Skip to content
RCreddit.com·
Not on the current live radar

A Strata fork for IBM AC922 running Qwen3.8-FN UD-Q4_K_XL is doing up to 7,357 tk/s prefill and 113 tk/s decode

AI summary

A developer forked Strata, making significant changes to Opus 5.5, to run it on an IBM AC922. This 2018-era IBM AC922, equipped with two POWER9 20-core SMT4 CPUs and 4 or 6 NVIDIA Tesla V100 SXM2 GPUs, achieved impressive performance. Running Qwen3.8-FN UD-Q4_K_XL, the system reached up to 7,357 tk/s for prefill and 113 tk/s for decode, leveraging the 150GB/s CPU-GPU bandwidth and unified memory access.

Why this one

This report details a Strata fork that achieves 7,357 tk/s prefill and 113 tk/s decode on an IBM AC922, unlike typical benchmarks on newer hardware.

Time & source

Times shown in UTC

Display time zone: UTC

Local time zone unavailable; showing UTC.

IngestedOffset at this time: UTC+0Oct 4, 2026, 21:00 UTC

Ingested
Oct 4, 2026, 21:00
Source type
Dev community

Full text isn't available here.

Read at source →
Source·reddit.com