Working on disabling Throttlestop undervolt when sleeping - Dell XPS 15

[boofbuscus]

I have a Dell XPS 15 and have successfully undervolted it which has drastically improved temperatures and performance which is great!
The downside, is that when the laptop is put to sleep it will most often crash due to the reduced voltage and fail to sleep requiring a hard reset.
This occurs during any small or large undervolt and is common across Dell XPS 15 owners. So my mission is to have the undervolt only applied when inside Windows and to turn off before the laptop sleeps.
Previously the fix was to go into the BIOS and set 'Force S3 Sleep', however Dell in their infinite wisdom have removed this option.

I already have task scheduler set to start Throttlestop when I login which works beautifully setting my -0.125mV undervolt (confirmed with HWmonitor) however now I need a solution to set the undervolt to 0mV offset before it sleeps.
I observed the event logs when the laptop sleeps, and it sends a Kernel Power Event with ID 42 just before going to sleep. I can use this trigger to try and run a script to set the Undervolt to 0mV right before it sleeps to save it from crashing. Then once I login it will resume the -0.125mV underclock.

I can think of two ways of doing this:
1) Run a script to change the .ini file to set an underclock of 0mV, and hopefully this change would be immediate and apply before Sleep
2) Run a script to turn off Throttlestop before the laptop sleeps

I am thinking Option 1 would be best, as for some reason no matter if I turn on or off Throttlestop (whether it is green or red in the task bar) the underclock is still applied (confirmed with HWmonitor). So simply turning off or killing the Throttlestop process would not disable the underclock. I would need to actually TELL the laptop to remove the voltage offset, does that sound right?

If all the above sounds good, can someone help me write a script that is simple and can run very quickly to remove the underclock before the laptop goes to sleep?

Thanks!

 

[unclewebb]

1) Running a script to change the INI configuration file is not going to work because ThrottleStop does not read the INI file before the computer goes to sleep.
2) ThrottleStop does not reset the voltages when it exits so this is not going to work either.

Can you post a screenshot of your ThrottleStop FIVR window so I can see your under volt settings? What CPU model do you have? Is your under volt 100% stable? What sort of stress testing have you done? Can you pass a 1 Thread 1024M TS Bench test with 0 errors as well as a full load TS Bench test with 0 errors? I just want to make sure that your sleep issues are not being caused by too big of an under volt.

For testing purposes, try setting up a second profile in ThrottleStop. In the FIVR window for your second profile, you will need to check the Unlock adjustable voltage option and set the offset voltage to zero. When you switch to this profile, it should zero your offset voltage. Now go into sleep mode and see if you have any problems resuming. Try this 5 or 10 times to see what your success rate is. When you resume, ThrottleStop will still be at zero offset voltage so you will have to manually switch back to your profile that has the under volt applied.

If this fixes the problem you are having with sleep mode then I will add a "zero voltage when going to sleep" option to ThrottleStop. This would be way easier than trying to cobble together some scripts. Send me a private message so we can do some testing.

Edit - Checking the new Voltage option should reset the offset voltages to zero just before Windows enters sleep mode.

 

[boofbuscus]

Edit - Checking the new Voltage option should reset the offset voltages to zero just before Windows enters sleep mode.

How do I find the new voltage option? I seem to have the latest version of Throttlestop 8.70.6 but cannot see this option

Thanks

 

[unclewebb]

How do I find the new voltage option?

I added a zero offset voltage option to ThrottleStop just for you but I have not publicly released this updated version yet. In the next day or two I will send you a download link so you can give it a try.

Initial testing looks good. It takes my laptop about 1.5 seconds to shutdown from the time that Windows gives the signal to ThrottleStop. In the log file, you can see the VID voltage increase so the zero offset voltage option is definitely working. When it resumes from sleep mode, the CPU immediately goes back to using offset voltage.

I brought this topic up in the NBR ThrottleStop forum and another user said that he had a similar problem on a couple of his low power laptops. Hopefully the updated ThrottleStop version will allow users to maintain their under volt without having any issues when going to and resuming from sleep mode.

   DATE       TIME    MULTI   C0%   CKMOD  CHIPM   BAT_mW  TEMP  GPU     VID   POWER
2019-10-29  14:33:44  35.91    1.2  100.0  100.0        0   45     0   1.0457    4.3
2019-10-29  14:33:44  35.83    1.3  100.0  100.0        0   44     0   1.0463    4.1
2019-10-29  14:33:44  35.90    1.1  100.0  100.0        0   44     0   1.0463    3.9
2019-10-29  14:33:44  35.57    2.3  100.0  100.0        0   45     0   1.0463    5.5
2019-10-29  14:33:44  35.89    1.9  100.0  100.0        0   45     0   1.0463    4.5
2019-10-29  14:33:44  35.87    1.7  100.0  100.0        0   45     0   1.0463    4.6
2019-10-29  14:33:44  35.89    1.6  100.0  100.0        0   44     0   1.0463    4.2
2019-10-29  14:33:44  35.89    1.6  100.0  100.0        0   45     0   1.0463    4.3
2019-10-29  14:33:45  35.91    1.1  100.0  100.0        0   44     0   1.0458    3.7
2019-10-29  14:33:45  35.89    1.5  100.0  100.0        0   44     0   1.0457    4.0
2019-10-29  14:33:45  35.43    3.2  100.0  100.0        0   44     0   1.0458    6.7
2019-10-29  14:33:45  35.77    1.9  100.0  100.0        0   44     0   1.0458    4.7
2019-10-29  14:33:45  35.67    1.9  100.0  100.0        0   44     0   1.0458    4.7
2019-10-29  14:33:45  34.72    8.2  100.0  100.0        0   45     0   1.0458    9.7
2019-10-29  14:33:46  35.87    4.2  100.0  100.0        0   46     0   1.0463    6.1
2019-10-29  14:33:46  33.98   55.7  100.0  100.0        0   53     0   1.0853   34.3 - Windows shutdown signal received
2019-10-29  14:33:46  34.00   86.0  100.0  100.0        0   57     0   1.0841   34.1 - overclock is decreased and VID voltage has increased
2019-10-29  14:33:46  34.00   94.6  100.0  100.0        0   56     0   1.0841   35.6
2019-10-29  14:33:46  34.00   55.3  100.0  100.0        0   56     0   1.0841   30.6
2019-10-29  14:33:46  34.00   33.6  100.0  100.0        0   53     0   1.0845   23.6
2019-10-29  14:33:46  34.00   27.8  100.0  100.0        0   56     0   1.0853   22.2
2019-10-29  14:33:47  34.00   33.1  100.0  100.0        0   58     0   1.0841   23.6
2019-10-29  14:33:47  34.00   26.9  100.0  100.0        0   58     0   1.0841   22.3
2019-10-29  14:33:47  34.00   67.3  100.0  100.0        0   59     0   1.0804   28.5
2019-10-29  14:33:47  34.00   97.0  100.0  100.0        0   60     0   1.0804   35.8
2019-10-29  14:33:47  34.00   80.0  100.0  100.0        0   60     0   1.0804   33.1
2019-10-29  14:33:47  34.00   80.8  100.0  100.0        0   60     0   1.0792   33.2
2019-10-29  14:33:47  34.00   49.9  100.0  100.0        0   60     0   1.0792   33.9 - computer heading into sleep mode with 0 offset voltage
2019-10-29  14:33:56   1.00  100.0  100.0  100.0        0   58     0   1.0414   33.9 - Approximately 9 seconds later, computer wakes up
2019-10-29  14:33:56  32.01   67.3  100.0  100.0        0   58     0   1.0414   68.1 - CPU is coming back up to speed
2019-10-29  14:33:56  32.00   51.7  100.0  100.0        0   57     0   1.0408   32.4 - VID offset (-40 mV) has been re-applied
2019-10-29  14:33:56  32.13   37.4  100.0  100.0        0   58     0   1.0414   32.2

 

[unclewebb]

ThrottleStop 8.71

ThrottleStop 871.zip

drive.google.com

New Features
- new option to set FIVR offset voltages to zero when entering Windows Sleep mode.
- reporting of maximum multiplier in FIVR Turbo Ratio Limits.
- updated link to The ThrottleStop Guide (2019) at Ultrabook Review.

Only a couple of minor updates. This new feature will zero the offset voltage just before entering Sleep mode which might help the CPU resume from Sleep mode more reliably. Some laptops have problems resuming when using under volts that are on the edge of stability.

The other feature should show the default maximum turbo multiplier.

This version is not signed. If you do not need these new features, I would continue using TS 8.70.6.

@boofbuscus - Thanks for your patience. Let me know if this new feature solves the problem you are having with your Dell XPS.

 

[boofbuscus]

Great thankyou so much! I will test it out over the next few days and report back

 

[boofbuscus]

This has fixed the problem, the laptop now sleeps perfectly thankyou so much!

 

[unclewebb]

@boofbuscus - Your suggestion and feedback are appreciated. There are a lot of different laptops that are going to benefit from this new feature.

 

[Notbillgates]

@unclewebb I was having similar issues. My Lenovo Y540 9750h was 100% stable while up and running, but it would sometimes randomly crash and reboot during sleep. Because of that issue, I had been using a smaller undervolt.

I began using the test version you posted here on Saturday. I wanted to test it well before reporting back my results. I again lowered my undervolt and have not yet had one crash or reboot while in sleep or wake. This appears to have fixed my issue!

Thank you for your work, time, and amazing program. You are a credit to the tech community. Please, let me know if I can do anything for you, testing or whatever else.

 

[unclewebb]

@Notbillgates - Perhaps Intel includes a little bit of extra voltage so CPUs will resume reliably. Resetting the offset voltage to default values seemed like a good idea. There are probably quite a few laptops that are going to benefit from this new feature. Thanks for the feedback.

 

[rrock_15]

@Notbillgates - Perhaps Intel includes a little bit of extra voltage so CPUs will resume reliably. Resetting the offset voltage to default values seemed like a good idea. There are probably quite a few laptops that are going to benefit from this new feature. Thanks for the feedback.

Hi, just stopping by to say thanks for the update! I switched to throttlestop on my new XPS 15 7590 with i7 7590U after XTU gave me grief with sleeping while undervolted, only to find that throttlestop has been doing the same thing when I opened up my laptop today. Went to google looking for a solution only to find that a fix was implemented just last week! Can't believe my timing here haha, thanks again for the work you're doing

 

[unclewebb]

I switched to throttlestop on my new XPS 15 7590 with i7 7590U

Is your CPU the Core i7-9750H? I do not think Dell is installing any 7590U CPUs in any recent laptop.

It is good to hear that this new feature has helped solve your problem. Your under volt is probably right on the edge of stability. Try running the TS Bench test and set it to 1 Thread. This will get the CPU using the highest multiplier. You should not see any errors during any test.

Post an update in a week or two if your CPU is still resuming from sleep without any issues.

 

[rrock_15]

Is your CPU the Core i7-9750H? I do not think Dell is installing any 7590U CPUs in any recent laptop.

It is good to hear that this new feature has helped solve your problem. Your under volt is probably right on the edge of stability. Try running the TS Bench test and set it to 1 Thread. This will get the CPU using the highest multiplier. You should not see any errors during any test.

Post an update in a week or two if your CPU is still resuming from sleep without any issues.

Yes indeed, its a 9750H "Right on the edge" is a good way to put it, I wanted to maximize performance to power consumption ratios, while improving battery/normal AC usage thermals. Currently running a -110mv cache and -220mv core undervolt that is now quite stable with the new zero offset whilesleep feature. Only consuming 45w under heavy load at 2.6ghz with this undervolt as opposed to 56w, and have 3.7ghz stable @56w. EST set to 255 on battery profile, 0 on AC and performance beause dynamic clock on wall power is pointless to me. Really good stuff here!

 

[mkdr]

Tested this on my Dell XPS 15 9570, doesnt work. @unclewebb Laptop still crashes during standby with undervolt.

 

[unclewebb]

@mkdr - The new Sleep Defaults feature has two options. Did you try both of them together?

The Cache Ratio option drops the cache down to 800 MHz. That helps with my Lenovo laptop. Maybe that or a new feature that drops the CPU MHz might help. Does your 9570 always resume from sleep without any issues when not under volting?

 

[mkdr]

@unclewebb No I have just tried the Voltage selection, not both. Why would that matter if the voltage is reset to default?

Yes I have no other issue with it, no crashes while non sleeping. I had a very long time working sleeping too, over thelast 6 months, but over the last week, I had two crashes again during sleep, after I changed a little bit of the undervolt values (or the temperature maybe got a bit colder in my room now).

I had very long these values

which led to a stable standby for my laptop. I kinda added a positive pffset for the the iGPU values some months ago, and it led to a stable behavior. Then I removed those again the other week and also rose the core and cache to -0.110 , and then I had 2 crashes again during sleep.

The crash manifests that I cant turn on the laptop anymore, the keyboards would light up but thats it. It also just happens after a longer sleep, about 8-14h (on ac).

Could you maybe please add debug outputs to the log file, like those marks you made in the upper screenshot, so I can see in the log file myself, if TS works correctly (like triggering the standby on/off resets).

 

[mkdr]

Since I have added 0.010mV offset to iGPU, the laptop hasent crashed anymore in standby, really weird. It has run 100% stable now again with this:

 

[Uncle Zingolo]

Hi @unclewebb

First I'd like to thank you for all your hard work.

I'm new to this and I feel quite overwhelmed to be honest. Just got my first laptop in over a decade and I'm looking to give this thing a go on my new XPS 7590 (i7, 16gb ram).

The version you posted with the fix for the sleep issue was from back in November. Do you recommend to use this 8.71 version or is there a more up to date one?

I've found a very helpful video on how to undervolt using your program and I plan to follow it step by step. I'd appreciate it if you can have a look and let me know if the settings the uploader used is optimal for the laptop I have.

One last thing. Every time I try to install the latest stable version or the beta version, I get the following message after copying and extracting the files.

"The code execution cannot proceed because mfc120u.dll was not found. Reinstalling the program may fix this problem"

Not sure how to work around this system error.

Thanks again

 

[unclewebb]

The latest stable and beta versions of ThrottleStop can always be found on TechPowerUp.
https://www.techpowerup.com/download/techpowerup-throttlestop/

Update for Visual C++ 2013 Redistributable Package
https://support.microsoft.com/en-us/help/4032938/update-for-visual-c-2013-redistributable-package

If you have a 64 bit operating system, you will need to install both the x86 and x64 versions of that library.

Dell started releasing updates last month through Windows Update that disable under volting. To find out if under volting is working on your computer, start by using a very small under volt. Press the Apply button and see if the under volt appears in the monitoring table in the top right corner of the ThrottleStop FIVR window. If you try to under volt and the Offset voltage column in the monitoring table continues to show +0.0000 then under volting has been disabled.

The only way to find optimal settings is by you testing your laptop. The settings that you learned about while watching a YouTube video might be wonderful for someone's laptop but might not be so wonderful for your laptop.

Watch a variety of videos and try using ThrottleStop for a day or a week to get an understanding of it.

As for the video above, the first thing he does is install Core Temp to monitor the CPU's temperature. ThrottleStop already has this feature so installing a second program to do the same thing that ThrottleStop is doing makes no sense. His ThrottleStop settings will also significantly reduce the performance of your laptop.

 

[Uncle Zingolo]

Hi Kevin,

As soon as I received the laptop, the first thing I did was disable BIOS auto updates through BIOS setup -> security -> UEFI capsule firmware updates -> disable.

The latest 1.6 firmware is the one causing the undervolt restrictions. Based on some brief research, going into the bios menu and loading the setup default values makes it possible to undervolt again.

I managed to install all the missing dll's and the ThrottleStop program worked instantly. That was last night so I wasn't aware that the settings on the video posted below end up greatly reducing the performance of my laptop.

The temperatures on my laptop have been reduced significantly, but I would love to revert back to the original settings from before making the modifications that were recommended on the shared YouTube video above. I can't remember the original values and I'm not sure if it's best to uninstall Throttle Stop and reinstall again? or is there a factory reset option on the program?

When I first performed the HeavyLoad test, the CPU temperatures went as high as 99/100 Celsius to as low as 60/65 Celsius. This sounds great of course, but I wouldn't want to compromise a lot of performance in the process.

Please see the screenshots below.

On the right box, the results on the table (6, 24, 109) were from a TS Bench test that was preformed prior to undergoing the undervolt process. Once I followed the steps given on the YouTube video, I performed the same Bench Test again and noticed that the results were increased by a few seconds on the 64M, 256M and 1024M. It went from (6, 24, 109) to (8, 31, 125).

This increase only tells me that the system is now slower than it previously was before changing any of the set values.

I will continue to learn more about the program over the coming days and weeks but if anyone out there has the same XPS 15 (7590) model as myself with i7-9750H 16GB Ram 512GB SSD, please share some of your experiences once you get the chance.

Many thanks for reading.

Zingolo

 

[unclewebb]

Before you make any adjustments in ThrottleStop, I would go back to the default values. You can do this by exiting ThrottleStop, remove the ThrottleStop.INI configuration file and then shut down your computer to reset all of the registers within your CPU. When you boot up and run ThrottleStop, it will read the default values from your CPU.

I know the default turbo ratio limits for the 9750H are 45, 44, 43, 42, 41, 40 but I am not sure what the default turbo power limits are. Different manufacturers use different default settings depending on the laptop model. The default Intel values are usually 45 Watts for the long turbo power limit and 54 to 60 Watts for the short turbo power limit. Some laptops built on the Clevo chassis have much better cooling. Their default values are something like 60 Watts for the long limit and 80 W or 90 W for the short limit.

Your screenshots show you have these power limits both set to 24 Watts which was recommended in the video. This will definitely allow your CPU to run cooler but it is going to sacrifice maximum performance. Intel mobile CPUs run hot. Intel is well aware of this. They also run very reliably even when they are running at over 90°C. Intel considers any peak core temperature under 100°C to be a "safe operating temperature". Those are their words, not just the words of some guy in a forum. Unless the cooling in your laptop is really horrible, I would not set either of the turbo power limits to anything less than 45 W. Definitely not 24 W.

Set the Speed Shift Max value back to 45. Are you using the Windows High Performance power profile? The Speed Shift EPP value on the main screen of ThrottleStop is what you are requesting from the CPU. The Speed Shift EPP value in the FIVR monitoring table is the Speed Shift value that the CPU is actually using. In Balanced mode, Windows usually controls this value. If Windows is setting this to 127 (as per your screenshots) then there is no reason to check the Speed Shift EPP value in ThrottleStop. It is being ignored. If you want ThrottleStop to control EPP then try try using the Windows High performance power profile.

Your undervolt looks fine. It is working according to the FIVR monitoring table so that is good. Go do some testing. See how much heat is OK. Some people use ThrottleStop to reduce their CPU speed simply because they do not like a lot of fan noise. You can do whatever you want with the program. Just make sure that it is your decision and you are not just blindly following a video (or blindly following my advice).

Edit - Setting the core and cache offset voltages to -125 mV is a great place to start testing.

 

[Shriukan]

Hello @unclewebb
Thanks for such a wonderful program. I am loving the improved performance on my devices. I have a question and possible request.

My MSI laptop handles sleep mode and resuming without issues with the Sleep Default - Voltage option ticked, however that is not the case for my Surface Pro 3 tablet.
I did some digging and saw that those devices don't actually go into Sleep Mode but rather into Connected Standby which trigger different IDs. I have 2 Tasks set to help by having one launch ThrottleStop on login and resume (Kernel-Power event ID 507) and one set to taskkill it on event 506. The issue is sometimes the program doesn't get killed in time causing the tablet to still crash.

The question comes down to, is ThrottleStop's Zero Voltage option configured to take into account CS and if not, would you mind adding it to see if it fixes the crashing for all of us with CS-enabled devices? (That way we don't have to disable CS)

Cheers,
Shriukan

 

[unclewebb]

@Shriukan - I do not think that ThrottleStop's Zero Voltage Option takes into account Connected Standby. I do not own a device that uses connected standby so I have no way to test and develop this feature to make sure that it properly supports CS.

What are your undervolt settings? Some people having trouble are using offset voltages that are not 100% stable. Can you pass a 1 or 2 thread TS Bench test with 0 errors? Are you undervolting both the CPU and the Intel GPU? If you are undervolting both of these, try undervolting only one of these at a time to see if it makes a difference. Try to isolate what is causing the problem. Try backing your undervolt off -10 mV or -20 mV. If you are undervolting by -100 mV, use -90 mV or -80 mV and see if that makes a difference.

Undervolting is a compromise. Sometimes using just a little more voltage than the bare minimum helps with stability problems when resuming.

 

[Shriukan]

Thanks for the quick reply. I am running -70mV on the CPU Core and iGPU and -125mV on the CPU Cache. I have already fine tuned to the point of BSODs and pulled down each by 15mV to stay on the safe side. I did pass the stress tests on 1,2 and 4 threads with nothing else and while running a game to purposefully crash and it survived without errors in all cases and in each and every condition, after passing the test, the tablet will crash when going into standby.

From what I understand, the issue with the undervolt is specific to CS. Apparently the state already undervolts the machine close to its minimum operational values and the offset just pulls it below operating voltage, hence the crash.
I will however try and decrease the undervolt to see where it might survive and compare its performance. If it is negligible i might stick with it. If it isn't I prefer living with a few occasional crashes due to task scheduler's bad timing.