Whole-PC freezes 1-12 s + ExitType 0x322327458D1ED790 kills (Javelin DPC data)
System: Ryzen 7 9800X3D, ASRock B650E Taichi (BIOS 3.50), RTX 5080 (driver 610.88, later 621.14), 64 GB, Windows 11 Pro 25H2 build 26200.9711, HVCI off, Secure Boot off, Killer E3100G 2.5GbE (1125.5.112.2022), 3 USB audio devices. Apex (Steam) updated 2026-11-05, EAAntiCheat 1.0.16371190.0.
Symptom: While Apex is running (including idle in the lobby at 1% GPU), the entire PC freezes for 1.5–12 s: all user-mode threads stop, audio stutters, and three independent logging processes (0.1 s / 0.25 s / 1 s timers) all miss ticks at the same moment. If the freeze lasts more than about 5–8 s, Javelin terminates the game: Event 1000 "ExitType: 0x322327458D1ED790 (0x39E94F9EEFD00483)" (7 kills on 11/05, 5+ on 11/06). From 10/09 to 10/31 I could play 2–5 hour sessions with only occasional kills (ExitType 0xE178DCBC5D1D0BCE, 0xC1C73E72028557A6); after the 11/02 Cipher patch and 11/05 update it became frequent again. With Apex closed: zero freezes in more than 70 hours of the same monitoring.
What the per-core sampling shows (PDH per-CPU DPC/Interrupt/Privileged time every 0.25 s + System-process thread kernel times via NtQuerySystemInformation; no handles are opened on the game or the anticheat):
- Every freeze (13/13 across 4 episodes on 11/06) coincides, within the same 0.25 s, with (a) a System-process kernel thread whose start address is ntoskrnl!IopThreadStart (i.e. created by a driver via IoCreateSystemThread) running at 100–2000% (the >100% values are the lost-tick catch-up and match the freeze length, e.g. 8.7 s in a 8.3 s freeze), and (b) DPC time pinned at 60–95% on the GPU's MSI core and/or the NIC/USB cores, plus one core at 80–96% Interrupt time with a low interrupt count (a single long ISR, not a storm). The other ~13 logical CPUs are idle yet no user-mode thread runs.
2. These IopThreadStart threads appear only while Apex runs (first seen 2 s before the game process is detected, 0 occurrences after exit; ~6/min, median interval 7.9 s). A periodic ~2-minute heavy pass (24–73% for 0.25–0.5 s, same TID for the whole session) never caused a freeze. The freezes come from different TIDs created at game start that occasionally run for seconds (e.g. at +2.2 min, +2.6 min, +41 min, +100 min).
3. Heavy Javelin thread alone (67 occurrences): no freeze. Multi-core DPC alone (30 rows): no freeze. Both together: 11/12 froze. The ACPI worker thread (acpi.sys!ACPIWorkerThread) is co-active in 11/13 freezes and 0/67 non-freezes.
4. No WHEA, no TDR (nvlddmkm 4101/153), no disk/memory pressure, GPU clocks/temps normal before each freeze.
5. Device attribution by interrupt pinning (2026-11-07): I pinned the GPU's MSI to logical CPU 14, the Killer E3100G NIC to CPU 12 and all xHCI controllers to CPU 6 (Affinity Policy, reboot). In the next freeze cluster (Apex +45 min, ended with ExitType 0x322327458D1ED790), DPC time was pinned on CPU 14 (GPU: 95–355% catch-up) and CPU 6 (xHCI: 75–137%) while CPU 12 (NIC) stayed ~0, so the spinning DPCs belong to nvlddmkm and the USB host controller, not the NIC. dxgkrnl's VidSchiWorkerThread was also at 93–1097% together with the Javelin worker (960–1218%). Two unrelated drivers' DPCs stalling at once points to a shared resource (e.g. PCI config-space serialization during the anti-DMA scan) rather than a single-driver bug.
6. After updating the GPU driver to 621.14 (clean install, 2026-11-07): the first session was clean for 74 minutes, then a freeze cluster hit at +74–78 min (28 freezes of 1.0–3.1 s in 4 min) with the same Javelin worker thread (IopThreadStart, one TID) at 100–2000%. This time the GPU's pinned core (CPU 14) stayed idle; the DPC load was on CPU 6 (xHCI USB host controllers: 65 of 90 quarter-second samples above 50%), CPU 11 and CPU 0. So the GPU driver was only one of the DPC partners; the lock-up still happens with the USB host controller driver alone. A second cluster followed at +187 min (30 freezes of 1.0–2.1 s in 8 min, same thread); in several of them per-core DPC was below 20%, i.e. the anticheat worker alone stalls the system. This points to a shared kernel resource held by the anticheat worker rather than any single device driver.
Interpretation (hypothesis): a Javelin worker (possibly the PCIe/anti-DMA inspection) holds a resource that the GPU driver's DPC (and sometimes NIC/USB DPCs) spin on at DISPATCH_LEVEL for seconds; same signature as forum threads 13525840 (FC26, nvlddmkm DPC 5.7 s spin with eaanticheat.sys active) and 13518090 (RTX 5080). Please check whether the 11/01 anticheat build changed the PCIe/driver inspection and whether a fix is planned. Happy to provide the raw 0.25 s CSVs, DxDiag, EA app logs and the Event 1000 list, and to run any additional tracing you can allow (xperf/WPR is denied while the game runs).
Note: this report was written with the help of an AI assistant (Claude) based on my own monitoring logs; the measurements and events are from my PC.
