfortably 128K 579
Q4_K_S 4.58 0.7 GB 1.7 GB Runs comfortably 128K 600
Q4_0 4.55 0.7 GB 1.7 GB Runs comfortably 128K 602
IQ4_XS 4.25 0.7 GB 1.7 GB Runs comfortably 128K 627
Q3_K_M 3.91 0.6 GB 1.6 GB Runs comfortably 128K 658
IQ3_M 3.70 0.6 GB 1.6 GB Runs comfortably 128K 678
IQ3_XXS 3.06 0.5 GB 1.6 GB Runs comfortably 128K 750
Q2_K 2.63 0.5 GB 1.5 GB Runs comfortably 128K 807
IQ2_XXS 2.06 0.4 GB 1.4 GB Runs comfortably 128K 897
IQ1_M 1.75 0.4 GB 1.4 GB Runs comfortably 128K 956

Also worth checking

Llama 3.2 1B on every GPU
Full hardware tableEverything that runs on a L40S
Full model tableWhy the KV cache is the part that bites
Guide