llamaperf

Unknown family

on NVIDIA RTX 5080 · Strata

Tone: mixedNVIDIA hardware
Nov 7, 2026
Throughput
100.0 t/s gen
Quant
IQ3
System RAM
128 GB

Summary

User compares Strata and Basalt on an RTX 5080 with 128 GB of system RAM, running an IQ3 quant on Windows. Strata reaches around 100 tok/s while Basalt reaches only about 80 tok/s with comparable settings. The user asks whether the slower Basalt result is expected on Windows.