inference/research
verifiedmedium · 0.70updated 2026-10-15

Argonne: PVC x1 llama.cpp Mixtral-8x7B FP16 128io b1

7.0 tok/s aggregate (batch*(in+out)/latency)

Fields

id
benchmarks/argonne-pvc-llamacpp-mixtral-8x7b-fp16-b1-128io-tps
Metric
tps_aggregate
Value
7
Unit
tok/s aggregate (batch*(in+out)/latency)
Model
mistralai/Mixtral-8x7B-v0.1
Methodology
THIRD-PARTY MEASURED (Argonne LLM-Inference-Bench arXiv:2411.00136). llama.cpp (ggml SYCL backend), FP16, 128in/128out, AGGREGATE over batch. 1 PVC tile, bs1, lat 36.34s.
Measured by
third_party
Reproducible
yes
notes
CSV column 'Intel PVC GPU' = ALCF Aurora/Sunspot Intel Data Center GPU Max (Ponte Vecchio, filed under xpu-intel-dc-gpu-max-1550; the PVC product line maps to the Max series). vLLM xpu backend, 1/2/4-tile configurations. MoE 8x7B.

Notes

CSV column 'Intel PVC GPU' = ALCF Aurora/Sunspot Intel Data Center GPU Max (Ponte Vecchio, filed under xpu-intel-dc-gpu-max-1550; the PVC product line maps to the Max series). vLLM xpu backend, 1/2/4-tile configurations. MoE 8x7B.

Cross-references