Skip to content

[Engine] gfx1151 Windows measurements on engine 0.1.48: PF_F16 +12% prefill, the tuning table at its local optimum, and the bf16 GEMM gap #3861

Closed as not planned

Description

@HaiYunChuang

Measurements from one Strix Halo Windows machine, run against engine 0.1.48
(Control 0.2.3). This continues the Windows deployment notes in #3029 and the
Strix Halo discussion in #3032; every number below is from this machine and each
A/B was interleaved (A/B/A/B, fresh process per arm, median), because same-session
drift here reached 3-5%.

Three findings that may matter upstream:

  1. STRATA_PF_FUSED_KQ=1 + STRATA_PF_F16=1 is worth +12.3% prefill on a
    6.4K-token prompt on IQ3_XXS (854.3 against 760.4 tok/s, paired +12.5/+10.5/
    +12.0 across three starts), and flat at 3.2K. IQ3_S gives +8.0%. It is the
    largest untapped lever I found on this card.
  2. The shipped hipBLASLt table is already at this machine's optimum: retuning
    30 shapes picked the same solution on 22, and the engine A/B of the retuned
    table is 629.0 against 629.7 tok/s. The two bf16 N=1 rows crash
    tune_hipblaslt with 0xC0000005 (M=1 matvec), so those rows cannot be
    recalibrated at all.
  3. The remaining gap is the bf16 dense GEMM: strata_wmma_gemm_bf16 is 2.45-2.6x
    slower than hipBLASLt on the two dominant prefill shapes (22.42 against
    8.76 ms for 4096x10240x2560), and STRATA_WMMA_GEMM=1 costs 2.4% overall.

Also in the document, with measurements: DFlash2 unlocked under --serve (a
one-line gate change) is -9.5% prefill and -1.7% decode at cap 5, -15.8% at cap 8,
with draft acceptance identical to MTP; the PLE table is already IQ4_NL (90 B/row)
and --ple-io mmap/ram change nothing (+0.9% / -5.0%); engine 0.1.47 -> 0.1.48
is parity on a configuration that already sets STRATA_HC_Q8, the tuning table and
--prefill 16384.


gfx1151 on Windows: measurements against engine 0.1.48

Notes from tuning one Strix Halo machine. Every number here was measured on this
machine; where a claim comes from the release notes it says so. The point of the
document is to save the next person the dead ends, and to hand the author a
reproduction path for the two things that still look open.

0. Machine, model, method

  • Ryzen AI Max+ 395 / Radeon 8060S (gfx1151), 128 GB LPDDR5X, Windows 11 x64,
    ROCm 10.2.0a20260930, engine 0.1.48 (windows-x86_64-hip-rdna3 package),
    Control 0.2.3, started through serve/server.py (strata.exe --serve).
  • Models: swift-abl-1.5-iq3_xxs (IQ3_XXS, 77.87 GB) and IQ3_S; config as in
    STRIX_HALO_WINDOWS.md section 3 plus STRATA_PF_FUSED_KQ=1 and
    STRATA_PF_F16=1 (section 2 below).
  • Protocol: interleaved A/B (A/B/A/B/A/B), one fresh engine process per arm, one
    warm-up request, 3-4 measured requests per process, median. Prompts are one
    Chinese text repeated to 1,646 / 3,218 / 6,362 tokens; temperature 0; a unique
    nonce per request. Same-session drift reached 3-5%, so only paired numbers are
    used. Prefill is prompt_per_second, decode predicted_per_second.

1. The shipped hipBLASLt table has no headroom here

tools/hip/tune_hipblaslt over the shipped 32-row table, 3 passes, per-solution
median of the three:

  • 22 of the 30 non-degenerate rows pick the same solution as the shipped table.
  • The 8 rows that differ look dramatic in a single pass (f16 512x2560 T=8192:
    solution 2542 measures 13-52 ms against 0.7-1.6 ms for 2541) but the same
    solution is 0.4-1.0 ms in another pass. The timings are bimodal, not stable.
  • Engine A/B, shipped table against the retuned one: prefill 629.0 against 629.7
    tok/s, decode 47.5 against 46.8. No change.

Reproduce: tune_hipblaslt <cases> --tuning-out table.txt, then diff the two
tables and A/B them with STRATA_HIPBLASLT_TUNING.

Bug found on the way: the two bf16 N=1 rows (bf16 1 2560 1 4096/8192) crash
the tuner with 0xC0000005 (access violation inside hipBLASLt) for the M=1 matvec
shape. The tuner cannot recalibrate those rows; the shipped values are all a
rebuild can have. A guard that skips N=1 with a note would be better than a
crash.

2. STRATA_PF_F16 + STRATA_PF_FUSED_KQ is a real long-prompt win

Opt-in today. Measured on IQ3_XXS, 6,362-token prompt, three paired starts:

arm prefill tok/s paired differences
baseline 760.4 -
STRATA_PF_FUSED_KQ=1 773.6 (+1.7%) -2.2 / -0.8 / +3.3 %
+ STRATA_PF_F16=1 854.3 (+12.3%) +12.5 / +10.5 / +12.0 %

At 3,218 tokens the same arms are flat (618.5 against 621.5). On IQ3_S the two
switches give +8.0% (893.0 -> 964.7 tok/s at 6,362 tokens). Decode is unchanged.
Output text stays coherent; a fixed prompt at temperature 0 gave 334 against 329
characters with the same argument structure, i.e. wording-level differences, as
the rounding note says.

The path runs for IQ3_XXS (no "off"/fallback line in the log), although
STRIX_HALO.md section 5 describes the switch with IQ3_S / IQ4_NL numbers.

Suggestion: document this pair as the long-prompt setting on gfx1151 (it is
what --prefill chunks above ~4K tokens need), or enable it automatically for
T >= 4096 behind the existing KL gate. Today a reader has to find two switches
and two conditions; the gain is larger than anything else left in the table.

3. DFlash2 under --serve: possible, not profitable

The gate at src/program/generate.cpp:5478 rejects o.serve. Removing that one
condition lets DFlash2 serve (log: windows capped at 5 tokens, no fallback).
Measured with the same local 0.1.48 build in both arms:

arm prefill tok/s decode tok/s
MTP 880.4 55.0
DFlash2 cap 5 792.9 (-9.9%) 51.7 (-1.7%)
DFlash2 cap 8 792.9 (-9.9%) 46.3 (-15.8%)

Draft acceptance is the same in both arms (MTP 25-27 of 33-38, DFlash2 25-27 of
33-38), so the drafter's quality is not the problem: the loss is per-round
drafting cost (CLI: 9.49 ms/round against MTP 8.06) plus the tap write of
df2_tap_rows for every layer of every chunk (the prefill 9-10%) plus 420 MiB
of VRAM. Widening to 8 makes it worse, and kVerifyMaxT = 8 caps the window
anyway; the "wider verify" hypothesis is not supported on this model.

Suggestion: keep the gate, but the release note could say that serve is
excluded because of the taps' cost, not because DFlash2 is untested.

4. The PLE table is already 4-bit; its I/O is not the bottleneck

include/strata/ngram/ple_reader.hpp says the table is 320,001,536 rows of 90 B
(IQ4_NL, 26.8 GiB); that is what this IQ3_XXS shard is. There is no q8 -> q4
gain to take (hipfire's +14% is against its own q8 PLE).

--ple-io A/B at 6,362 tokens, three starts per arm:

mode prefill tok/s
direct (default) 863.5
mmap 871.2 (+0.9%)
ram (mlock 26.8 GB) 820.4 (-5.0%)

The default's unbuffered reads plus the 1M-row cache, page dedup and sorted
submission are already off the critical path; locking the table in RAM only
steals page cache from the experts. No change recommended.

5. The real gap is the bf16 dense GEMM

Standalone benchmark (same buffers, same GPU, rel_l2 2e-6 against hipBLAS):

shape, bf16 strata_wmma_gemm_bf16 hipBLASLt best ratio
T=4096 N=10240 K=2560 22.42 ms (9.6 TFLOP/s) 8.76 ms (24.5 TFLOP/s) 2.6x slower
T=4096 N=2560 K=2560 5.48 ms (9.8 TFLOP/s) 2.24 ms (24.0 TFLOP/s) 2.45x slower

The numbers are the tuned-table medians for the same shapes (the 64x64 WMMA
kernel versus hipBLASLt solution 1254 on the first shape). In the engine,
STRATA_WMMA_GEMM=1 costs 2.4% prefill (867.6 -> 846.6) and 1.6% decode, i.e.
the generic bf16 WMMA path is not a substitute for the library either.

This is where the remaining speed is. Prefill at 780-880 tok/s against a 250 GB/s
machine that can do 24.5 TFLOP/s in the library; another engine on the same
hardware reports 2,100-2,400 tok/s with 4-bit weights, which implies roughly a
2.4x better dense GEMM plus the smaller weight stream. Matching hipBLASLt is not
enough (it is the current path); a bf16 tile kernel would have to be ~2.5-3x
faster than the existing one to matter.

Question for the author: is a 128x256 (or 256x128) bf16 tile kernel with
split-K and double-buffered LDS on the roadmap, or is hipBLASLt considered the
ceiling for bf16 on gfx11? The 64x64 kernel is 2.5x off, and the measurement
above is a ready-made target.

6. Engine 0.1.47 -> 0.1.48 on this configuration: parity

Three paired starts, 6,362-token prompt: prefill 877.3 against 874.3 (-0.3%),
decode 51.8 against 51.5 (-0.6%). Expected: this config already sets
STRATA_HC_Q8=1 (the 0.2.3 default), already points
STRATA_HIPBLASLT_TUNING at the table, and passes --prefill 16384 rather than
auto, so none of the three 0.2.3 Strix Halo changes apply. The release note's
+6-19% / +5-14% figures are against a 0.2.2 default configuration.

7. Local changes used for these measurements

  • strata-swift-abl.json and strata-iq3_s.json: added
    "STRATA_PF_FUSED_KQ": "1", "STRATA_PF_F16": "1". Kept.
  • src/program/generate.cpp: removed o.serve from the DFlash2 gate for the
    section 3 A/B, then reverted. Not proposed as a patch as it stands.
  • Built tune_hipblaslt and a small bench_gemm_shape (links
    src/prefill/wmma_gemm.cu + hipBLAS, same allocation, event timing).
  • The locally built 0.1.48 engine needs the packaged DLL closure to start
    (0xC0000139 otherwise); the package's rocm/bin next to the exe is what makes
    it load.

8. Summary

lever measured effect keep?
18 gfx1151 defaults off -6.2% prefill, -17% decode keep on
STRATA_SH_STREAM=0 on Windows +13% decode (earlier doc) already set
hipBLASLt table retune +0.1% (noise) no change
STRATA_PF_FUSED_KQ + STRATA_PF_F16 +12.3% prefill at 6.4K, 0 at 3.2K yes (opt-in)
DFlash2 in serve (patched) -9.5% prefill, -1.7% decode no
PLE --ple-io mmap/ram +0.9% / -5.0% no change
bf16 WMMA GEMM (STRATA_WMMA_GEMM=1) -2.4% prefill no

Open: a bf16 dense GEMM that beats hipBLASLt (section 5), and the tuner's N=1
crash (section 1).

Activity

  1. github-actions commented on Oct 23, 2026

    @github-actions

    Thanks for the issue. To keep the tracker readable, every item needs an area prefix in the title and the parts of the issue form filled in. This one is missing:

    • Start the title with its area: [Engine], [GUI], [TUI], [Benchmark], [Docs] or [Other], then a short description. Example: "[Engine] decode stops after 2 minutes".

    Edit the issue (title and description) and this note goes away on its own. If it is still missing parts in 5 days it will be closed, which is not a rejection: fix it and comment /reopen.

    Details: CONTRIBUTING.md

  2. mistral13 commented on Oct 23, 2026

    @mistral13

    Are the improvements from STRATA_PF_FUSED_KQ=1 and STRATA_PF_F16=1 also present in unsloth-ud-iq4_xs?

    Do you also observe that only a single CPU core is being utilized during decoding, or is this behavior specific to unsloth-ud models?

    I'm also interested in finding the optimal parameters for this model.
    I haven't tested version 0.2.3 yet, but in earlier versions, using --mmap-experts together with STRATA_SH_STREAM=0 resulted in a decoding performance improvement of more than 10%. Have you tested this combination yet?

  3. changed the title [-][gfx1151] Windows measurements on engine 0.1.48: PF_F16 +12% prefill, the tuning table at its local optimum, and the bf16 GEMM gap[/-] [+][Engine] gfx1151 Windows measurements on engine 0.1.48: PF_F16 +12% prefill, the tuning table at its local optimum, and the bf16 GEMM gap[/+] on Oct 24, 2026
  4. github-actions commented on Oct 24, 2026

    @github-actions

    Thanks, this follows the format now.

  5. HaiYunChuang commented on Oct 24, 2026

    @HaiYunChuang
    ContributorAuthor

    On UD-IQ4_XS: I have not run that model, so I cannot confirm the gain. The F16 prompt-expert path covers IQ4_XS gate/up (native_f16_kernel) and IQ4_NL down in moe_fused_iq.cu, but UD mixes tensor types, so the only reliable check is the log plus a long-prompt A/B: with STRATA_PF_F16=1 there should be no fallback line, and a 6.4K-token prompt should be clearly faster than the same run with STRATA_PF_FUSED_KQ=0 STRATA_PF_F16=0. On IQ3_XXS the gain is 0 at 3.2K and +12% at 6.4K, so measure a long chunk, not a short one.

    On one CPU core during decode: in our runs the expert cache holds all 24,576 slots and the profile line reads CPU experts 0.00 distinct / 0.00 routed per layer, so no expert work reaches the CPU and the pool stays idle; the single busy core there is the host thread. If your line shows CPU experts above zero (fewer slots than experts, or a small VRAM budget), then --pool-tasks / --pool-workers are the knobs, and it is not model-specific.

    The configuration behind the numbers in this issue: --mmap-experts --expert-cache auto --prefill 16384 --spec 4 --spec-min-p 0.5 --lookup-chain 3 --mtp-q4 all --mtp-window 8192 --kv int8 --max-context 262144, with STRATA_PF_FUSED=1 STRATA_PF_FUSED_KQ=1 STRATA_PF_F16=1 STRATA_PF_GEMM=1 STRATA_HC_UPMIX=1 STRATA_PA_FAST=1 STRATA_HIP_WMMA=1 STRATA_SELECT_WMMA=1 STRATA_HC_Q8=1 STRATA_PF_SWITCH_MIN_T=4096 STRATA_PREFILL_STREAM_MIN=128 STRATA_SH_STREAM=0 and the shipped gfx1151 hipBLASLt table. On Windows, STRATA_SH_STREAM=0 is the one inverted default (+13% decode here).

  6. HaiYunChuang commented on Oct 27, 2026

    @HaiYunChuang
    ContributorAuthor

    Update on 0.2.7 / engine 0.1.52: the new PLE residency policy costs 15-16% prefill on this machine, and one switch removes it

    Same machine as section 0 (Ryzen AI Max+ 395, 128 GB, Windows 11, ROCm 10.2.0a20260930), source v0.2.7 started through serve/server.py, swift-abl-1.5-iq3_xxs (IQ3_XXS) with the section 2 config (STRATA_PF_FUSED_KQ=1, STRATA_PF_F16=1, STRATA_SH_STREAM=0, --prefill 16384, --mmap-experts). Prompt 6,362 tokens, interleaved A/B, one fresh engine process per arm, warm-up plus 3 measured requests, median. Engine 0.1.50 (0.2.6 package) against engine 0.1.52 (0.2.7 package), both under the 0.2.7 server.

    round arm order 0.1.50 prefill 0.1.52 prefill delta
    1 A/B 874.9 730.3 -16.5%
    2 B/A (reversed) 869.4 737.1 -15.2%
    3 A/B, STRATA_PLE_AUTO_RAM=0 878.9 870.7 -0.9%

    Round 3 paired deltas: -1.3% / -0.1% / -0.1%.

    Cause. Every 0.1.52 start logs:

    strata generate: --ple-io ram chosen by the residency policy (the table is 26.8 GiB, the expert arena still to fill 40.0 GiB and 75.8 GiB of RAM is available; ...)
    strata generate: PLE table locked in RAM (--ple-io ram chosen by the residency policy) in 21.8 s
    

    and only in the 0.1.52 arms:

    strata serve: conversation cache off: not enough free RAM to park conversations
    

    So ple_residency.hpp admits the table (26.8 GiB + arena + 8 GiB headroom fits in 128 GB) and locks it on every start. That is the mode measured in section 4 at -5.0% (--ple-io direct 863.5 against --ple-io ram 820.4). On this machine the direct reads plus the 1M-row cache were already off the critical path, and the locked table takes the RAM that the mmap'd experts read through the page cache; the conversation cache is then dropped as well. Net -15 to -16% prefill.

    Suggestion. On Windows gfx1151 (unified memory) the "table + arena + 8 GiB" test is not the right admission rule: the lock's cost is paid from the same pool the mmap'd experts read through. Either leave the policy off for gfx1151/Windows by default, or require the table to be a small fraction of available RAM. STRATA_PLE_AUTO_RAM=0 restores the direct reads and gives parity here, but a reader who does not know the switch loses 15%.

    Nothing else regresses: with the switch set, 0.1.52 is within -0.9% prefill of 0.1.50, and decode is inside this machine's same-session drift (0.1.50 51.6-58.1, 0.1.52 46.0-56.8 across the three rounds; some requests ended at 28/29 completion tokens, so the decode medians are not clean).

  7. Niko1221 commented on Oct 29, 2026

    @Niko1221
    Owner

    Thank you for the interleaved numbers: 874.9 / 869.4 against 730.3 / 737.1 tok/s, and 878.9 / 870.7 with STRATA_PLE_AUTO_RAM=0, is a clear result. We read the residency policy (ple_residency.hpp) and agree with your cause: it has no check for a GPU that shares the PC's memory (Strix Halo) or for a Windows HIP build, and it locks the table out of the pool the memory-mapped experts read through. We have not changed the default. Until then STRATA_PLE_AUTO_RAM=0 (or --ple-io direct) is the setting to use on your machine.

  8. Niko1221 commented on Nov 5, 2026

    @Niko1221
    Owner

    Thanks for taking the time to report this.

    Strata has changed a lot since this was opened, and 0.2.11 is out now, so many older reports no longer apply. We're closing issues older than a week that have had no new reports in the last week. Closed here does not mean rejected.

    To reopen it takes about two minutes:

    1. Re-test on 0.2.11. It may already be fixed. Write down what you see.
    2. Start the title with its area: [Engine], [GUI], [TUI], [Benchmark], [Docs] or [Other].
    3. Fill in the parts of the issue form: what happened, the Strata version, your graphics card, memory and system.
    4. Comment /reopen. A check runs and either reopens the issue or lists what is still missing.

    The rules are in CONTRIBUTING.md.

  9. HaiYunChuang commented on Nov 9, 2026

    @HaiYunChuang
    ContributorAuthor

    Update (Nov 3–8, same box): engine 0.1.52 → 0.1.58, the BIOS carve-out finding, and a DFlash2 negative result on a Swift fine-tune

    Same hardware (Ryzen AI Max+ 395, 128 GB, Windows 11, gfx1151). All numbers are interleaved independent-process A/B (arms' logs checked identical each round), temperature 0, the 6.4K prompt from my earlier runs, median of 3 per start. The four guards stayed in place throughout (STRATA_PF_FUSED_KQ=1, STRATA_PF_F16=1, STRATA_SH_STREAM=0, STRATA_PLE_AUTO_RAM=0).

    1. Engine upgrades

    comparison model prefill tok/s decode tok/s
    0.1.52 → 0.1.56 (on v0.2.10 source) Swift-abl IQ3_XXS 921 → 1,234 (+34%) 55.5 → 52.2 (−6%, reproduced in both orders)
    0.1.56 → 0.1.58 (on v0.2.12 source) Swift-abl IQ3_S 1,311 → 1,330 (+1.5%) 45.5 → 44.2 (−2.9%, inside noise)
    • The 0.1.56 prefill jump is visible in the log: the PLE direct-read probe went from 43,000 to 133,000–154,000 rows/s.
    • The 0.1.56 decode regression was real (reproduced reversed, clean ctok subset); 0.1.58 is back inside noise.
    • The +1.5% at 0.1.58 confirms, on real hardware, that the 0.2.12 Strix Halo prompt gains are K-quant/UD-specific: 0.1.58 adds just two gfx1151 default switches (SMALL_UNR, GR_Q8_PFS, 18 → 20), and that is all an i-quant set picks up.

    2. The BIOS UMA carve-out was worth +27–29% prefill — possibly the largest single knob on a Strix Halo Windows box

    This box shipped with UMA Frame Buffer Size = 32 GB. Your test machine runs 64 GB ("128 GB (64 GB set aside for graphics)", 0.2.10 notes). With 32 GB the IQ3_S expert cache (46.84 GiB) overflows ~15 GB into WDDM-managed memory:

    UMA Frame Buffer IQ3_S prefill @6.4K @8k long-answer decode
    32 GB (as shipped) 1,011 1,017 38–46
    64 GB (after the BIOS change) 1,286 1,310 46.9 (ctok 128, four rounds)

    One BIOS setting: +27% / +29% prefill, and most of the gap to your published 1,424 @8k closes (the remaining −8% is plausibly the Swift fine-tune and variant differences). Cost: Windows sees 63.6 GB, so the conversation cache stays off. This may deserve a line in the Strix Halo guide — a user on the default 32 GB carve-out silently loses ~25% prefill and ends up reporting "much slower than advertised."

    3. DFlash2 tuned pack on a Swift fine-tune: negative — and the exclusion is exactly right

    Swift-abl IQ3_S (SC117's abliterated Swift 1.5) is my main set, so model_name doesn't match _SUPPORTED_NAME and serve correctly fell back to MTP ("DFlash2 is for Qwen3.8-Flash-Next only"). I ran it anyway by naming the config qwen3.8-flash-next-iq3_s, which applies SET_DEFAULTS[("hip","iq3_s")]: the tuned selq4 pack (sha256-verified against the pin), conf 0.35, --spec 6 (windows capped at 5 by the engine). One setup note: the serve side finds the drafter via data_dir_of, which walks two levels up from --pack — i.e. the model data folder, not the install's data\ directory where the stock pack lives.

    MTP (--spec 4) DFlash2 tuned (conf 0.35)
    chat decode (ctok 26) 45.6 44.1 (−3.3%)
    long-answer decode (ctok ≈240) 49.1 44.9 (−8.7%)

    All three DFlash2 reps landed below all three MTP reps. Draft acceptance was only 50–58%: the drafter is trained on the stock distribution, and Swift's shorter thinking plus the abliteration shift costs more in verification than the accepted drafts return. Conclusion: Swift sets stay on the MTP — the SUPPORTED_FAMILIES exclusion is right, and this fills in the one number behind it. The tuned pack stays in the model data folder; the stock ISTA IQ3_S config matches by name and can use it any time.

    Happy to run anything specific on this box — the interleaved A/B harness is reusable.

  10. HaiYunChuang commented on Nov 22, 2026

    @HaiYunChuang
    ContributorAuthor

    Update (Nov 22): 0.2.16's VRAM regression is fixed in 0.1.65 — but the prompt path on gfx1151 + IQ3_S is now 23% slower than 0.1.60, so this box rolled back again

    Same hardware and method as before (Ryzen AI Max+ 395, 128 GB, Windows 11, 64 GB UMA carve-out, swift-abl IQ3_S, the 6.4K/8K prompts, medians).

    1. The good news first: the 0.1.62 memory regression is gone. With 0.1.65 the engine loads to 45,247 MiB of VRAM free — the same figure as 0.1.60 (0.1.62 left only 24,969; it also held ~20 GB more with everything loaded). Thank you for the quiet fix.

    2. The prompt path, however, regressed on this configuration. 0.2.14+ moved the prompt experts to Strata's own kernels, and on 0.1.60 the old guard pair still won: STRATA_PF_F16=1 + STRATA_PF_FUSED_KQ=1 was 5.5% faster than the new path (my Nov 16 numbers here: 1,332 vs 1,260 @6.4K). On 0.1.65 that old path no longer engages — the log now says "prompt experts on the fused int8 kernels" with the same config that used to take the F16 route — and nothing I tried recovers the speed:

    engine / settings prefill @6.4K prefill @8k
    0.1.60, PF_F16=1 + PF_FUSED_KQ=1 (rolled back to) 1,350 – 1,357 1,377 – 1,380
    0.1.65, same config (falls to the int8 fused path) 985 – 995 990 – 995
    0.1.65, PF_F16=2 (the new native_h11p_kernel route) 1,003 – 1,016 1,014 – 1,016
    0.1.65, guards off 936 – 950 948 – 950

    So the ceiling on 0.1.65 is ~1,015 tok/s — 23% below 0.1.60. Decode is unchanged (44–48 throughout). The release notes' RDNA3 speedups are all NVIDIA figures; on gfx1151 with an i-quant set the new kernels have not caught up with the old F16 fused route yet.

    3. What this box does now: stays on v0.2.15 / 0.1.60 (second rollback; the first was 0.2.16 for the VRAM issue). The retest trigger I keep is simple — when a release looks relevant, measure PF_F16=2 once; if it clears ~1,300 @6.4K, the new path has arrived and the guards can retire.

    Happy to run any specific build or setting you want measured on this machine — the interleaved harness takes minutes per arm.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions