Gemma on NVIDIA Ada

Can I run Gemma 3 27B on an L40S?

Yes. Gemma 3 27B at Q4_K_M uses 17.5 GB of the 44.3 GB available on a L40S, and runs at about 31.4 tokens per second. You can push the context to 128K.

Runs comfortably

17.5 GB of 44.3 GB · 39%
044 GB
Weights 15.6 GB
KV cache 1.1 GB
Runtime overhead 0.8 GB

Gemma 3 27B at Q4_K_M leaves 26.8 GB spare on a L40S. There is room to raise the context length or move up a quantisation level.

Generation31.4tok/s
Prompt processing1387tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Gemma 3 27B on a L40S

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 51.0 GB 53.0 GB Won't fit — 3.10
INT8 / W8A8 8.50 26.8 GB 28.7 GB Runs comfortably 128K 18.7
Q8_0 (GGUF) 8.50 26.8 GB 28.7 GB Runs comfortably 128K 18.7
FP8 (E4M3) 8.00 25.5 GB 27.4 GB Runs comfortably 128K 19.6
Q6_K 6.56 20.9 GB 22.9 GB Runs comfortably 128K 23.7
Q5_K_M 5.67 18.1 GB 20.0 GB Runs comfortably 128K 27.2
Q5_K_S 5.52 17.6 GB 19.5 GB Runs comfortably 128K 27.9
Q4_K_M 4.85 15.6 GB 17.5 GB Runs comfortably 128K 31.4
AWQ 4-bit 4.25 15.5 GB 17.4 GB Runs comfortably 128K 31.6
GPTQ 4-bit 4.25 15.5 GB 17.4 GB Runs comfortably 128K 31.6
MXFP4 4.25 15.5 GB 17.4 GB Runs comfortably 128K 31.6
Q4_K_S 4.58 14.8 GB 16.7 GB Runs comfortably 128K 33.0
Q4_0 4.55 14.7 GB 16.6 GB Runs comfortably 128K 33.2
IQ4_XS 4.25 13.8 GB 15.7 GB Runs comfortably 128K 35.3
Q3_K_M 3.91 12.7 GB 14.7 GB Runs comfortably 128K 37.9
IQ3_M 3.70 12.1 GB 14.0 GB Runs comfortably 128K 39.8
IQ3_XXS 3.06 10.2 GB 12.1 GB Runs comfortably 128K 46.7
Q2_K 2.63 8.9 GB 10.8 GB Runs comfortably 128K 52.9
IQ2_XXS 2.06 7.1 GB 9.1 GB Runs comfortably 128K 64.2
IQ1_M 1.75 6.2 GB 8.1 GB Runs comfortably 128K 72.6

Also worth checking

Gemma 3 27B on every GPU
Full hardware tableEverything that runs on a L40S
Full model tableWhy the KV cache is the part that bites
Guide