Repository navigation
[Benchmark] RX 6700 XT (gfx1031, Windows 10) - Flash-Next IQ3_XXS self-built HIP engine: prefill 293-315 tok/s, decode 33-38 tok/s #3598
Description
What did you measure, or what went wrong
Rerun of the community benchmark on a native-Windows self-built gfx1031 engine, posted here per the 0.2.0-beta tracker reset (originally a comment on #1993, which was closed in the reset).
Hardware and software
- GPU: AMD Radeon RX 6700 XT 12 GB (gfx1031 / Navi 22)
- CPU / RAM: Core i5-12490F (6P+4E) / 64 GB DDR4-3200
- OS / driver: Windows 10 Pro 22H2 (19045) / Adrenalin 32.0.21045.11001 (2026-09-25)
- Strata: v0.1.45, engine self-built with
tools\hip\build_windows.bat(gfx1031 only); BUILD.json: ROCm 10.2.0a20260930, hipBLASLt 100500, vision none; llama.cpp snapshot 3cf03257 - Model: Qwen3.8-Flash-Next IQ3_XXS (ModelScope, SHA256-verified), MTP draft layer on, 64K context (
--max-context 65536 --kv int8 --kv-resident 32768; the speed table below was measured at 32K)
Method: the community harness from bench/results/2026-10-03-community-r9700-windows/benchmark.py (synthetic code filler + unique nonce, greedy, 256-token cap, reasoning_effort:"none"), 3 reps at 4K, 1 at 16K; numbers below are the engine's own log lines.
| prompt tokens | prefill | decode (with MTP) | TTFT |
|---|---|---|---|
| 4,196 x3 | 292.7 tok/s (292.6-292.8) | 33.9-38.1 (median 35.5) | 14.4 s |
| 16,680 x1 | 314.7 tok/s | 35.7 | 53.0 s |
- Expert load: 39.97 GiB at 1.96 GiB/s (42 s); decode expert-cache hit 74.9-75.9% (~10% of routed experts over PCIe)
- MTP drafts accepted ~60% (e.g. 145/239, 137/248, 153/221)
- Confirms the ~33 tok/s reported for this card in Community Bench: 2026-10-11-community-rx-6700-xt-qwen3-8-flash-next-iq3-xxs #1993; for reference, the archived Linux run (
2026-10-05-community-rx-6700-xt, IQ2_XS withHSA_OVERRIDEto gfx1030) measured 32.1 tok/s decode at 4K - this native Windows build, two quants higher, is on par or faster - Poor-results note: nothing failed at 4K/16K; a separate intermittent crash during long-prompt prefill is filed as a [Engine] issue
Reproduction status
Measured locally (every number above is from my machine; logs and bench-output JSON available on request).
Activity
Niko1221 commentedon Oct 22, 2026
Thanks — I'll add the self-built Windows gfx1031 RX 6700 XT numbers (prefill 293-315, decode 33-38 tok/s) to docs/COMMUNITY_BENCHMARKS.md, with the BUILD.json/ROCm details.
Niko1221 commentedon Oct 29, 2026
Thank you for measuring this and writing it up. We are building a Strata website with a page for community benchmarks, and results like these will be added there when it launches, so we are closing this here. If your numbers change with a new version, you are welcome to post them again.
Vop3p commentedon Oct 29, 2026
Thanks for the Windows gfx1031 numbers. On the official
rdna2package (no self-build needed, it covers gfx1031) a handful of settings changed a lot on Windows for an RX 6900 XT, standard protocol (#4012):vram-reserve 1536,kv-resident 32768,pcie-frac 0.2,--prefill 8192,STRATA_HIP_PROMPT_F16=1and the switch set,STRATA_HOST_CORE=siblingThe biggest steps on Windows were
STRATA_HIP_PROMPT_F16=1with the switch set (16K 489 -> 798) and then--pcie-frac 0.2with--prefill 8192together (16K 799 -> 1,076, decode 35 -> 42;autostops at 6,144 on Windows).STRATA_SH_STREAMwas worse on Windows. A PowerShell script that sets all of this up for an RX 6700 XT / 6800 / 6900 XT, with what each setting does: windows-rx6000.Developed with an AI coding assistant; every number above was measured on the RX 6900 XT Windows machine described in #4012.