Throttling T15g

[AlfredPaolo]

Hello, I hope to find some informations here. English is not my native language, so please excuse any potential weird phrasing.

I have the T15g Thinkpad with the following specs since a few days days: Core i7-10750H, 2x16GB RAM, 1TB SSD, Geforce RTX 2070 with a 1080p display. I installed Windows 10, updated bios/vbios and some recommended drivers with Lenovo System Update and the graphics driver via Geforce Experience. I use the SpeedShift option from Throttlestop. Maximum is set to 32 (maybe a bit conservative). Undervolting seems to not be possible with my Laptop. The voltages in the FIVR window are all locked.

Office work/surfing is generally fine. Tried to play Witcher 3 and got some weird fps drops from 60 to around 20. Even when I was just standing in the game and nothing was going on. When I was playing path of exile, the same happened. And Path of Exile is not that ressource demanding. The System is on "Maximum Performance". So I tried to analyze it further, the state of the system when the fps drops in witcher 3 occured was mostly as follows:

• GPU temp: ~69 °C
• GPU usage: ~73%
• GPU core clock: > 15xx MHz. Does sometimes go up as high as 18xx MHz (!!).
• GPU Power: ~88-90 W
• CPU temp: ~7x °C (there's one exception in the logs, where it's 94 for a second)
• CPU usage: ~30 %
• CPU power: ~ 14 - 22W with a few occasional spikes

Then the fps drops occured and power/clock rates from CPU and GPU fell. I even recorded it with OSD, The drops occur at: 1:22, 1:31, 1:41: https://streamable.com/ctsf0x
If it's too blurry to read (streamable resizes to 720p) here's the source: https://easyupload.io/4r9p5c

Intel XTU utility says there was Thermal Throttling, Power Limit Throttling and Current/EDP Limit throttling. See the Screenshot below:

I experimented a bit and disabled the "Hybrid Mode" (integrated graphics) in BIOS. I noticed, that the GPU core clock is in average much lower then with enabled hybrid mode. The throttling occurs much less, but it's still there, I saw a few "BD PROCHOT". I think the BD PROCHOT sensor is tied in some way to the GPU. When I disabled BD PROCHOT the GPU core clock still fell sometimes and the fps rate dropped, but the CPU core clock did not seem to fall that much. So even with disabled "BD PROCHOT" I still get some fps drops here and there.
With disabled Hybrid Mode and disabled BD PROCHOT Throttlestop shows these limits:

The only strange thing regarding temperatures is the "x7d" sensor in TPFanControl. Seems to sometimes go as high as ~90°C and then slowly down again. I often read, that you can ignore it and it's some nonsense value from a non-existing sensor. Didn't have that in my old Thinkpad.

I really don't understand why there is any throttling. The temperatures look fine to me (except that x7d in TPFanControl), the consumed power too. Could it have something to do with the (in my opinion) very high GPU core clocks? Nvidia specifications state 930 - 1155 MHz with boost. Mine are much higher, seems Lenovo does some overclocking there. I added some logs from Throttlestop. With hybrid mode on, hybrid mode off and hybrid mode off + BD PROCHOT unchecked. Any suggestions/ideas? Do I overlook something? Thank you in advance!

 

[unclewebb]

   DATE       TIME    MULTI   C0%   CKMOD  BAT_mW  TEMP   NVIDIA GPU     VID   POWER
2021-04-03  22:44:24  32.00   54.9  100.0       0   78     855    63   0.9211   24.3
2021-04-03  22:44:25   9.29   91.1  100.0       0   65    1080    63   0.5830    5.7   BDPRO
2021-04-03  22:44:26   8.00   87.2  100.0       0   65    1080    63   0.5880    4.1   BDPRO
2021-04-03  22:44:27  29.21   33.3  100.0       0   71    1365    64   0.9482   14.6
2021-04-03  22:44:28  31.68   21.1  100.0       0   70    1365    64   0.9352   13.0
2021-04-03  22:44:29  31.55   23.0  100.0       0   71    1425    65   0.9404   13.1
2021-04-03  22:44:30  30.88   22.6  100.0       0   71    1425    65   0.9352   12.6
2021-04-03  22:44:31  32.00   33.3  100.0       0   71    1215    64   0.9471   17.1
2021-04-03  22:44:32  32.00   26.2  100.0       0   72    1215    64   0.9528   14.5
2021-04-03  22:44:33  32.00   25.8  100.0       0   72    1155    64   0.9559   14.5
2021-04-03  22:44:34  32.00   25.9  100.0       0   71    1155    64   0.9414   14.6

Your computer has a BD PROCHOT throttling problem. Some sensor within your computer is telling the CPU to drop down to 800 MHz for a second or two before recovering. In one of your other log files, the amount of BD PROCHOT throttling happens much more frequently. The Core i7-10750H has a 45W TDP rating. Power consumption and your CPU and GPU temperatures are all OK so there appears to be no legitimate reason for any throttling like this, especially down to 800 MHz.

If you want to try and eliminate this type of throttling, you need to use ThrottleStop. Clear the BD PROCHOT box on the main screen. This tells the CPU to ignore any external throttling messages. Your CPU will still thermal throttle and slow down to protect itself if it ever gets too hot. Clearing the BD PROCHOT box only blocks the external throttling signals, not the internal ones. There is no other solution. Intel XTU does not give you access to the BD PROCHOT setting so I would uninstall XTU so it does not interfere with ThrottleStop.

There is no way to determine exactly why this happens. The engineer that designed your laptop will know but no one else in a forum or answering the phone for Lenovo is going to know what the trigger is for this type of throttling. One bad sensor misfiring can make a computer unusable.

Maximum is set to 32 (maybe a bit conservative).

The cooling system in your laptop is really good. There is no reason to slow your CPU down so much. Intel CPUs are designed to run reliably up to 100°C. That is why Intel sets the thermal throttling temperature to this value for virtually all of their CPUs. Reducing performance so you can run your CPU at a comfortable 70°C to 75°C is not necessary.

The other problem some Lenovo laptops have is they vary the thermal throttling temperature. To fix this problem, go into the ThrottleStop Options window and make sure PROCHOT Offset is checked and check the box above that one, Lock PROCHOT Offset. This prevents the randomly changing thermal throttling temperature. I have seen some Lenovo laptops drop this down to 65°C. That makes for a cool running laptop that performs miserably.

The 10750H supports a feature called Thermal Velocity Boost. This will slow the CPU down 100 MHz when it reaches 70°C. If you do not want your CPU to slow down, open the ThrottleStop FIVR window and clear the check mark beside the Thermal Velocity Boost box.

Try running another log file with BD PROCHOT disabled and with your CPU allowed to run at its rated speed. Post some screenshots of how you have ThrottleStop setup. In the TPL window, you might need to increase Power Limit 4 or you can set this to 0 so it does not cause any throttling.

 

[AlfredPaolo]

First of all, thank you for your time analyzing my problem.

The cooling system in your laptop is really good. There is no reason to slow your CPU down so much. Intel CPUs are designed to run reliably up to 100°C. That is why Intel sets the thermal throttling temperature to this value for virtually all of their CPUs. Reducing performance so you can run your CPU at a comfortable 70°C to 75°C is not necessary.

At first I set the Speed Shift max that low because I don't want to hear fans while im working. Without limitations, there are occasional spikes into the 4xxx MHz range and occasional (albeit quiet) fan noises. Thought that would be a way to get rid of that. And after noticing, that it's still more than enough for the games I played (usage was never near 100%), I just left it that way.

I cleared the BD PROCHOT checkbox and locked the PROCHOT offset box. The Thermal Velocity Boost was already unchecked (could be that I already did that and forgot). I also cleared the Speed Shift box (I guess that's what you mean with "with your CPU allowed to run at its rated speed") in the main window and the TPL window. Did not yet change the Power Limit 4. Here are the images of my Throttlestop configuration:

I attached a logfile of some gaming session with this configuration. There seem to be alot of TVBs, even though I disabled it in Throttlestop. CPU reached 100°C a few times too. CPU clock does not seem to drop, but FPS drops (to about 10/20 FPS) still happened quite a bit. The FPS drops align with drops in the GPU core clock. Am I right in the assumption that the GPU behaves strange? Don't see any problem with the power (got that from Afterburner) or the temperature of the GPU. Throttlestop displays nothing for the GPU in the Limits window.

 

[unclewebb]

there are occasional spikes into the 4xxx MHz range

Occasional spikes? Ignore the on screen CPU MHz data while gaming. It does not accurately track the CPU multiplier when Intel Speed Shift is enabled. The log file shows that your CPU is consistently running between 4200 MHz and 4400 MHz while gaming. Not sure about all of the Thermal Velocity Boost (TVB) messages in the log file. I do not think my 10th Gen desktop CPU shows any TVB messages when Thermal Velocity Boost is disabled in ThrottleStop.

The FPS drops align with drops in the GPU core clock.

Your problem now is the GPU. Early in the log file you can see your GPU boosts up to over 1400 MHz. Later on, it throttles down to only 330 MHz. This will kill your gaming performance and your FPS. Limit Reasons only reports throttling reasons for the Intel GPU. It does not track the Nvidia GPU throttling reasons. Maybe try running a log file with GPU-Z and see if it reports this information. The GPU might have a low power limit set internally.

Now that you know how to get full speed out of the CPU, it is OK to reduce the Speed Shift Max variable in the TPL window to slow it down. If you reduce the power going to the CPU, maybe there will be more power left over for the GPU. The GPU might actually run more consistently with a slower CPU or it might not make any difference.

Lenovo has been building various throttling schemes into their laptops for a long time. It is possible that the power adapter is not capable of fully powering both the CPU and GPU so they came up with a plan to throttle both of them. Fixing the CPU throttling only fixes half of the problem. Dell used to do shady stuff like this so they could ship their laptops with a smaller power adapter. If the GPU power limit is hard coded into its BIOS, using a higher rated power adapter might not make any difference.

 

[AlfredPaolo]

Thanks again for your reply!

Occasional spikes? Ignore the on screen CPU MHz data while gaming. It does not accurately track the CPU multiplier when Intel Speed Shift is enabled. The log file shows that your CPU is consistently running between 4200 MHz and 4400 MHz while gaming.

I meant while I'm on the desktop surfing or something like that (light work), the occasional spikes were display in Throttlestop. I used Afterburner for tracking the GPU/CPU ingame. The FPS drops perfectly aligned with the throttling that was displayed ingame. The GPU data from afterburner seemed to be relative useful to me, the GPU core clocks were approximately the same as in the Throttlestop log. Do you by any chance know the reliability of data from afterburner?

Your problem now is the GPU. Early in the log file you can see your GPU boosts up to over 1400 MHz. Later on, it throttles down to only 330 MHz. This will kill your gaming performance and your FPS. Limit Reasons only reports throttling reasons for the Intel GPU. It does not track the Nvidia GPU throttling reasons. Maybe try running a log file with GPU-Z and see if it reports this information. The GPU might have a low power limit set internally.

Ah, ok, makes sense that it's the integrated GPU in a tool for the CPU. I'll try to run a log file with GPU-Z, thanks for the tip. The ~300MHz rates occur when there's loading ingame or when I shortly switched to the desktop. I'm relatively sure that's no throttling. Afterburner displays rates that are around ~1080, when the throttling occurs (if this data is credible). When logging via GPU-Z I will avoid loading screens and switching, forgot that while logging before.

Now that you know how to get full speed out of the CPU, it is OK to reduce the Speed Shift Max variable in the TPL window to slow it down. If you reduce the power going to the CPU, maybe there will be more power left over for the GPU. The GPU might actually run more consistently with a slower CPU or it might not make any difference.

This seems to be the case. There are currently two things, that stabilize my FPS. 1. Disabling hybrid mode in BIOS, then the GPU runs at ~300 MHz lower and drastically less fps drops occur. 2. Limiting the CPU clock rate then leads to less fps drops, too. I did set the Speed Shift max to 32 again and did a bit of gaming, I attached the log. Within 3 hours, there were a few fps drops (maybe ~5 - 10), but it seems only one through PL2. Is the PL2 message relevant enough to try to fix it? With Speed Shift enabled, there were no TVB messages anymore. I could try to go lower with the Speed Shift max, too.

Disabling BD PROCHOT subjectively does not lead to less fps drops. In that case CPU almost never throttles with my Speed Shift settings on (only some PL2 here and there), but the GPU still throttles as much as before. So it makes sense to me, that the BD PROCHOT signal comes from the GPU.

Lenovo has been building various throttling schemes into their laptops for a long time. It is possible that the power adapter is not capable of fully powering both the CPU and GPU so they came up with a plan to throttle both of them. Fixing the CPU throttling only fixes half of the problem. Dell used to do shady stuff like this so they could ship their laptops with a smaller power adapter. If the GPU power limit is hard coded into its BIOS, using a higher rated power adapter might not make any difference.

If that's really the case I would feel ripped off. I have a 230W adapter, did not find one with more from lenovo. According to afterburner (again, if the data is at least approximately accurate) the GPU uses ~88 W and according to Throttlestop log the CPU uses at most ~22W with my Speed Shift settings. Lenovo also released a new vBIOS, which enabled the GPU to draw up to 90W, before it was up to 80W. So there should be plenty left, or do I miss some other big energy consumer?

 

[unclewebb]

Do you by any chance know the reliability of data from afterburner?

The Afterburner GPU data looks great. ThrottleStop does a better job of tracking the CPU multiplier when Speed Shift is enabled and the CPU is not fully loaded.

Disabling BD PROCHOT subjectively does not lead to less fps drops.

Disabling BD PROCHOT fixes the CPU throttling problem but it does not fix the GPU throttling problem. It could possibly make the GPU throttling problem worse. If the throttling algorithm discovers that the CPU is using more power than it should, it could decide to throttle the GPU more.

I would feel ripped off.

You should feel ripped off. When you buy a laptop, it should be able to run trouble free at its rated specs without any CPU or GPU throttling and without it overheating and thermal throttling. Too many manufacturers are cutting way too many corners. All consumers should feel ripped off. They should return their under performing laptops for a full refund. No one complains so manufacturers keep turning out duds.

Run GPU-Z to track your GPU power consumption. You might notice that your GPU is power limit throttling at less than 90W. It might be a GPU thermal limit that is causing problems.

 

[AlfredPaolo]

You should feel ripped off. When you buy a laptop, it should be able to run trouble free at its rated specs without any CPU or GPU throttling and without it overheating and thermal throttling. Too many manufacturers are cutting way too many corners. All consumers should feel ripped off. They should return their under performing laptops for a full refund. No one complains so manufacturers keep turning out duds.

I absolutely agree. I could send the laptop back, but I don't think I would find something equal for the price I paid, it was a special offer here in a students shop in germany. Are there any manufacturers, that don't do this unnecessary throttling?

Run GPU-Z to track your GPU power consumption. You might notice that your GPU is power limit throttling at less than 90W. It might be a GPU thermal limit that is causing problems.

I tried to underclock my CPU further, as I had relatively good results with the 35 multiplier before. So I did run GPU-Z while gaming. Log is attached. Throttlestop settings were the same as I posted pictures of above. Except, that Speed Shift was on with a maximum of 25. Still had two big FPS drops, at 00:03:06 and at 00:03:25. 1080 MHz is the default clock according to GPU-Z. When the GPU throttles, it seems to fall down to that speed. So I guess an indicator for occured throttling in the log is, when the GPU runs at 1080 MHz and GPU load is 100% at the same time. Throttling occured, although GPU power consumption was at ~60W, board power consumption at ~88W and GPU temperature at 69 °C (hot spot 76 °C). Don't see any problems with these values.

I then checked "Disable Turbo" in Throttlestop and set the Speed Shift max to 20. "Disable Turbo" is actually not necessary in this case as the base frequency is greater then 2GHz, right? Or is it possible that there are scenarios in which the multiplier goes over the Speed Shift max? After I limited my CPU to 2GHz I could finally play without GPU throttling. So there really seems to be an energy issue or there is potentially enough energy available and some throttling alorithm screws this up. Can not imagine, that this is unintentionally.

 

[unclewebb]

Have a look in the GPU-Z log file for the column,
PerfCap Reason

That column shows the value 1 quite often which indicates GPU power limit throttling. It seems to be triggered by board power draw which is about 85W to 88W.

Here are the definitions.

GPU-Z PerfCap Log Number Meanings

Hello, everyone. I used GPU-Z today to log the activities of my video card to a file while I was playing a game because I'm getting freezing and artifacts at times, and I was wondering what 1 of the items meant. I know PerfCap in GPU-Z means Performance Cap, and there are text values in GPU-Z...

www.techpowerup.com

Looks like Lenovo has decided to limit both the CPU and the GPU. Too bad that information was not included on the sales sheet before you bought this laptop.

Can not imagine, that this is unintentionally.

That is the hard part. When you realize that all of these throttling schemes were done deliberately and no one you ever talk to will admit that these schemes exist. This is an industry wide problem. Some thick laptops have proper cooling and can run at their rated speed. Many of the thin and light laptops that most consumers buy are throttle city. They have a hard time living up to their specs when both the CPU and GPU are active.

I guess your only option is to play around with ThrottleStop to limit the CPU performance so you do not exceed the power budget.

 

[AlfredPaolo]

Have a look in the GPU-Z log file for the column,
PerfCap Reason

That column shows the value 1 quite often which indicates GPU power limit throttling. It seems to be triggered by board power draw which is about 85W to 88W.

Here are the definitions.

GPU-Z PerfCap Log Number Meanings

Hello, everyone. I used GPU-Z today to log the activities of my video card to a file while I was playing a game because I'm getting freezing and artifacts at times, and I was wondering what 1 of the items meant. I know PerfCap in GPU-Z means Performance Cap, and there are text values in GPU-Z...

www.techpowerup.com

Thank you for that link. Yes, the value 1 appears a lot. But at times in which I got no noticeable fps drops. The only two big FPS drops I had (at that time the Speed Shift multiplier was still at 25) seem to have occured with PerfCap Reason 16.
I don't fully understand this, I wouldn't expect the GPU do drop from 1485 MHz and 82% GPU load to 1080 MHz and 88% GPU load and then stay on 1080 MHz with 100% GPU load (big fps drop occured here) with PerfCap Reason 16. Would expect that to be PerfCap Reason 1, if I indeed exceeded my power budget.

Is this still the right place or should I open a new thread in the GPU-Z section?

I guess your only option is to play around with ThrottleStop to limit the CPU performance so you do not exceed the power budget.

I fear that's true.