gpu4.pics
data via LocalScore
NVIDIA RTX 3090
llamafile 0.9.1
24 GB VRAM NVIDIA 350 W TDP
running Llama 3.2 1B Instruct· Q4_K_M· 1.5B
330
tokens / sec
3,662 · Excellent
Generation
330tok/s
higher is better
Prompt / prefill
14,061tok/s
higher is better
Time to first token
95 ms
lower is better
Efficiency
1.14 tok/s/W
Electricity cost
$0.042/Mtok
Power draw
296 W
AMD EPYC 7352 24-Core Processor (znver2) · 126 GB RAM · Linux · llamafile
gpu4.pics/r/ls-result-3

All benchmarks on this rig

ModelQuantGenPromptTTFTPowerPerf/W$/MtokScore
Llama 3.2 1B Instruct1.5BQ4_K_M33014,06195 ms296 W1.14 tok/s/W$0.042/Mtok3,662

How it stacks up — Llama 3.2 1B Instruct

NVIDIA RTX 3090
330
NVIDIA L40S
355
NVIDIA RTX 3090 Ti
354
NVIDIA H100
335
NVIDIA RTX A6000
315
NVIDIA A100 80GB
308