Geerling llama.cpp Vulkan on rtx2080ti (pi host): deepseek-r1-distill-qwen-14b Q4_K_M pp4096+tg128 143.29 tok/s
143.29 tok/s (pp4096+tg128)
Fields
id
benchmarks/geerling-vulkan-rtx2080ti-pi-deepseek-r1-distill-qwen-14b-q4km-pp4096-tg128Metric
tps_aggregate
Value
143.29
Unit
tok/s (pp4096+tg128)
Model
DeepSeek-R1-Distill-Qwen-14B-Q4_K_M.gguf (8.37 GiB, 14.77 B params)
Methodology
THIRD-PARTY MEASURED (Jeff Geerling, geerlingguy/ai-benchmarks). llama.cpp build f9ec8858e (7474), backend Vulkan, default llama-bench flags (ngl 99). Host: Raspberry Pi CM5 16GB, Pi OS Trixie, PCIe-attached card. Model file size 8.37 GiB, 14.77 B params, Q4_K_M. llama-bench combined pp4096+tg128 figure: one prompt of 4096 tokens followed by 128 generated tokens, reported as a single tok/s. This is neither pure prefill nor pure decode; stored as tps_aggregate with concurrency 1. Whole-system wall power during the run 83.6 W (node scope, not device).
Measured by
third_party
Reproducible
yes
Mean power (W)
83.6
Power scope
node
Thermal state
unknown
notes
Source issue prints the llama-bench mean and the +/- spread; the stored value is the mean. The Pi-host rows are the same physical GPU on a PCIe-constrained host, so they are not a card ceiling.
Notes
Source issue prints the llama-bench mean and the +/- spread; the stored value is the mean. The Pi-host rows are the same physical GPU on a PCIe-constrained host, so they are not a card ceiling.