The ThrottleStop Guide

Spoiler

 

Maybe TGL-H doesn't use FIVR, but U and L absolutely does. Core voltage provided is 1.7V at all times.

 

Thanks @Prema for the pic. On my Asus desktop board, I have to enable SVID support in the BIOS. After that, CPU voltage control works in ThrottleStop. If you are using a fixed voltage set in the BIOS, ThrottleStop voltage control will probably not do anything.

Does Limit Reasons appear to work? I think I remembered to enable it this time. Try setting a low power limit and see if something lights up red.

Everything else looks good, especially considering I do not own an 11th Gen CPU.

If you run a TS Bench - 1 Thread test, is Windows smart enough to schedule it on one of your preferred cores? This does not work correctly on my 10850K. I assume it is a Windows bug.

I hope you typed that wrong. 1.7V is a lot of voltage for a CPU.

I have not seen any screenshots of an 11th Gen U or G7 CPU that shows anything for voltage readings or voltage adjustments in the ThrottleStop FIVR window. I know the 10th Gen U support voltage control. What CPU Generation are you talking about?

 

This is 11th gen. It must have an integrated voltage regulator to step down the 1.7V input.

Throttlestop doesn't have any ability to control the voltage. It is able to force x41 on all cores for longer than default on a 1165G7 if I run an older bios version.

 

An input voltage of 1.7V makes sense. The CPU cores get a lot less than that.

That is the problem. The CPU might still use an integrated voltage regulator but Intel has blocked any access to it. This means that software no longer has any control of the CPU voltages. ThrottleStop will never work on 11th Gen U or G7 and Intel XTU does not support the low power CPUs so it will never work either.

Intel CPUs are not much fun when CPU voltage control is locked out.

 

Hello I undervolted my brother 6200U and was doing that for the first time, so sorry for perhaps stupid question.
It seems that UV - 100 mV on core and cache went well, it decreased some Watts and generally the notebook is running at full speed. Than I searched for msr 0x150 and the others for fun and 0x150 is all zeros.
Is it OK? Was UV done just on MMIO base?
Also I have seen some pics of 8 Gen U cpu's running beyond 15W TDP but 6200U is not that case, it could draw 11W at 27 multi under load all day long. So my question is if I could force this cpu beyond the TDP somehow (simmilar to 8 gen) would the performance benefit from it, or is the performance limited just by max multiplier?

 

I understand them disabling it but disabling the bios option for voltage control or voltage control support completely is terrible.

Also I have terrible UV with 10875H even -50mv worked for some time but gave a bsod -40 gave me a bsod too it is either I have a terrible silicon or cpu needs lots of volts to boost as high and not really designed well...

 

The 27 multiplier is the maximum multiplier when a 6200U has 2 cores active.

https://www.cpu-world.com/CPUs/Core_i5/Intel-Core i5-6200U Mobile processor.html

If the CPU is already running at full speed then that is all it can do. It does not make any sense to make the CPU consume more power if it is not going to run any faster.

This msr works differently compared to most other msrs. This one msr allows software to communicate with the voltage regulator. You can send hundreds of commands to the CPU and get lots of information from the CPU through this one register. There is a whole world of stuff hiding behind this one door. If you only do a read msr on this register, that only shows you the door. It does not show you everything that is hiding behind the door.

If Intel set the default VID voltage close to what the CPU actually needs then your CPU does not need to undervolt very much. My 4th Gen 4700MQ is the same. It is only stable at -40 mV or -50 mV.

On my desktop computer, the load line calibration value makes a big difference to how much I can undervolt and still be stable. When the IA loadline is set to 0.25 mOhm, I can hardly undervolt at all. I can only undervolt -10 mV when running an OCCT SSE stress test.

 

This could be that need to UV is eliminated.

Intel arranges the volts correctly and everyone gets chips that do not need undervolt but at the same time some people wrote 10875h can undervolt much over -100mv but some people like me can not get any undervolt.

Cpu uses a lot of power without undervolt which I wouldnt care but it is getting throttled.

My 10875h uses 100W on load and it is a 45W tdp that is ec limited when gpu is active so at 45W I can only get 3Ghz all core then why do I have a 4.3Ghz all core boost this is very scammy...

 

Thank you!

 

And some will do the opposite. All your added features in ThrottleStop is used make his $3 grand laptop performing as laptops from yesterday. I don't think this was the reason you started up with your TS project for over a decade ago

Finally found a low temp undervolt for my new M15 R3, the trick was the turbo ratio limit.

 

Not sure why anyone would pay top dollar for a laptop and then be forced to undervolt it and underclock it to try and prevent it from over heating.

Maximum performance has always been my only concern. Throttle Stop. I wanted to stop the nonsense. All these modern laptops that put out more heat than an Easy Bake oven are not for me.

Just helping another Alienware owner on another forum at the moment. His 8th Gen laptop seems to be power limit throttling below the 45W TDP rating. Another Dell feature. If I paid $3,000 U.S. for something like this, it would be going back in the box before the sun went down.

Edit - It looks like his overheating Alienware is constantly power limit throttling. Dell is limiting his 45W CPU to 38W. If it was allowed to run at its full rated power, it would overheat and thermal throttle instead. CPU voltage control is Locked so he is out of luck.

Code:

   DATE       TIME    MULTI   C0%   CKMOD  BAT_mW  TEMP   NVIDIA GPU     VID   POWER
2021-04-05  20:54:00  30.02  100.0  100.0       0   94    1443    61   0.9341   38.0   PL2
2021-04-05  20:54:01  30.30  100.0  100.0       0   93    1443    61   1.0089   37.9   PL2
2021-04-05  20:54:02  30.29  100.0  100.0       0   92    1443    61   1.1002   38.0   PL2
2021-04-05  20:54:03  30.85  100.0  100.0       0   96    1443    60   1.0485   35.9   TEMP
2021-04-05  20:54:04  33.24  100.0  100.0       0   99    1443    60   1.1759   34.8   PL2
2021-04-05  20:54:05  29.78  100.0  100.0       0   94    1443    61   0.9816   36.1   PL2
2021-04-05  20:54:06  31.15  100.0  100.0       0   91    1443    61   0.9525   36.9   PL2
2021-04-05  20:54:07  27.88  100.0  100.0       0   99    1443    60   1.1997   37.9   PL2
2021-04-05  20:54:08  30.90  100.0  100.0       0   93    1443    60   1.1880   37.7   PL2
2021-04-05  20:54:09  29.18  100.0  100.0       0   93    1443    60   0.9879   37.6   PL2

Thanks @Papusan for continuing to enlighten the blind.

 

If you think the above Dell throttling problem is bad, it is nothing compared to this throttling Dell Latitude. It features continuous power limit throttling. The PL1 power limit is being set so low internally by the EC that the CPU is forced to run at its lowest speed, 800 MHz. CPU and GPU temperatures and power consumption are all OK. There is no reason for this insanity. It continues on for two minutes, there is a brief recovery and then the user gets slammed again with another round of severe throttling. You cannot use ThrottleStop to fix Dell's latest throttling blunders.

Code:

   DATE       TIME    MULTI   C0%   CKMOD  BAT_mW  TEMP   NVIDIA GPU     VID   POWER
2021-04-06  00:54:00   8.00   45.8  100.0       0   71     696    69   0.5334   10.2   PL1
2021-04-06  00:54:01   8.00   81.8  100.0       0   72     696    69   0.5376    8.5   PL1
2021-04-06  00:54:02   8.00   67.1  100.0       0   72     139    69   0.5409    8.5   PL1
2021-04-06  00:54:03   8.00   56.2  100.0       0   71     139    69   0.5326   10.8   PL1
2021-04-06  00:54:04   8.00   52.1  100.0       0   71     139    69   0.5366    9.4   PL1
2021-04-06  00:54:05   8.00   51.6  100.0       0   72     139    69   0.5349    9.0   PL1
2021-04-06  00:54:06   8.00   52.1  100.0       0   71     696    69   0.5359   10.2   PL1
2021-04-06  00:54:07   8.00   49.4  100.0       0   71     696    69   0.5408    8.7   PL1
2021-04-06  00:54:08   8.00   48.5  100.0       0   70     139    68   0.5345   10.4   PL1
2021-04-06  00:54:09   8.00   62.6  100.0       0   71     139    68   0.5439    8.4   PL1
2021-04-06  00:54:10   8.00   56.8  100.0       0   72     747    69   0.5392    9.3   PL1
.....
2021-04-06  00:56:05   8.00   44.6  100.0       0   65     696    63   0.5442    9.3   PL1
2021-04-06  00:56:06   8.00   45.7  100.0       0   65     696    63   0.5468    9.3   PL1
2021-04-06  00:56:07   8.00   50.9  100.0       0   65     139    63   0.5459    9.8   PL1
2021-04-06  00:56:08   8.00   46.9  100.0       0   65     139    63   0.5425    9.7   PL1
2021-04-06  00:56:09  20.63   39.3  100.0       0   67     747    63   0.7473   10.7
2021-04-06  00:56:10  26.00   43.8  100.0       0   70     747    63   0.7288   11.8
2021-04-06  00:56:11  26.00   50.9  100.0       0   67     696    63   0.7402   13.9
2021-04-06  00:56:12  26.00   53.0  100.0       0   69     696    63   0.7419   14.1

 

Hi all!

The fans of my laptop Lenovo ThinkPad X1E (gen 1) was starting and making a lot of noise as soon as the CPU was staring working. It was annoying.

Following the advices in this thread ( @unclewebb) I tweaked and undervolted it, using ThrottleStop. Also disabled Lenovo Thermal Solution.

The laptop is working much better now. However, sometimes (sparsely) I see the yellow EDP OTHER in RING appear. Is it okay to ignore it?

Here are my configurations:





Any insight would be much appreciated. Cheers,

 

EDP OTHER lighting up under ring is common and nothing to worry about. On some CPUs, it seems to be hyper sensitive. The CORE column is the important one. Boxes lighting up red under the CORE column indicate that throttling is in progress. As long as that is not happening, everything should be OK.

Setting the cache to -0.1377 is likely right on the edge of stability. Most users set that one to -0.125. If you ever have a BSOD, reduce your cache offset first.

The core offset to cache offset ratio does not need to be set to a 2:1 ratio. Most users find that there is little to no advantage going beyond -100 mV more than the cache. If you set the cache to -0.1250 then there is probably going to be no advantage to setting the core to more than -0.2250. Setting the core too high will not hurt anything. Any excessive core request will simply be ignored by the CPU.

You get a thumbs up for doing that.
Some manufacturer's control software is garbage. If you do not need it, get rid of it.

That is the important thing. Glad to see another happy ThrottleStop user.

 

Quick sanity check for a noob: are there motherboards that don't support Package C6 state or higher? My cores all enter C7 but the package never does, and I think I've tried just about every driver/bios combination.
(i7-10875h with unlocked bios. Aida64 screenshot included)

https://imgur.com/a/tqSdZv8

 

I do not have any documentation about what can block the deeper package C states or how to get them working correctly. My desktop Z490 board is also limited to a maximum of package C3.

Saving a watt when plugged in is not too important. For laptops running on battery power, it seems like a good idea to have all of the available low power package C states working correctly. Laptop manufacturers do not seem to put a lot of effort into getting them working. Either that or they deliberately disable them so their laptops have better SSD benchmark numbers.

If you ever find a way to solve this mystery, post an update so I can learn something new.

Did you try all of the tricks listed in this guide?

http://forum.notebookreview.com/thr...e-on-windows-enabling-deeper-c-states.815602/

Getting rid of the default Microsoft driver helped me find Package C3.

 

Thanks a lot @unclewebb . I increased the cache to -0.125 just in case .

 

I did a fresh install of W10 and it seems like the laptop just cannot use Package C states above C3. Makes sense because the bios had package demotion and undemotion disabled by default.
I guess this stupid laptop wasn't meant to be used like a laptop

 

By the way, increasing cache seems to fix EDP OTHER yellow label under ring. Since then, no more EDP OTHER.

 

Hello @Papusan
I found this interesting article about how some Uefi (Dell) drivers are tied up with Windows and installing some of ACPI specific persistent table.
https://standa-note.blogspot.com/2021/04/reverse-engineering-absolute-uefi.html?m=1
(I read it at speed anyway, I've never heard of Absolute security software but it could have some performance consequencies also)

 

Treksdot on the MSI forum discovered that his 11700K supports Thermal Velocity Boost.

https://forum-en.msi.com/index.php?...-missing-tvb-bios-options.362093/post-2055129

Using ThrottleStop to disable TVB throttling allowed him to run his CPU at full speed. Intel says that only the 11900K supports TVB but that is obviously not true. At full load he went from 4.6 GHz all core to 5.0 GHz all core by disabling a feature that Intel says does not exist in the 11700K. Interesting. An updated BIOS made this feature disappear.

 

Did anyone had an issue where setting PROCHOT offset to 0 actually makes it more restrictive? Manufacturer of my laptop set the prochost to 95 C but with 0 offset in thorttlestop it actually activates at 90 C.

 

When I set the PROCHOT offset to 0 my laptop just shuts down even if we do not see it there are other things that throttle can you set it to 1 or 2 and check

My laptops work on that if you have good cooling it should work better but if it does not have good contact it could get higher than 100 and in my case shut down or in your case aggressively throttle to keep it under

 

It is not supposed to do that. Can you post a screenshot of Limit Reasons when this throttling is in progress? A Log File that shows this throttling would also be useful. It might be some other type of throttling that is kicking in.

 

This might not find a good answer from this thread - still going to ask:

Anyone have any ideas that is Dell still continuing the trend of blocking undervolting of XPS machines with their new Bios versions?

I did update mine last summer and managed to get around their UV block that time. Now there's ofc newer Bios versions but I haven't found any valid information from the web if the same problem still persists.

I would hate to find out that my dear Throttlestop wouldn't work anymore after and update...

 

There's no reason to assume that they have stopped this behavior.
You might be able to work around the limitation via the UEFI variables, but it is not guaranteed to work and is quite experimental.

 

On my i5 6200u laptop with Windows 10 20H2, when I enable Speed Shift in TS, Windows power plan max frequency and max processor state settings are completely ignored. Any way to fix it?

The problem was caused by Windows 10 power slider.

 

New heatsink, trying to solve a core to core temp issue. Highest R20 score I've managed so far, fresh repaste using MX-4. Here are the temps idle and peak:



R20 score:


And my TS settings:


Pointers are welcome.

 

Based on the difference in temps from core to core, either your heatsink or CPU is not flat or you did a bad job with the thermal paste or it is pumping out. Some popular pastes can pump out within a week.

When loading the cores run something consistent like the TS Bench test.

Intel rates their core temperature sensors to be accurate to +/- 5°C. Everyone assumes that these sensors are perfectly calibrated and are 100% accurate from idle to full load. Definitely not. The third core from the top in your screenshot might always read too high even when the thermal paste and contact is perfect.

 

Skylake or KabyLake models with locked BGA i.e H/HQ and not HK have been locked after latest BIOS patches. If we go by VMware cpu patcher and apply the same uCode Dell uses in their XPS/Inspiron/Alienwares the undervolting feature is working fine. I noticed it after BIOS 1.11 shot my temps to 85C during startup/OS booting. I just update MEI FW and CPU ucode separately w/o installing Dell patches.

 

Repaste was done yesterday and thank you for the info about the sensors. The laptop runs fine, no stuttering or anything indicating a throttling condition. I've literally tried everything under the moon with trying to target that 3rd core; tried LM, tried ICD, tried different thermal pads, tried replacing the pads with a silicone putty (so damn messy), and have now replaced the heatsink. That core to core difference is actually the best I've seen so far. My TS Bench performance is within 5% of what it was before I replaced the heatsink. I'm starting to think this is the limit for my laptop.

 

When running cinebench what is the throttling reason on TS Limits? It should be only temp if it is power you can tune it more

You should be able to get more performance out of it since my 8750H could get same results with less all core boost.

You should put the windows to maximum performance power plan

on TS main page you can enable Speedshift and set it to 0

disable BD prochot

there are other things to do on the FIVR and TPL if you want the max performance too.

 

I'm getting thermal and EDP OTHER under ring.

I've set it to max performance, I've tried Speedshift set to 0 or leaving it off, doesn't seem to make much difference. Win10 does appear to dynamically adjust Speedshift dependent on load. I've also tried BD Prochot off or on. As far as pushing the tune further, I start getting BSODs if I go much further than -140 mV on the cache and anything beyond -135 doesn't seem to make a difference. I'll try messing with the core more...

 

If you are using ThrottleStop to manage Speed Shift EPP, use ThrottleStop to switch to the Windows High Performance power plan. It seems that Windows interferes less with the ThrottleStop Speed Shift EPP setting when using the High Performance power plan.

When THERMAL is triggered under CORE in Limit Reasons, you will usually see EDP OTHER under the RING column at the same time. The high temp is the reason for throttling. You need better cooling to solve this problem. That one core that either reads hotter or is hotter than the other ones will trigger this.

 

So, I cracked open my laptop again and repasted with Coollaboratory Liquid Ultra, using kapton tape to keep everything protected. Here are my thermals now:

This is after TS Bench 12 thread, 960M size run. These are thermals I can live with and are much better than the last time I used liquid metal. My R20 score hasn't changed much, but it's doing it without any throttling at all:


That one core is still toasty but not like it was before. I should add that the temps above were after multiple R20 and R23 runs. I'm finally happy.

 

Your temps look awesome but liquid metal on copper could dry out monitor your temps in couple months you may need to apply again

-unlike aluminium which forms a reaction and worst idea-

gallium and copper just do some electron exchange it doesnt affect it thermally but you will get staining of copper and might need to apply again

Also you could have some heavy background apps since your cpu could do better R20 score when testing is your cpu always boosting to 4.1 ish GHZ right ? if it is then your background apps are hogging it can you test it again with the cinebench affinity set at realtime at task manager advanced tab

 

Not too worried about LM. I've used it before and have had good results, most of the time. There's another user here that just did a repaste after having LM on his laptop for six years and from what I understand, the chemical changes that happen between copper and gallium do eventually stabilize with an alloy forming on the heatsink. It isn't very pretty, but doesn't affect performance.

I set R20 to realtime and the screen stopped updating. When it unfroze, this is what I saw:

With these temps:

I think I got this sorted now. Thanks to the team here and @unclewebb, I doubt Intel realizes what you've done for them.

 

Has anyone had any luck with undervolting a DELL latitude with a i5-10310U ? ...i really don't like getting instantly to 100°C when loading all cores.

It just doesn't feel right.

I can get to limit the actual TDP via throttlestop to reduce peak temperature but I'd like to give undervolting a try. Heck, I can't even disable hyperthreading to lower temps on scientific workloads. The current bios is so useless.

 

If CPU voltage control is locked, you need to edit a couple of EFI variables to unlock the BIOS.

https://brendangreenley.com/undervo...hermals-battery-life-and-speed/#cpu-undervolt

Scroll down until you get to this section:

Edit EFI Variables to Enable CPU Undervolting

 

Works great if you're under warranty. Too risky for common folks.
There is forced downgrade for BIOS using Usb I don't know if Dell has blocked it on newer PCs.

 

I thought it was nearly risk-free given a CMOS battery cycle will reset NVRAM and next boot will load back stock default values from the unmodified BIOS if any problems arise after editing the NVRAM?
ie. worst case, be sure you know how to access, disconnect, and reconnect your CMOS battery

 

I tried searching and found nothing up to now. Are there any ways to open up options in ThrottleStop for the 7600U in my Dell 7480 Latitude? The FIVR panel is mostly grayed-out.

 

Does the top middle of the FIVR window show that CPU voltage control is Locked?

You might be able to find a previous BIOS version which is not locked. There is nothing ThrottleStop can do if your BIOS locked your CPU.

 

It's all grayed out, yes.

 

I found an interesting review of an 11400F. It is rare to see a picture of ThrottleStop in a desktop review let alone the reviewers actually using ThrottleStop to get the most out of a desktop CPU. It is about time.

A better than 30% increase in their Cinebench R23 score had them smiling. Not only was there a significant increase in performance but they were also able to drop power consumption almost in half. They were running bench stable with a -300 mV undervolt!!

I think the review is Polish. Google translate showed me the important stuff. Their review includes lots of benchmarks, with and without the undervolt.

https://www.komputerswiat.pl/recenz...alnosci-tani-i-z-ogromnym-potencjalem/vn3knpb

 

Try resetting the BIOS to defaults.

 

Wow, thank you. Will get into it and hope I don't brick it since our IT would probably kill me.

By the way, is there any humanly valid reason why the would lock a BIOS so hard?

Like, if you get into the bios it's probably because you know what you are doing and it's not you can't destroy the laptop by having the CPU at 100°C 24/7.

EDIT: Unlocking EFI variables worked out smoothly. The CPU-cooler combo is horrible though. Even with -0.1V it still skyrockets to 100°C in a couple of seconds with TS bench.

 

Removing CMOS needs entire disassembly of my PC!

 

Good to hear. I hesitate to recommend stuff like this but with modern locked down laptops, you gotta do what you gotta do. Editing EFI variables seems to work OK for many Dell laptop models.

Pathetic heatsinks are a real problem. Everyone wants thin and light laptops, not thin, light and inadequate heatsinks. It would take very little effort and additional cost to make 100% better laptops. I think there would be a market for something like that but manufacturers disagree.