CPU goes from 4 GHz to 798 MHz for no apparent reason, disabling power throttling not working

[Blaez]

In some circumstances, my CPU suddenly starts throttling from 3.5-4.5 GHz all the way down to 798 MHz, for no clear reason. Sure, the CPU was running a bit hot before that within the range of 77-91°C, but why does that make it drop down to a fifth of its performance, resulting in a stable 70°C? At that point it seems to be running at around 15-17 W, while PL1 seems to be 45 W and PL2 109 W, both well above the power actually going to the CPU.

For context, this happened a few times while playing Age of Empires IV, as it went from nicely playable on 4K with High graphics settings to a frustratingly unplayable ~5 FPS which does not improve even if I go down to FHD resolution, as it is so bottlenecked by the CPU at that point. After that it also seemingly randomly shifts between these two states, having good fps one second and very low the next. To be honest even a stable 2 GHz would sound better than this.

The device is a Dell XPS 17 9710, with an Intel Core i9-11980HK CPU and NVIDIA GeForce RTX 3060 Laptop GPU.

I have had what seemed to be a very similar issue with an earlier generation of this device in the past. However, I am pretty sure back then the throttling happened when the temperatures were higher and then the PL1 did shift down I think to 15W, explaining why the throttling was happening. Back then, ThrottleStop worked perfectly to eliminate this power throttling entirely, which made things run a bit hotter and resulted in one or two thermal shutdowns over a year but at least it worked.

Sadly with this device and its 11th-gen CPU, despite having played around with all the settings ("Disable and Lock Turbo Power Limits" and all the others I have seen mentioned in relation to it), ThrottleStop never seems to have an effect on the power limit throttling and seems to be unable to prevent it. Does anyone know why, or how I could try to make it work?

 

[unclewebb]

Dell XPS

The Dell XPS line has had throttling problems for over 10 years. It is the reason why ThrottleStop was created.

Back then, ThrottleStop worked perfectly to eliminate this power throttling entirely

Those were the good old days. For years, Dell would develop a new throttling scheme and I was able to add a new feature to ThrottleStop to prevent the throttling. I think the latest Dell laptops use an embedded controller to force feed low power limits to the CPU. There is nothing ThrottleStop can do to prevent this new throttling method.

Your screenshot shows that the MSR and MMIO power limits are being ignored. This confirms that a third set of power limits are being used. ThrottleStop does not have access to this third set of power limits.

The 11980HK has a 45W TDP rating.

Intel® Core™ i9-11980HK Processor (24M Cache, up to 5.00 GHz) Product Specifications

Intel® Core™ i9-11980HK Processor (24M Cache, up to 5.00 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.

ark.intel.com

Your screenshot shows that the Dell throttling algorithm is PL2 power limit throttling your CPU down to only 22.5W which, coincidentally, is exactly half of its rated TDP. Is there any Dell power plan / fan control software running on your computer? Make sure this is set to Ultimate performance or whatever the highest performance option is. If you cannot find a way to run your 45W CPU at its full 45W indefinitely then you need to contact Dell. Tell them that you are not impressed with being robbed of the performance that you paid for.

all the way down to 798 MHz

Completely exit HWiNFO. Turn on the ThrottleStop Log File option and go play a game for at least 15 minutes. When finished testing, exit ThrottleStop so it can finalize your log file. Your log file can be found in the ThrottleStop / Logs folder with today's date in its name. Attach one of those to your next post if you want me to have a look.

Lowering the turbo power limit to 22.5W is bad. Some Dell laptops are doing something far worse. It seems when the average temperature gets high, the power limit is forced to 5W or less. This causes the CPU to crawl along at its minimum speed of 798 MHz. The log file should confirm this throttling method. ThrottleStop can no longer be used to fix what Dell is doing but at least it can document this problem.

The last person I tried to help with a similar throttling problem contacted Dell and was told that this throttling was "by design". Don't accept that explanation. Ask to have this matter escalated.

 

[Blaez]

Thank you very much for the prompt reply, I really appreciate it.

Completely exit HWiNFO. Turn on the ThrottleStop Log File option and go play a game for at least 15 minutes. When finished testing, exit ThrottleStop so it can finalize your log file. Your log file can be found in the ThrottleStop / Logs folder with today's date in its name. Attach one of those to your next post if you want me to have a look.

I am attaching three parts of the log file. It turns out that while I was taking my original screenshots with the HWiNFO charts I already had ThrottleStop logging turned on, so I am attaching that old one as well just in case it helps comparing it to the charts. ("ThrottleStop-originalWithHWiNFO.txt")

I previously had the Thermal Mode set to Optimized in the Power Manager section of My Dell, as I tried Ultra Performance multiple times in the past but it didn't seem to make a difference. Just to compare, I logged about 15 minutes again still with the Optimized power settings ("ThrottleStop-newOptimizedPower.txt") and then changed to Ultra Performance, restarted the laptop twice just in case and logged the same situation again. ("ThrottleStop-newUltraPerformancePower.txt") Both of these were logged with no application running other than AoE IV and ThrottleStop.

I believe in all of these three scenarios, the CPU power went from hovering around 20-30 W (with occasional peaks around 40-45 W) down to 15-19 W which is when things started lagging really bad.

The 11980HK has a 45W TDP rating.

One thing to note about this laptop specifically (the XPS 17) is that it has a more power-hungry GPU than its XPS 15 counterpart. The RTX 3060 is set to run at 60 W in this device, while the power cable is still limited to taking 130 W in total. This could be an issue if the CPU was taking a lot of power as well and Dell has now configured it so it can drain some battery to go over 130 W temporarily but not too much as there was a lot of negative press about this in the past. (NotebookCheck article about the original battery drain issue / NotebookCheck article about the issue being fixed) This should be no reason to throttle all the way down to 15 W however, which took the laptop's total power usage to around 95 W..

If you cannot find a way to run your 45W CPU at its full 45W indefinitely then you need to contact Dell. Tell them that you are not impressed with being robbed of the performance that you paid for.

I bought this device with "ProSupport" aimed at business users and they're fairly competent, I am more than happy to reach out to them. I am not really interested in returning it though as sadly there is no alternative device I'd be happy with, but hoping that by escalating the matter eventually they could come up with a solution that lets us work around this limitation to some extent.. Maybe that's too optimistic but I guess it is worth a try.

 

[unclewebb]

@Blaez - All three log files show the same problems. The CPU is reaching 100°C and thermal throttling. This can happen even when the CPU load is as little as 10%. You could ask Dell to replace the thermal paste. This might not make any difference though if the real problem is a heatsink that is completely inadequate. Some heatsinks look more like a folded piece of tin foil.

There is lots of power limit throttling down to 15W. It looks like the design engineers knew the cooling system was not adequate so they included some severe throttling schemes in the hope that no one would notice. You noticed.

There is no fix for terrible design. The 11980HK is too much CPU to install in a thin and light chassis. This CPU cannot be fully utilized. Not even close. It is silly to also include a 60W Nvidia GPU. Either the CPU or the GPU or both will be forced to throttle to prevent a melt down.

The ProSupport team may be competent but I do not see any way that they are going to be able to re-engineer this laptop model. They cannot release a BIOS update that disables all of the throttling schemes. If they did that, there would be complaints about constant sky high temperatures.

 

[Blaez]

@Blaez - All three log files show the same problems. The CPU is reaching 100°C and thermal throttling. This can happen even when the CPU load is as little as 10%. You could ask Dell to replace the thermal paste. This might not make any difference though if the real problem is a heatsink that is completely inadequate. Some heatsinks look more like a folded piece of tin foil.

There is lots of power limit throttling down to 15W. It looks like the design engineers knew the cooling system was not adequate so they included some severe throttling schemes in the hope that no one would notice. You noticed.

There is no fix for terrible design. The 11980HK is too much CPU to install in a thin and light chassis. This CPU cannot be fully utilized. Not even close. It is silly to also include a 60W Nvidia GPU. Either the CPU or the GPU or both will be forced to throttle to prevent a melt down.

The ProSupport team may be competent but I do not see any way that they are going to be able to re-engineer this laptop model. They cannot release a BIOS update that disables all of the throttling schemes. If they did that, there would be complaints about constant sky high temperatures.

I was fortunate enough to be able to compare it to another identical machine earlier and there did not seem to be much difference between them in terms of performance or thermal/throttling behaviour. Because of that I imagine this is not an isolated incident and better paste would probably only help marginally, though it may be worth a try. Also, this device has a vapor-chamber which should work slightly better than a regular heatsink though I think it's also a bit harder to work with as there seems to be a lot more surface with thermal paste/pads from what I have seen.

What else do you think can be done to mitigate the issue this causes in games?

1. Is there no way at all to eliminate this kind of throttling happening to the CPU? I've seen some other third party solution mentioned in a different thread to change PL1 and PL2 levels for 11th gen processors where ThrottleStop was not working. Is that different since in my case it does not seem to be an issue with PL1 or PL2 but Dell's own limitation? So the only solution would be to replace the BIOS? If so, how feasible could it be to reverse-engineer and modify that small part of the BIOS, has something like that ever been done before in a similar scenario? (I'm a skilled software engineer myself with a lot of determination though not so much experience in this area..)
2. Is it possible to undervolt this CPU, and if so how? I think you mentioned in a different thread that it is possible to undervolt 11th gen H and K processors, but I'm not sure how. I used XTU in the past but on this device it does not seem to allow undervolting, even despite the processor being the "unlocked" HK version.. When I reached out to Intel in the past asking about this they said it would depend on Dell's implementation and they could not guarantee anything.
3. Could it make sense and would it be possible to create a workaround where the CPU is manually throttled a bit to prevent overheating, so that Dell's aggressive throttling never kicks in? If there's no better solution, at least by downclocking the CPU entirely, if that prevents it.

 

[unclewebb]

it does not seem to be an issue with PL1 or PL2 but Dell's own limitation

That is correct. There is no easy way to reprogram an embedded controller (EC). I have not seen anyone solve this problem. RW Everything might allow you to access the EC.

replace the BIOS

I do not think it is possible to modify the BIOS on any recent Dell laptop. You might end up with an expensive brick if you try. Pretty sure the BIOS is locked down and protected from any tampering.

undervolt this CPU

It is possible to undervolt an 11th Gen H series CPU. Open the FIVR window and see if it says Locked at the top. If Dell has locked this feature, then you need to modify a couple of UEFI variables. Once again you need to be careful not to brick your laptop. Here is the general procedure explained.

Undervolting 2020 Dell Laptops like the Vostro 7500 and More Tips to Improve Thermals, Battery Life, and Speed - Brendan Greenley

2021 Update: Based on post and Reddit comments, the approach to unlock undervolting works for a number of models, including the XPS, Inspiron, G5, 11th Generation Intel processors, and 2021 Dell models and other lines. If it works for you, please leave a comment below so others may know...

brendangreenley.com

it would depend on Dell's implementation and they could not guarantee anything

That sounds like something Dell would say. Dell generally frowns on anyone that wants to do any mods to one of their laptops. They forget that it is your laptop. You paid for the laptop and you own it.

create a workaround where the CPU is manually throttled

This might work. Other users have done similar things. If you can reduce the average temperature, you might be able to avoid the severe throttling. Try running the CPU with Disable Turbo checked on the main screen. This will slow the CPU down significantly. It will reduce power consumption and heat. Run another log file to see if this makes your laptop a little more useable.

 

[Blaez]

I do not think it is possible to modify the BIOS on any recent Dell laptop. You might end up with an expensive brick if you try. Pretty sure the BIOS is locked down and protected from any tampering.

I see.

It is possible to undervolt an 11th Gen H series CPU. Open the FIVR window and see if it says Locked at the top. If Dell has locked this feature, then you need to modify a couple of UEFI variables. Once again you need to be careful not to brick your laptop. Here is the general procedure explained.

It is locked indeed. I'll look into that, thank you.

This might work. Other users have done similar things. If you can reduce the average temperature, you might be able to avoid the severe throttling. Try running the CPU with Disable Turbo checked on the main screen. This will slow the CPU down significantly. It will reduce power consumption and heat. Run another log file to see if this makes your laptop a little more useable.

I went ahead and disabled turbo from ThrottleStop. This resulted in a slightly lower than expected 2.6 GHz clock at a very stable 70°C. Everything looked great for a while and this allowed the game to run perfectly smoothly for 50 minutes. And then out of nowhere, the same throttling started to happen. I am attaching both the ThrottleStop logs and HWiNFO charts. It seems insane, the CPU temperature never seemed to go above 73 °C any second around when the throttling started (or before), so I honestly have no idea what triggered it or why. At this point it really seems completely broken to me unless I am missing some other metric that could cause it.

EDIT: And just in case there might be some other missing sensor value (I wouldn't be surprised as in previous generations of this device there used to be a VRM that received zero cooling and used to overheat, however no such sensor is shown in HWiNFO on this device unlike in the past) which triggers this throttling, what software should I use to make sure I am not missing any temperature sensor's readings? HWiNFO literally only shows CPU, GPU and drive temperatures and nothing else, which is a bit suspicious as the previous generation of the device used to have way more. Also even fan RPMs are not shown there anymore. Just something that has all the data, even if in an inconvenient format. I am thinking there is a chance it could be a similar issue as with the VRM in the past, some overlooked chip overheating and causing everything to throttle, which could maybe be fixed with some extra/replaced thermal pads etc. (especially since disabling turbo made the throttling kick in after 50 minutes rather than 10, which indicates that the trigger is not completely trivial)

 

[unclewebb]

which indicates that the trigger is not completely trivial

Only an XPS engineer deep inside Dell is going to know what sensor or sensors are used to trigger throttling. The problem you are having seems to be unique to Dell. Other manufacturers are guilty of throttling their laptops but not quite like this.

I have read about some designs that include a laser temperature sensor on the motherboard that points up towards the backside of the keyboard. This temperature data can be used to trigger throttling. Overheating VRMs are also a possibility. When you run the CPU at a reduced speed, it takes longer for the VRMs, etc. to heat up before the throttling begins.

Some test results while using your laptop in a frozen meat locker might be interesting. Your laptop might work perfectly fine in that environment.

I gave up trying to solve Dell's throttling laptop problems. Let me know if you ever find a solution.

 

[bxr]

Did you manage to solve it? I had the exact same issue with my XPS 9710 though with a different CPU (i7-11800H). Dell gave me a motherboard replacement and nothing changed.

For me, it seemed like its a problem with the integrated GPU. The frame drops happened only when using the internal display. I managed to "solve" it by using an external display and turning on "Direct Graphics Controller Direct Output Mode" in BIOS. In F1 2021 benchmark mode using the external display yielded 140fps vs 100fps with the internal screen, and the throttling is gone.

In some circumstances, my CPU suddenly starts throttling from 3.5-4.5 GHz all the way down to 798 MHz, for no clear reason. Sure, the CPU was running a bit hot before that within the range of 77-91°C, but why does that make it drop down to a fifth of its performance, resulting in a stable 70°C? At that point it seems to be running at around 15-17 W, while PL1 seems to be 45 W and PL2 109 W, both well above the power actually going to the CPU.

For context, this happened a few times while playing Age of Empires IV, as it went from nicely playable on 4K with High graphics settings to a frustratingly unplayable ~5 FPS which does not improve even if I go down to FHD resolution, as it is so bottlenecked by the CPU at that point. After that it also seemingly randomly shifts between these two states, having good fps one second and very low the next. To be honest even a stable 2 GHz would sound better than this.

The device is a Dell XPS 17 9710, with an Intel Core i9-11980HK CPU and NVIDIA GeForce RTX 3060 Laptop GPU.

I have had what seemed to be a very similar issue with an earlier generation of this device in the past. However, I am pretty sure back then the throttling happened when the temperatures were higher and then the PL1 did shift down I think to 15W, explaining why the throttling was happening. Back then, ThrottleStop worked perfectly to eliminate this power throttling entirely, which made things run a bit hotter and resulted in one or two thermal shutdowns over a year but at least it worked.

Sadly with this device and its 11th-gen CPU, despite having played around with all the settings ("Disable and Lock Turbo Power Limits" and all the others I have seen mentioned in relation to it), ThrottleStop never seems to have an effect on the power limit throttling and seems to be unable to prevent it. Does anyone know why, or how I could try to make it work?


View attachment 234449


View attachment 234451