Thinkpad T480 throttling CPU and GPU once GPU gets hotter (?)

[erfz]

Hello,

I have the T480 with 8350U and MX150, BIOS version 1.37 (dated 4/20/21).

When I run CSGO for a few minutes, both the CPU and GPU get throttled heavily as can be seen in the ThrottleStop log file attached.
In particular, at about timestamp "21:14:32" things take a turn for the worse.
I've attached screenshots of my ThrottleStop settings; apologies for so many screenshots, but I really can't tell what's going on.
I also attached my ThrottleStop.ini as a .txt file since .ini isn't an allowed extension.
Most notably, I see some PL1/2 throttling in both the CPU and GPU after checking the Limit Reasons once the throttling is in effect.

I should also note that I'm using ASUS GPU Tweak II as mentioned in this reddit thread in order to increase the MX150 temperature limit from 70 C to 75 C.
It does seem to work properly since the log shows GPU temps higher than 70 C. Before doing this, it would stay below 70 C but downclock quite aggressively to maintain this.

I haven't disabled any of the Intel DPTF drivers or Lenovo Intelligent Thermal Solution, if that's relevant.

Appreciate the help!

 

[unclewebb]

The TPL window shows that you have set the PL1 power limit to 18W and your CPU is PL1 power limit throttling right at 18W so that is working as intended. You could try increasing that a little more to 20W or 25W. Your CPU will run a little hotter.

21:14:32

At this time the GPU starts throttling. It goes from 1620 MHz down to 800 MHz to 850 MHz. The CPU power consumption drops. I think the CPU power goes down because the GPU has slowed down so much. The load on the CPU is significantly reduced when this happens.

The GPU seemed to be bouncing off the 75°C limit for a while before it really slowed down. There might be some sort of average temperature limit that kicked in. You could try bumping your GPU throttling temperature up another 5°C. In theory, I think the MX-150 is rated to approximately 94°C so there should be a little more room to experiment. You could leave the PL1 limit at 18W. A little bit of CPU throttling is not as bad as the GPU dropping its speed in half.

GPU-Z might be able to show you some more info about the default throttling temperature. 70°C and even 75°C seems low.

No need to mess with the DPTF driver. It is not limiting you yet.

 

[erfz]

The TPL window shows that you have set the PL1 power limit to 18W and your CPU is PL1 power limit throttling right at 18W so that is working as intended. You could try increasing that a little more to 20W or 25W. Your CPU will run a little hotter.


At this time the GPU starts throttling. It goes from 1620 MHz down to 800 MHz to 850 MHz. The CPU power consumption drops. I think the CPU power goes down because the GPU has slowed down so much. The load on the CPU is significantly reduced when this happens.

The GPU seemed to be bouncing off the 75°C limit for a while before it really slowed down. There might be some sort of average temperature limit that kicked in. You could try bumping your GPU throttling temperature up another 5°C. In theory, I think the MX-150 is rated to approximately 94°C so there should be a little more room to experiment. You could leave the PL1 limit at 18W. A little bit of CPU throttling is not as bad as the GPU dropping its speed in half.

GPU-Z might be able to show you some more info about the default throttling temperature. 70°C and even 75°C seems low.

No need to mess with the DPTF driver. It is not limiting you yet.

So I increased the GPU temp limit to 80 C and went through a few matches of a lighter game (League of Legends).
The GPU never really went above 70 C (perhaps it went to like 71/72 for a few moments, but certainly not close to the 80 C limit).
The CPU TDP fluctuated between like 13-16 W; I guess this game won't really push the CPU that hard in comparison to CSGO.
Despite this, eventually the GPU did throttle down heavily despite this -- to something like 600-700 MHz. It effectively halved my frame rates from ~110-120 to ~50-60.
Nothing was out of the ordinary in the Limit Reasons -- only PL2 on Core and EDP Other on Ring.
I unfortunately did not get a log of this happening but I bet it'll look the same as what's in my original post. I'll try to replicate with logging ASAP.

EDIT: I instead replicated the scenario in CSGO since it is easier (gets hotter faster). This is with the GPU temp limit at 80 C (verified in GPU-Z). Log attached, as well as the Limit Reasons (nothing out of the ordinary I think). I did change the SpeedShift - EPP to be handled by windows on the "Best Performance" power mode, which corresponds to a SpeedShift - EPP value of 0 on my machine. But I don't think this is related. Sad..was hoping the GPU temp limit was the answer.

 

[unclewebb]

GPU temp limit at 80°C

I would try setting the GPU temp limit to the full 94°C. That is the default setting for your GPU.

You could also try running a GPU-Z log file. It might show what the reason is for GPU throttling. It almost seems like it is based on the average GPU temperature but it could be current or some other sensor that is not reported.

Whatever it is, this does not look good. You might not be able to fix this GPU throttling issue.

 

[erfz]

I would try setting the GPU temp limit to the full 94°C. That is the default setting for your GPU.

You could also try running a GPU-Z log file. It might show what the reason is for GPU throttling. It almost seems like it is based on the average GPU temperature but it could be current or some other sensor that is not reported.

Whatever it is, this does not look good. You might not be able to fix this GPU throttling issue.

I set the GPU temp limit to 94 C and logged both GPU-Z sensors and ThrottleStop this time. Go to around time 14:00:21 for when the throttling occurs.
It does seem that GPU-Z unveiled something interesting...PerfCap Reason = 1 (meaning PWR throttling on the GPU, according to this post).
Not sure why it would be power throttling though, since the voltage is constant at 0.85 V. I've attached my custom Afterburner curve below.

Also, it seems that not having this GPU throttling should (?) be possible from this reddit post. However, this person's likely to be on a different BIOS version than me (hence why I asked them). EDIT: They've responded and we are on the same BIOS.