Issue Summary: On a Lenovo Legion Gen 10 with an RTX 5070 Mobile, the GPU is artificially capped at 35W and locked into the P4 performance state under load. This occurs despite having 115W TGP available via the motherboard and using the mandatory nvidia-open kernel modules.
Diagnostic Findings:
Telemetry:nvidia-smi reports the GPU Shutdown and Slowdown Thermal Limits as -5°C and -2°C respectively, suggesting a failure in sensor communication with the GSP firmware on this specific chassis.
Failsafe: The GSP firmware appears to trigger a “Software Power Cap” failsafe due to the invalid thermal data, forcing the card into P4.
ACPI Spoofing: I have verified the motherboard power tables are unlocked by using acpi_osi="Windows 2022", which allowed nvidia-smi to see the 115W cap, but the driver immediately countermands this under load.
Cross-Platform: The hardware functions correctly at full TGP on Windows 11, even with the same issue in the GPU Shutdown and Slowdown Thermal Limits as reported by nvidia-smi, indicating this is a Linux-specific GSP firmware/driver telemetry issue.
System Environment:
As per dmidecode -t system:
Manufacturer: LENOVO
Product Name: 83NN
Version: Legion Pro 5 16IRX10
Wake-up Type: Power Switch
SKU Number: LENOVO_MT_83NN_BU_idea_FM_Legion Pro 5 16IRX10
Family: Legion Pro 5 16IRX10
Driver 595.71.05
Command Output: Performance State
==============NVSMI LOG==============
Timestamp : Thu May 7 20:51:55 2026
Driver Version : 595.71.05
CUDA Version : 13.2
Attached GPUs : 1
GPU 00000000:01:00.0
Performance State : P4
Clocks Event Reasons
Idle : Not Active
Applications Clocks Setting : Not Active
SW Power Cap : Not Active
HW Slowdown : Not Active
HW Thermal Slowdown : Not Active
HW Power Brake Slowdown : Not Active
Sync Boost : Not Active
SW Thermal Slowdown : Not Active
Display Clock Setting : Not Active
Clocks Event Reasons Counters
SW Power Capping : 462540 us
Sync Boost : 0 us
SW Thermal Slowdown : 0 us
HW Thermal Slowdown : 0 us
HW Power Braking : 0 us
Sparse Operation Mode : N/A
Command Output: Temperature Sensor
==============NVSMI LOG==============
Timestamp : Thu May 7 20:52:22 2026
Driver Version : 595.71.05
CUDA Version : 13.2
Attached GPUs : 1
GPU 00000000:01:00.0
Temperature
GPU Current Temp : 52 C
GPU T.Limit Temp : 34 C
GPU Shutdown T.Limit Temp : -5 C
GPU Slowdown T.Limit Temp : -2 C
GPU Max Operating T.Limit Temp : 0 C
GPU Target Temperature : 87 C
Memory Current Temp : N/A
Memory Max Operating T.Limit Temp : N/A
That definitely isn’t normal behavior for a 115W TGP setup—being stuck at 35W and locked in P4 under load suggests a driver or power management bug, not a hardware limit. The fact it’s happening with nvidia-open modules makes it even more likely a GPU firmware/driver interaction issue. Hopefully NVIDIA addresses this soon, because it severely cripples real-world performance
What do you get from sudo nvidia-smi -q -d POWER ?
Also, your log has entries like this:
kernel: NVRM: API mismatch: the client 'nvidia-smi' (pid 88721)
NVRM: has the version 595.44.06, but this kernel module has
NVRM: the version 595.71.05. Please make sure that this
NVRM: kernel module and all NVIDIA driver components
NVRM: have the same version.
Forgive me, at the time I was testing other drivers to find out whether it was a driver issue. I can assure that the result is the same even without the mismatch error.
Running the command sudo nvidia-smi -q -d POWER results in:
==============NVSMI LOG==============
Timestamp : Tue May 12 15:14:14 2026
Driver Version : 595.71.05
CUDA Version : 13.2
Attached GPUs : 1
GPU 00000000:01:00.0
GPU Power Readings
Average Power Draw : 1.00 W
Instantaneous Power Draw : 12.49 W
Current Power Limit : 100.00 W
Requested Power Limit : N/A
Default Power Limit : 50.00 W
Min Power Limit : 5.00 W
Max Power Limit : 115.00 W
Power Samples
Duration : Not Found
Number of Samples : Not Found
Max : Not Found
Min : Not Found
Avg : Not Found
GPU Memory Power Readings
Average Power Draw : N/A
Instantaneous Power Draw : N/A
Module Power Readings
Average Power Draw : N/A
Instantaneous Power Draw : N/A
Current Power Limit : N/A
Requested Power Limit : N/A
Default Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A
EDPp Multiplier : N/A
I installed the current drivers through the arch package manager, I tested the vulkan dev and other drivers by building and installing them from the open kernel modules github page.
Odd, your power limit is set to a reasonable level (100 W), but an average power draw of 1 W looks broken.
The temp sensor values shouldn’t be an issue; mine are like that too.
In your log, nvidia-powerd seems to exit with an error, which is odd: nvidia-powerd[999]: ERROR! Error(1f) in setting the short timescale limit (0)
Not really sure how to dig deeper into this. Hopefully NVIDIA or someone more knowledgeable can advise.
But in the meantime, I would test another distro just to make sure there isn’t some driver screwup from the build/experimentation you did earlier. Fedora (with RPM Fusion drivers) or Ubuntu 26 LTS (with ubuntu-drivers) would be what I’d try, as a sanity check.
After analyzing the bug report I could clearly see that GPU performance state is P0 also Max Power Limit is shown as 115W. At the time of logs capture the GPU utilization was at 24% only which does not does not demonstrate a persistent 35 W cap. Please check on latest driver if you are still seeing wrong thermal values and update here. Please revert back with correct state nvidia-bug-report to proceed further on this bug. Archiving this bug as of now.
Confirming this issue on Windows 11 - it is NOT Linux-specific.
Lenovo Legion 5 15IAX10 (Type 83F0), RTX 5070 Laptop GPU, Windows 11 Home 26200, driver 596.49 (clean install), BIOS S2CN26WW (latest), Legion Performance mode + dGPU mode, on AC.
Same corrupted telemetry as OP, and on my unit the failsafe does trip on Windows - with a reproducible pattern:
After a reboot the GPU runs normally: 75–80W, ~2300 MHz, 98% util (86–88°C sustained).
After ~25–30 minutes of sustained gaming load, power collapses in a single sample: 79W/2422MHz → 31W/907MHz, throttle reason flips to SW Power Cap (0x4), GPU locked in P4.
Key detail: the GPU then cools to 61–65°C while still under 99–100% load and remains clamped at 30W indefinitely. Only a full reboot recovers it. Power limits per nvidia-smi: min 5W / default 50W / max 115W, current forced to 30W.
Not the AC adapter (battery discharge = 0 under load), not thermal (all HW slowdown flags inactive).
So the trigger appears to be heat-soak dependent - the invalid EC thermal telemetry seems to degrade further once the chassis warms up, then the power-cap failsafe wedges permanently. This may explain why OP saw full TGP on Windows: short test, cold chassis.
I have a 15-second-interval nvidia-smi CSV log capturing the exact collapse moment (attached)