RTX PRO 1000 Blackwell (ThinkPad P14s Gen 6) — GPU power locked at 25W on Linux, PlatformRequestHandler fails to query SBIOS

NVIDIA Open GPU Kernel Modules Version

610.43.02

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

Haven't yet tested against the proprietary driver - Blackwell laptop GPUs
(including this one) ship with the open kernel modules only as far as I'm
aware, so I'm not sure a proprietary-driver comparison is even possible
here. If it is possible, happy to test.

Operating System and Version

Ubuntu 26.04 LTS

Kernel Release

Linux ThinkPad-P14s-Gen-6 7.0.0-30-generic #30-Ubuntu SMP PREEMPT_DYNAMIC Fri Jul 31 18:22:54 UTC 2026 x86_64 GNU/Linux

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

GPU 0: NVIDIA RTX PRO 1000 Blackwell Generation Laptop GPU (UUID: GPU-85870bda-2813-18ad-bfa8-ae4ed9007837)

Describe the bug

GPU power draw is hard-capped at ~24-25W under Linux regardless of driver-reported
limits (default 35W, max 50W per nvidia-smi -q -d POWER), nvidia-smi -pl,
power-profiles-daemon mode, or ACPI platform_profile. nvidia-powerd runs
cleanly with no errors (v2.0, DBus connected), but a kernel-level NVRM assertion
at boot shows PlatformRequestHandler failing to query the platform power mode
from SBIOS:

NVRM: GPU0 nvAssertOkFailedNoLog: Assertion failed: Invalid data passed
[NV_ERR_INVALID_DATA] (0x00000025) returned from PlatformRequestHandler
failed to get platform power mode from SBIOS @ platform_request_handler_ctrl.c:2117

This fires at driver init, before nvidia-powerd negotiates anything - so the
daemon is running against an already-broken foundation. Under identical
benchmark settings, Windows 11 on the same hardware reaches at least 38.9W GPU
power draw and delivers ~47% higher FPS (57.4 vs 39.x FPS in Total War:
Warhammer III, 2560x1600, High, 66% resolution scale). Thermal headroom is not
the limiting factor - GPU temp was 63C against an 87C thermal limit ceiling
(per Windows telemetry) during the capped Linux run.

This matches a pattern reported across multiple laptop vendors on this and
adjacent GPUs, where NVIDIA's Dynamic Boost/PlatformRequestHandler mechanism
depends on a vendor-specific ACPI/WMI method that Windows OEM tooling (Lenovo Commercial
Vantage in this case) calls correctly, but which fails or is unavailable via
the generic Linux ACPI/WMI path:

• Acer Nitro ANV15-51 (RTX 4050 Laptop): identical platform_request_handler_ctrl.c
  assertion. https://forums.developer.nvidia.com/t/nvrm-failed-to-get-target-temp-from-sbios-failed-to-get-platform-power-mode-from-sbios-on-rtx-4050-laptop-gpu/375988
• HP OMEN Max 16 (RTX 5080/5070 Ti Laptop): GPU locked at 80W vs 175W on Windows,
  root-caused to the Linux hp-wmi driver skipping a required WMI unlock call.
  https://forums.developer.nvidia.com/t/hp-omen-max-16-rtx-5080-laptop-gpu-tgp-locked-at-80w-on-linux-nvidia-powerd-receives-sbios-disable-signal-board-8d41/369242
• Dell Pro Max 16 Premium, same exact GPU (RTX PRO 1000 Blackwell / GB207GLM),
  a related GSP-RM heartbeat/GC6 bug on driver 595.x affecting the same power
  management subsystem. https://forums.developer.nvidia.com/t/gsp-rm-firmware-bug-heartbeat-stops-after-gc6-exit-leads-to-fatal-gpu-loss-xid-79-and-broken-display-recovery/366591

Note: on driver 595.x, nvidia-powerd crashed on nearly every boot on this
system, consistent with the Dell Pro Max GC6/GSP-heartbeat bug above (same
GPU). Switching to 610.43.02 resolved the crash, but the underlying power
negotiation now fails differently (SBIOS query error above) rather than being
fixed.

Checked the full Lenovo BIOS changelog history for this platform (UEFI 1.03
through the latest 1.26, Apr 2025-Aug 2026) - no release mentions GPU power
delivery, Dynamic Boost, NvPCF, or ACPI/WMI power-mode fixes, so a BIOS update
is not expected to resolve this.

To Reproduce

• System: Lenovo ThinkPad P14s Gen 6, Core Ultra 7 265H, NVIDIA RTX PRO 1000
  Blackwell Laptop GPU (GB207GLM), hybrid graphics (PRIME on-demand)
• Driver: 610.43.02 (open kernel modules), installed via Canonical's
  precompiled path
• BIOS: R2WET43W (UEFI 1.25 / EC 1.15)
• Boot the system normally on AC power
• Check dmesg -T | grep -iE 'thermal|throttl|power' -> observe the
  PlatformRequestHandler/SBIOS assertion above at boot
• Run nvidia-smi -q -d POWER -> Current/Default Power Limit reports 35W,
  Max Power Limit reports 50W
• Run any sustained 100% GPU load (used: Total War: Warhammer III benchmark,
  2560x1600, High settings, 66% resolution scale) and monitor nvidia-smi
  during the run
• Observe: GPU stays pinned at P4 performance state, ~24-25W power draw,
  100% utilization, well below both the reported 35W default and 50W max
• Attempt sudo nvidia-smi -pl 50 -> fails with "Changing power management
  limit is not supported in current scope for GPU"
• Attempt switching power-profiles-daemon to performance mode -> no change
  in GPU power draw
• Attempt setting ACPI platform_profile to performance (already at max of
  low-power/balanced/performance) -> no change
• Confirm nvidia-powerd is running without errors
  (systemctl status nvidia-powerd) -> daemon is healthy, no crash
• Reboot into Windows 11 on the same hardware, same AC power, same benchmark
  settings -> GPU power draw reaches at least 38.9W, benchmark averages
  57.4 FPS vs 39.x FPS on Linux

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz.txt

More Info

Attached log is named nvidia-bug-report.log.gz.txt because I couldn't upload
a .gz file - it's the unmodified output of
sudo nvidia-bug-report.sh, just rename back to .gz before decompressing.

Additional context:

• Secure Boot: enabled
• prime-select: on-demand (hybrid graphics)
• Currently installed BIOS (R2WET43W / UEFI 1.25) is one release behind the
  latest available (UEFI 1.26, released 2026-08-11), but that release's
  changelog contains no power-management-related fixes, so I don't expect
  it to change this behavior. Happy to update and re-test if a maintainer
  thinks it's worth ruling out anyway.
• Haven't yet tested against the proprietary driver - Blackwell laptop GPUs
  (including this one) ship with the open kernel modules only as far as I'm
  aware, so I'm not sure a proprietary-driver comparison is even possible
  here. If it is possible, happy to test.
• What I expected: GPU power draw to scale up toward the driver-reported
  35W default (and ideally the 50W max) under sustained full load, matching
  what's observed on Windows on the same hardware. What actually happens:
  power draw stays pinned at ~24-25W regardless of load, with the SBIOS
  query failure in dmesg as the likely root cause.

Start with the boot dmesg assertion in platform_request_handler_ctrl.c at line 2117 and inspect the attached nvidia-bug-report.log.gz.txt for the surrounding initialization details. Compare the reported SBIOS/platform power-mode failure with the nvidia-powerd status and reproduction steps. Done means identifying and fixing the Linux SBIOS query failure so sustained GPU load can exceed the observed 24–25W cap.