Themen / AMD / HIP
AMD / HIP
Auto topic: AMD / HIP
Issues
- #6442 [Engine] Inkling-2-Mini support (looped MoE from Thinking Machines)
- #6542 [Engine] Findings on Apple M5 Max 128 GB: Gyro-L and Hy4-Flash for coding agents on v0.2.18 (+ PRs #6538, #6539, #6540)
- #6537 [Engine] Keep evicted parked conversations on disk, also with --layer-split and --batch
- #5949 [Engine] HIP/Linux: make HSA_USERPTR_FOR_PAGED_MEM=0 the default on discrete AMD cards
- #6122 [Other] Auto drafting: a PC with several cards reports no card memory, so no default tier can apply
- #5939 [Engine] Windows CUDA12 0.1.59: Tesla V100 GPU is lost during single- and dual-GPU inference
- #6533 [Engine] A branching follow-up re-reads the whole prompt when the final turn is longer than 1024 tokens
- #6531 [Benchmark] gfx1100 (RX 7900 XTX, IQ3_S): existing switches give 1.9–3.7x prompt and 1.2x decode speedup
- #6521 [Benchmark] R9700 + Threadripper PRO 3975WX: Hy4-Flash at ~30 tok/s (SQ-64, 4K) vs. 62 tok/s in the 0.2.18 notes
- #6523 [Engine] Windows NUMA in Strata: Why Our 40-Core Server Initially Used Only 20 CPU Workers
- #6526 [Engine] gfx1200 (2x RX 9060 XT, ROCm 10.1): peer-GPU fused WMMA prefill, RDNA4 MMVQ load_int_b2/Q6KTraits, and hipBLASLt fallback notes
- #6525 [Engine] Drop in MTP head replacement trained against quantized model output: +3.7 / +4.4 % decode at --spec 3
- #5614 [Engine] RX 9070 XT gfx1201: 0.1.58 verify unspecified launch failure triggers MES REMOVE_QUEUE GPU reset
- #6435 [Benchmark] RX 9070 (gfx1201, Windows 11, 32 GB RAM) / Gyro-S TQ1_0 / 0.2.6 → 0.2.17: 0.2.17 defaults lose ~30% prompt and ~12% decode on a 32 GB PC; two settings bring it back
- #6511 [Benchmark] 2x Tesla P40 (sm_61) split: 0.2.17 gives +9% decode, but 220K prefill is still ~5% under the 0.2.13.1 MMQ path (same-day A/B)
- #6431 [Engine] 0.1.63 STRATA_L2_PF=1 default regresses NVFP4 batch decode by 8-12% on 2x RTX 5090 with PCIe-streamed experts
- #6515 [Engine] gfx12: `nw_barrier()` falls back to `__syncthreads()`, so the fused prompt experts' stage writes can race (latent)
- #6513 [Engine] Linux HIP: engine crashes at start (SIGSEGV in libamdhip64) with CachyOS's hip-runtime-amd 10.1 v3/znver4 builds; Arch's build works
- #6273 [Engine] 16 GB cards: guard the ~1 GB VRAM floor during KV growth and document the per-format context limit
- #6232 [Engine] 384K compaction re-reads 264K tokens (456 s): content-addressed, position-independent segment cache for near-zero-reread compaction
- #6074 [Engine] 0.1.60 prompt kernels: -13% at 20,000 tokens on a 16 GB card, +8% at 40,000, flat at 16,000
- #6483 [Engine] Long Gyro-S decode triggers HIP 719 / LiveKernelEvent 141 on RX 9070 XT
- #6500 [Engine] Vulkan backend: scope for the port (stage 1 landed: the loop and one Q4_0 kernel)
- #6327 [Engine] Compaction re-reads 331K-342K tokens five times (19.5-20.1 min each): the cache budget is the blocker (PR #4338)
- #5665 [GUI] Heavy typing lag inside the app window only; same UI smooth in browser (0.2.12, Wayland)
- #6496 [Other] Windows HIP rdna2 package (0.1.63) works on a real RX 6900 XT (gfx1030)
- #6493 [Engine] serve loop: GENI prompt can write past the position table; --cvec offsets unchecked (static analysis)
- #6491 [Other] Ray, an assistant and coding agent built on Strata: would any part of it help the app?
- #5787 [Engine] serve: python.exe dies natively under long agent sessions — every dump faults inside CPython on the same nested `any(genexpr)` path (`serve/frontend.py` `_has_tag`)
- #5751 [Engine] Model load hard-powers-off the whole PC on RTX 5070 Ti (Kernel-Power 41, no dump): autotune variants and 0.1.58 crash; 0.1.46 + conservative VRAM settings stable
- #6487 [Engine] 0.1.62: --vision cpu fails deterministically at startup with "kv_stream: reset: an illegal memory access" (sm_86, Swift IQ3_S) — --vision gpu on the same box is fine
- #6409 [Engine] Qwen3.8-Next-Flash on Strix Halo AI Max 395 (128GB UMA): tuning for 2 concurrent streams at >100 tok/s each @ 200k context
- #6478 [Engine] Expert-cache hit rate and decode throughput regress from engine 0.1.52 to 0.1.62/0.1.63 on two 11 GB Turing cards (Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS)
- #6472 [Benchmark] RTX 4090 24 GB VRAM headroom for long-context KV and IQ4 experts
- #6404 [Engine] gfx1030 (RX 6800 XT, Windows): 0.1.63 reads prompts 2x faster than 0.1.60 but decodes 11 % slower — same config, A/B on one host
- #5672 [Engine] gfx1100 (RX 7900 XTX) --adapt-async 1 results in timeout followed by a display-driver reset during long generations
- #5953 [Engine] Qwen3.6-35B-A3B fails at startup on gfx1150 (Linux, ROCm 7.15): prefill gemm cublasGemmEx f16 - cuBLAS status 6
- #6429 [Benchmark] hipBLASLt tuning table by default on gfx1100 (RX 7900 XTX, IQ3_S): speed, decode and logits
- #5990 [Engine] V100 (sm_70) on Linux: 13× Xid 79 + PCIe CmpltTO across engines 0.1.35–0.1.59.1 — not version-specific
- #6208 [Engine] 0.2.15 decodes 28-36% slower than 0.2.7 on RTX 3060 12 GB: --expert-cache auto starts with a much worse resident set (same pack and profile)
- #5995 [Engine] Exit code 0xC0000409: stack buffer overrun on V100 + RTX A4500
- #6421 [Benchmark] Ryzen 7 5700X, 64GB DDR4-2400, 2x RX 7800 XT 16GB (Ubuntu/ROCm) -- Qwen3.8-Flash-Next UD-IQ4_XS (256K, pinned 41.4 GiB, --spec 4)
- #6145 [Engine] Deterministic #879 with --ple-io direct on UD-Q6_K_XL (mmap passes; expert blob bytes certified clean; boot-state dependent)
- #6160 [Engine] STRATA_PIPED_MTP is turned off on Linux CUDA by the engine's own parallel defaults (STRATA_BATCH_SELF_COMMIT=1)
- #6175 [Engine] the 4-bit KV cache (q4u32) as the default on RDNA4 — scope agreement
- #6182 [Engine] Metal: --prefill auto sends 1024-token calls, so the hostpipe never overlaps the GPU (~10-12% slower prefill)
- #6413 [Benchmark] v0.2.17 on 2x Arc Pro B60 (layer split) and 1x Arc Pro B70 / Qwen3.8-Flash-Next UD-Q4_K_XL / 262K: same speed as v0.2.16, `--pipeline-windows 2` no longer produces `!!!!` (12 of 12)
- #6410 [Engine] Scope an opt-in compact two-slot mode for resident Metal
- #6392 [Engine] Linux HIP RX 7900 XTX (32 GB RAM, IQ2_XS low-RAM): engine hangs within 4-10 min of sustained serving, verify timeouts, also on 0.1.63
- #6088 [Engine] Nvidia v100 performance slow down -12% with "own kernels"
- #6244 [GUI] Changing an option on the model screen silently drops drafting: dflash2 and resets STRATA_RESIDENT_PIN to 0
- #5937 [Engine] Windows/gfx1030: working set doubles after ordinary serving (19.5 -> 36 GB) and never returns to baseline
- #6402 [Engine] serve: /v1/models reports the context window but no output-token limit - clients guess the answer cap
- #6269 [Engine] gfx103x: prompt throughput gaps left by the MMQ removal (dense projections, IQ4_XS down)
- #6265 [Engine] layer-split auto: the #2644 clamp picks a slower split on asymmetric same-VRAM pairs (measured +10% hand-pinned)
- #6264 [Engine] RDNA2 gfx1030: 37% prefill regression persists in 0.1.62 — MoE GEMM gate/up 3.4x slower than 0.1.59
- #6254 [Engine] Cold-boot first load of Qwen3.8-Flash-Next-Gyro-M crashes with 0xC0000005 on RX 9060 XT (gfx1200)
- #6226 [Engine] Hardware support: AMD Radeon 890M Graphics not recognized
- #6153 [GUI] App Host dies with V8 "Reached heap limit" (JS heap OOM) after ~9 h uptime — repeatedly on Arch Linux
- #6130 [Engine] RX 7900 XTX (gfx1100, Linux), 0.2.15: picture text misread ("STRATA 412" for "STRATA 42") after earlier requests on the default HIP build; the -DSTRATA_MMQ=ON build reads it right
- #6086 [Other] Terminal: shell tab completion for the strata command
- #6325 [Engine] RDNA4 16 GB + 6-core CPU, Gyro-S: the automatic PCIe share (0.90) is ~3 % slower than --pcie-frac 1.0; with --spec 6 --draft-conf 0.6 decode is +7.5 %
- #5996 [GUI] helpers on one card: three read-only helpers ran 163, 266 and 351 minutes, with no cost warning and no ceiling
- #5991 [Other] the shipped Harness 0.2.1-alpha.3 has no model-switch pressure check, added upstream in dsh-v0.2.1-alpha.4
- #5950 [Engine] STRATA_COALESCE_PREFILL never co-reads on a piped layer split or for a burst reaching an idle server
- #5936 [Engine] GB10 (sm_121): batched prompt attention hangs at context >= 64 tokens, watchdog kills the engine (Edge0-145B-NVFP4, 0.1.55) [AI-generated report]
- #5841 [Engine] --mtp-q4 all overflows the draft head when the output head is IQ4_XS (GB10 / Qwen3.8-Flash-Next IQ2_XS): "mtp: writing the draft head", engine exits 1
- #5822 [Engine] Flash-Next 0.1.58: native QSA indexer (batch) rejects arguments and exits on single-GPU CUDA
- #5789 [Engine] Windows HIP (gfx1201, RX 9070): `strata-device.exe --selftest` crashes 0xC0000005 in amdhip64_7.dll — engine 0.1.58 and 0.1.59, Adrenalin 26.10.2
- #5738 [Engine] Windows HIP (gfx1030): STRATA_ADAPT_OVERLAP=1 fails the first request with "never rang (unspecified launch failure)" when the whole expert arena is pinned
- #5718 [Engine] gfx1200 (RX 9060, Linux): "verify: timed out" (#267) during decode, not tied to spec decoding, RAM budget or pool workers
- #5653 [Engine] SYCL (Arc): systematic instruction-following gap vs CUDA on identical weights — 2 % vs 77 % on paired probe
- #5649 [Engine] gfx1200 (Radeon RX 9060 XT): engine intermittently stalls reading long prompts
- #5612 [Engine] conversation cache: on a layer split, a sub-agent that matches only a shared head takes the whole parked conversation and drops its checkpoints
- #5608 [Engine] RX 9070 XT gfx1201: repeatable GPU page faults during IQ3_S inference on HIP 0.1.57/0.1.58
Pull requests
- #5699 [Engine] session-identity: the whole STRATA_* environment and the device plan join the config fingerprint (#5379)
- #4310 [Engine] session files: format v2 with per-stage save/restore under --layer-split
- #6520 [Engine] Inkling-2-Mini from the official NVFP4/MXFP8 checkpoint, expert offload and CPU experts for 8-16 GB cards
- #6210 [Engine] Opt-in P2P q8_1 decode push, fused P2P down-projection, and GPU peer sync flag
- #6529 [Engine] Volta sm_70 decode optimizations, zero-VRAM in-place Q8_0, NVLink split head, and volta-forever detection
- #6548 [Engine] Add opt-in streamed-KV QSA slot-group batching, pre75 multi-slot attention, and batch SIDE_MIXER
- #6547 [Engine] Add opt-in multi-slot batched MTP drafting (STRATA_BATCH_MTP_FUSED)
- #6546 [Engine] Opt-in resize_for_overwrite and parallel host memcpy for kv_mode != 0 park/restore
- #6545 [Engine] Size STRATA_PARK_ASYNC VRAM stage by device-resident bytes under kv_mode != 0
- #6544 [Engine] Fix O(N^2) scan in getline_fd and avoid full-line copy in BGEN/parse_i64_list
- #6543 [Engine] Guard negative page_table entries in qsa_prompt_attn and track monotonic kv_stream overflow_count
- #4878 [GUI] Keep quiet desktop chat streams alive
- #6524 [Engine] SYCL: enable the conversation cache's SSD tier (0.2.17's ConversationSpill)
- #6439 [Engine] Persist the SSD conversation tier across restarts when the model fingerprint matches
- #6305 [Engine] fingerprint the resolved MTP files, so the SSD conversation tier stays on for GGUF and rt drafters
- #6540 [Engine] Hy4-Flash: SQ-96 starts on 128 GB Macs, --info no longer crashes, session reuse on by default, reasoning_effort max in configs
- #6539 [Engine] Hy4-Flash (Metal): resume at the longest common prefix, several conversation slots, SSD tier kept across restarts
- #6538 [Engine] SSD conversation tier: fingerprint the resolved MTP files and keep the tier across restarts (v0.2.18)
- #6534 [Engine] sycl: keep the branch point a follow-up can reach
- #4411 [Engine] session-file saves: ping-pong blocks and a writer thread so the source copy overlaps the previous block write
- #4307 [Other] serve+webui: unload that works — pre-staged checkpoints, polite waits, honest answers
- #4306 [Other] serve: session ladder — conversations survive an engine unload
- #3553 [Engine] let a HIP helper read pinned expert rows in place
- #3552 [Engine] skip file-tier experts while a helper GPU is attached
- #3551 [Engine] add an opt-in CUDA and HIP hybrid build
- #4551 [Engine] session-file v3: incremental append saves under --session-save-append
- #3773 [Engine] q36: MTP drafts for sampled requests behind STRATA_Q36_MTP_SAMPLED (stacked on #3678)
- #6536 [Engine] fix #3354: keep default API requests greedy under MTP, gate sampled drafting behind STRATA_Q36_MTP_SAMPLED
- #6532 [Engine] prefill: opt-in STRATA_I8_DEFER - unpicked streamed entries no longer cut the int8 expert groups into fragments (gfx1030 4K prompt +3.6%)
- #5897 [Engine] serve: let the "vision" section carry its own "env" for the image encoder
- #3800 [Engine] Token mask for JSON response formats: MTP drafts survive constrained decoding
- #5436 [Engine] opt-in STRATA_ARENA_NUMA: place the expert arena per NUMA node on two-socket Linux
- #6399 [Engine] Prefetch routed expert weight rows into L1 on GB10
- #3607 [Engine] native Windows support for Intel Arc SYCL, rebased onto 0.2.18 (engine 0.1.64)
- #6514 [Benchmark] RX 7600 XT (gfx1102): ROCm 7.2.4 vs 10.1, and the CachyOS ROCm 10.1 crash
- #4928 [Engine] long --lookup-max-t windows over a layer split (opt-in, single conversation)
- #6510 [Engine] --experts-wex4: read the routed experts from a checked WEX4 container (opt-in, file tier)
- #6102 [Engine] use runtime expert count in layer-major prefill
- #4619 [Engine] DS4_LOOKUP: draft a round from the request's own history (opt-in)
- #4376 [Engine] dsv41: the WMMA expert kernels compile on pre-Volta cards
- #4366 [Engine] DeepSeek prompt: the busy experts take FP16 GEMMs, not the tiles (DS4_FAT_ROWS, opt-in)
- #6197 [Engine] int8 expert kernels: IQ4_XS as a down format
- #6508 [Engine] q36: resume from the longest common prefix with GDN state checkpoints — 0 → 12,288 reused, 37.1 → 2.9 s
- #6468 [Benchmark] RX 9070 XT on Windows, Gyro-S at 131K: app 0.2.17 / engine 0.1.63 with a tuned config (standard 12/12)
- #6176 [Engine] q4u32: the 4-bit KV cache with a fitted step as the default on RDNA4
- #6501 [Engine] Give each layer-split stage its own two-pass check buffers
- #4255 [Other] serve: refuse a port another program holds on 127.0.0.1, and say which program
- #6490 [Engine] Load pack weights directly into shared Metal storage
- #6486 [Engine] serve: an mcp__ prefix added to a plain tool, or dropped from an MCP tool, is that tool
- #6419 [Engine] autonomous hardware-synchronized dual-GPU decode window (P2P zero-wait) [Draft speed Lab 230 tok/s+ RFC]
- #6485 [Engine] Add experimental gfx90a HIP build
- #6481 [Engine] keep the baked qwen think block when preserve_thinking is unset
- #2965 setup: a separate opt-in layers 8-33 projection vector for the original Qwen3.8-Flash-Next (any quantization)
- #3957 [Engine] serve: a long conversation survives an engine restart - parked conversations saved at a graceful stop and loaded at the next start, to files or (Windows) kept in RAM with no disk writes (opt-in STRATA_PARK_DIR, STRATA_PARK_RAMLIKE=1; layer split included)
- #6474 [Engine] serve: keep a restarted engine's batch slot state out of old request cleanup
- #6471 [Engine] serve: send STOP when closing a batch admission mid-prompt
- #5476 [Engine] gfx103x decode: opt-in one-warp MMVQ for K <= 4096 and a per-format expert layout table (STRATA_MMVQ_NW1, STRATA_EXP_TABLE)
- #5475 [Engine] opt-in int8 copies for the persistent hc read (STRATA_HC_PERSIST_Q8, STRATA_HC_I8ONLY)
- #2843 [Engine] Slot checkpoints for clients that rewrite the end of their history (3/3, depends on #2527 and #2528): STRATA_SLOT_CHECKPOINTS, STRATA_BRANCH_CHECKPOINT, BCHECKS
- #2528 [Engine] Slot placement policy for STRATA_SLOT_CACHE_AFFINITY (2/3, depends on #2527): protection by history length, retention scores, value decay, no wait
- #2527 [Engine] Keep conversations in pipelined batch slots (1/3): active-row windows, park=, cache-aware placement behind a policy interface
- #1614 [Engine] shared KV pool for all lanes (--kv-pool-tokens): at 2 x 256k the first token comes 2-5x sooner and lane moves ~10x faster; off = identical to main
- #6465 [Engine] bound host RAM during conversation swaps
- #6464 [Benchmark] Strix Halo 64 GB unified memory (Ryzen AI Max+ 395)
- #6453 [Engine] SYCL: an integrated Intel GPU is planned as one pool of memory with the system (Arc B390)
- #6452 [Engine] serve: an on-demand CPU picture reader starts without asking the expert cache for VRAM
- #6393 [Engine] admit oversized conversations to the existing SSD cache
- #6458 [Engine] STRATA_FILL_STAGED: copy mapped blobs to the GPU via pinned buffers
- #6455 [Engine] SYCL: an Xe3 GPU (Panther Lake, Arc B390) runs the Battlemage kernels; STRATA_INTEL_GEN sets the class
- #4334 [Engine] STRATA_HOST_TIMELINE host-timeline split + summary-time draft-profiler flush (opt-in, measurement only)
- #3919 [Engine] kv: size the resident ring from the run's depth (opt-in, STRATA_KV_RING_ADAPT)
- #3729 [Engine] SYCL: flag Intel iGPUs as integrated so the expert cache is sized from MemAvailable
- #3648 [Engine] --peer-device host route: STRATA_PF_PEER_PIECES=1 (opt-in) moves the activations and the sums in 8 pieces (+7 % prompt on 2x V100)
- #6432 [Engine] Reuse Metal resident prefill scratch buffers
- #6441 [Engine] Reduce resident Metal QSA prefill scratch