• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

ThrottleStop and undervolting - thermal throttling questions

Joined
Jul 10, 2020
Messages
25 (0.02/day)
Hi! I've been using a Clevo PB71EF-G (re-pasted with Kryonaut Thermal Grizzly, and also new thermal pads) for a while now and I've had it undervolted pretty much since the day I got it. For the games I've played it's been doing OK, doing what it's supposed to do, at least in how the games feel so even if I am thermal throttling then its surely not affecting me in real world usage. Lately though I have been kinda curious about the performance purely from a knowledge perspective, and I saw that other people have posted here and gotten some really in-depth analysis and tips on how to improve. I also see a vast difference between various guides out there and I just know from experience that can't really be good.

I've posted whatever information I've seen that other people have posted so please forgive me if I've missed something :)
 

Attachments

  • ThrottleStop 9.0.png
    ThrottleStop 9.0.png
    42.3 KB · Views: 1,533
  • FIVR.png
    FIVR.png
    70.9 KB · Views: 1,539
  • TPL.png
    TPL.png
    25.5 KB · Views: 1,538
  • Cinebench.png
    Cinebench.png
    42.9 KB · Views: 1,234
  • TS Stress Test.png
    TS Stress Test.png
    74.8 KB · Views: 1,881
  • 2020-07-10 Logs - Running Cinebench R20.txt
    22.1 KB · Views: 185
  • 2020-07-10 Logs - Running TS Stress Test.txt
    16.5 KB · Views: 149
Joined
May 7, 2020
Messages
244 (0.17/day)
You could try lowering the Core offset even more. 2:1 ratio Core:Cache should work fine and give you better results. The PL1 limit is triggered probably because you're using Balanced Windows Power Mode. Try to set it to Performance.
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
You could try lowering the Core offset even more. 2:1 ratio Core:Cache should work fine and give you better results. The PL1 limit is triggered probably because you're using Balanced Windows Power Mode. Try to set it to Performance.

Thank you so much for the helpful response. I'll try that immediately. I was actually curious about the core/cache ratio. Both are set at -134.8mV which is one step lower higher than what spits out errors when doing TS Bench at 12 threads, 1024 size. This is basically the reference frame I was going by for finding out what to set my offsets to. A couple of questions.

1. Is it an OK reference frame to go by, or should I ignore the errors there and benchmark stability in another way?
2. I tried setting my core offset to -200mV as I saw it being recommended in multiple threads. That did however, again, give me errors in TS Bench (same settings as above). Could it be ignored in this case, or does that mean it's probably not viable?
 
Last edited:
Joined
May 7, 2020
Messages
244 (0.17/day)
Thank you so much for the helpful response. I'll try that immediately. I was actually curious about the core/cache ratio. Both are set at -134.8mV which is one step lower than what spits out errors when doing TS Bench at 12 threads, 1024 size. This is basically the reference frame I was going by for finding out what to set my offsets to. A couple of questions.

1. Is it an OK reference frame to go by, or should I ignore the errors there and benchmark stability in another way?
2. I tried setting my core offset to -200mV as I saw it being recommended in multiple threads. That did however, again, give me errors in TS Bench (same settings as above). Could it be ignored in this case, or does that mean it's probably not viable?
Your errors are most likely from too much Cache offset. Find the right spot in which it gives no error. Lift it up with 10mV just in case and start tweaking the Core as low as it can handle. And yes - TS's TSBench test is a very good way to check for errors. You should play with the settings until there are no errors in tests. You could also use Cinebench R20. It's also a fast, but propper way to test the system.
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
Your errors are most likely from too much Cache offset. Find the right spot in which it gives no error. Lift it up with 10mV just in case and start tweaking the Core as low as it can handle. And yes - TS's TSBench test is a very good way to check for errors. You should play with the settings until there are no errors in tests. You could also use Cinebench R20. It's also a fast, but propper way to test the system.

Excellent!

I'll play around with it and see where I can find a sweet spot. I did change the Windows power plan to High performance, but correct me if I'm wrong, this pushes the CPU minimum frequency state to 100% at all times. Wouldn't this make the CPU always run att full speed and never down clock? I guess this doesn't matter though since I already put Speed Shift EPP to 0, but it would then also make me question why I would even need to change to the high performance power plan in Windows.
 
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,273 (1.26/day)
@HenryCase - If you want ThrottleStop to be in charge of the Speed Shift EPP variable, it is best to use the Windows High Performance power plan. This used to be the best way to prevent ThrottleStop and Windows getting in a fight over who gets to control EPP.

My opinion is that if Windows can control EPP, why not let it? Clear the Speed Shift EPP box on the main screen and then open up the FIVR window so you can monitor what Speed Shift EPP value the CPU is using. Switch through the Windows power plans back and forth and see how this changes the EPP value. If Windows is able to do an OK job at this then let it.

The Speed Shift EPP box in ThrottleStop was originally intended for older processors where the BIOS or Windows was not Speed Shift aware. You do not have to use this setting.

When testing voltages, use Cinebench R20. It is a good way to see the advantage of lowering the core voltage more than the cache voltage.

Edit - Your log files show Cinebench using a max multiplier of 36 and the TS Bench using a max multiplier of 38. Both tests should be at a solid 39.00. A Speed Shift setting of 128 can limit maximum performance. I would not set EPP higher than 80.

Also, a long term power limit of 55W is not going to be enough for your CPU to achieve maximum performance.
 
Last edited:
Joined
Jul 10, 2020
Messages
25 (0.02/day)
@HenryCase - If you want ThrottleStop to be in charge of the Speed Shift EPP variable, it is best to use the Windows High Performance power plan. This used to be the best way to prevent ThrottleStop and Windows getting in a fight over who gets to control EPP.

My opinion is that if Windows can control EPP, why not let it? Clear the Speed Shift EPP box on the main screen and then open up the FIVR window so you can monitor what Speed Shift EPP value the CPU is using. Switch through the Windows power plans back and forth and see how this changes the EPP value. If Windows is able to do an OK job at this then let it.

The Speed Shift EPP box in ThrottleStop was originally intended for older processors where the BIOS or Windows was not Speed Shift aware. You do not have to use this setting.

When testing voltages, use Cinebench R20. It is a good way to see the advantage of lowering the core voltage more than the cache voltage.

This makes a lot of sense. I'll try this out right away. In the meantime I have been fiddling around with things for a bit and I think I at least landed somewhere reasonable. Sorry for the wall of text.

This was what I was running when I first created the thread, and although it didn't give me any errors in TS Bench it did power limit throttle.

Core: -134.8mV
Cache: -134.8mV
Turbo Boost Long Power Max: 55
CPU Temp: 83C

So I played around with lowering the core offset and raising the cache offset because having the cache set to -134.8mV and the core to anything below that value would give me errors. I finally landed on the below. And this didn't give me any errors. I ran it multiple times. It was still however power limit throttling. So reading up on some more threads here and going through a bunch of other answers, based on that information I changed some other things too (I think I read that the PP0 current limit would not hurt to have at 100, so I just left it there. I did also alter the turbo ratio limits, setting all cores to 38). Next I'll try raising all cores to 39 and run Cinebench R20 to see where I'll land, but all in all this should be quite decent, right?

Core: -190.4mV
Cache: -130.9mV
CPU Temp: 78C
Turbo Boost Long Power Max: 65
PP0 Current Limit: 100

TS.png
TSLimits.png
FIVR.png
TPL.png


I am kinda befuddled though, mostly because of my own lack of knowledge, at what the difference would be between running both offsets at say -134.8mV versus running them like I've done in the second example. Are there any benefits of running the core lower than the cache if neither setting is throttling anything, or does it all come down to real world testing, I.e how does it feel in games and what are the temperature differences (for e.g. in the above examples I do get slightly better temperatures with the core set lower)? I feel like I have a gazillion questions like why are my defaults (as seen in the first screenshot) Turbo Boost Power Max settings different from others, and why was my Turbo Time Limit set to 48? I've seen other peoples being set at 28 (I changed mine to that as well).
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,273 (1.26/day)
I do get slightly better temperatures with the core set lower
That is the whole point of lowering the core more than the cache. If you are being power limit throttled, this can reduce power consumption and get you some more performance. If you are not being power limit throttled, you should see reduced temperatures. You do not have to do this. Most laptop owners are encouraged to try this because they like a fast and cool running laptop.

The default turbo time limit is 28 seconds. Individual manufacturers can set this however they like. This controls how long the CPU can use the short term 90W power limit before it switches to the long term 65W limit.

I would leave the Speed Shift Max value at 41. To control the maximum CPU speed, it is better to use the turbo ratio limits. Speed Shift Max is a good secondary way to control the CPU speed and can be used when the turbo ratios are locked.

Why not set your turbo ratios to their default values? 41, 41, 40, 40, 39, 39
You do not have any throttling or heat issues so why not run your CPU at its full rated speed like Intel intended?

You seem to be getting a good understanding of things. It takes a while to learn how all these various control methods interact. Intel made things way too complicated to try and keep all of their OEM customers happy. Hopefully your next tests will show some improved Cinebench numbers.
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
That is the whole point of lowering the core more than the cache. If you are being power limit throttled, this can reduce power consumption and get you some more performance. If you are not being power limit throttled, you should see reduced temperatures. You do not have to do this. Most laptop owners are encouraged to try this because they like a fast and cool running laptop.

The default turbo time limit is 28 seconds. Individual manufacturers can set this however they like. This controls how long the CPU can use the short term 90W power limit before it switches to the long term 65W limit.

I would leave the Speed Shift Max value at 41. To control the maximum CPU speed, it is better to use the turbo ratio limits. Speed Shift Max is a good secondary way to control the CPU speed and can be used when the turbo ratios are locked.

Why not set your turbo ratios to their default values? 41, 41, 40, 40, 39, 39
You do not have any throttling or heat issues so why not run your CPU at its full rated speed like Intel intended?

You seem to be getting a good understanding of things. It takes a while to learn how all these various control methods interact. Intel made things way too complicated to try and keep all of their OEM customers happy. Hopefully your next tests will show some improved Cinebench numbers.

You are of course 100% right!

I am getting the hang of it, it's so much fun, and I am actually getting improved results, which is an added bonus. What you said in your earlier post about hitting a solid 39.00 in both tests made something click for me in how I'll approach tinkering and testing this. New logs attached.

About the turbo ratios to their default values, I think why I set them to 38 was mainly a combination of not understanding things and hardcore following guides. I think my idea was to find a middle-point where the CPU always can perform at a stable clock speed and having temperatures in check. My logic was that a stable clock speed is going to give me a better experience (even if lower than the rated speeds) than a clock speed that is being throttled, making FPS drops.

However, this experience also made me aware of another logical fallacy of mine. I'm basing that idea off of my experiences with older laptops that weren't performing as well as this one, instead of starting fresh. Surely, this will be a good lesson for when I get another laptop (or maybe even help out friends with theirs). In any case, the attached logs are with turbo ratio limits all set to 39. I'll do some more testing with them changed to their default values.

I think I also understood something about the performance power profile in Windows. So.. to sum it up. Speed Shift EPP is unticked in the TS main window. I am using the performance profile in Windows. Do I have to untick "Enable Speed Shift when ThrottleStop starts" in TPL settings? I did have to raise the Turbo Boost Long Power Max to 70W, but I don't think it's a big deal honestly judging by the max temperatures shown in the logs. I saw you re-iterating the fact that Intel rates these CPUs safe up until 100C, so I think I'm actually quite satisfied with the latest results.

Thank you for explaining things so well. This is such a pleasant experience.

Cinebench R20.png
 

Attachments

  • Cinebench R20.txt
    7.3 KB · Views: 142
  • TS Bench.txt
    11.6 KB · Views: 138
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,273 (1.26/day)
A Cinebench R20 score of 3048 puts you tied for 6th best in the world at HWBot.


Not bad at all. It is always fun to watch someone's scores go from nothing special to something very special. The guys at the top of the charts might have faster (more expensive) memory or running Cinebench at a higher Windows priority. There are lots of tricks to get a few more points.

Your log file shows that your heatsink and fan are having no problem keeping your CPU at a reasonable temperature with power consumption hitting 76W. That means you can bump the long term power limit up to 75W or 80W without any worries.

No need to use the 38 multiplier to get some steady MHz. The 39 multiplier is also steady as long as the power limits are set appropriately.

Your laptop probably enables Speed Shift in the BIOS. If you like using Speed Shift, it will not hurt anything to use the option, Enable Speed Shift when ThrottleStop Starts. This will just double check to make sure Speed Shift is enabled. This was included for older laptops that do not enable Speed Shift in the BIOS. No problem checking this on newer laptops.

With Speed Shift EPP unchecked on the main TS window, open the FIVR window and look at the monitoring table data at the top right. The last line in that table constantly shows what EPP value the CPU is using. Switch Windows power profiles and you should see that number change. If you can get the EPP value you want by adjusting Windows, there is no need to check the Speed Shift EPP option on the main screen of ThrottleStop. Letting Windows control it is OK.

On older computers or older versions of Windows, changing power profiles did not change the Speed Shift EPP value that the CPU was using. It those situations, you can check the Speed Shift EPP option on the main TS screen to control EPP. Making sense? ThrottleStop has a lot of options to cover a lot of unique situations. No need to use any of these options if you do not need to.
You seem to be discovering the important options. :toast:
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
A Cinebench R20 score of 3048 puts you tied for 6th best in the world at HWBot.


Not bad at all. It is always fun to watch someone's scores go from nothing special to something very special. The guys at the top of the charts might have faster (more expensive) memory or running Cinebench at a higher Windows priority. There are lots of tricks to get a few more points.

Your log file shows that your heatsink and fan are having no problem keeping your CPU at a reasonable temperature with power consumption hitting 76W. That means you can bump the long term power limit up to 75W or 80W without any worries.

No need to use the 38 multiplier to get some steady MHz. The 39 multiplier is also steady as long as the power limits are set appropriately.

Your laptop probably enables Speed Shift in the BIOS. If you like using Speed Shift, it will not hurt anything to use the option, Enable Speed Shift when ThrottleStop Starts. This will just double check to make sure Speed Shift is enabled. This was included for older laptops that do not enable Speed Shift in the BIOS. No problem checking this on newer laptops.

With Speed Shift EPP unchecked on the main TS window, open the FIVR window and look at the monitoring table data at the top right. The last line in that table constantly shows what EPP value the CPU is using. Switch Windows power profiles and you should see that number change. If you can get the EPP value you want by adjusting Windows, there is no need to check the Speed Shift EPP option on the main screen of ThrottleStop. Letting Windows control it is OK.

On older computers or older versions of Windows, changing power profiles did not change the Speed Shift EPP value that the CPU was using. It those situations, you can check the Speed Shift EPP option on the main TS screen to control EPP. Making sense? ThrottleStop has a lot of options to cover a lot of unique situations. No need to use any of these options if you do not need to.
You seem to be discovering the important options. :toast:

Yeah, that makes complete sense.

I did fiddle with the Windows profiles with the FIVR window opened, and I mean the only conclusion I can draw is that the Performance profile puts the EPP at 0 pretty much the whole time while the Balanced profile puts it at 84 for the majority of the time. I think both of those are doing what they're supposed to do. I'm thinking that when I feel comfortable enough around the majority of these settings I'll go ahead and setup a couple of choice profiles in TS for specific situations. At home I have a cooling pad (which I excluded from all the testings done so far, just to be as accurate as can be with what the machine can do on it's own), but when I travel for work, which I do 1-2 times per month the cooling pad becomes a nuisance to drag along. The laptop itself is enough of a burden to carry. In any case, would there be any potential conflicts if I were to use TS to control the EPP, or would that just override Windows? For e.g. if none of the values Windows sets suits me.

And the tied 6th place thing, I dunno what to think of that. It's probably no big deal, but I have to admit it did make me giddy. Now imagine this. I changed the turbo ratios to the default (although calling them default is a misnomer since apparently on my machine the default is 41, 41, 40, 40, 39, 38. I'm using as suggested 41, 41, 40, 40, 39, 39) and lo and behold, not sure if I'm dreaming but according to that list this should grant me a bronze placement :twitch:

[Edit]: I set the Cinebench process priority to high and not sure if that made the log behave oddly. Looks rather short.

Cinebench R20_3098.png
 

Attachments

  • Cinebench R20_3098.txt
    2.9 KB · Views: 146
Last edited:
Joined
Jul 10, 2020
Messages
25 (0.02/day)
Sorry for dragging up an old thread.

I was playing some Death Stranding and I bluescreened once, unfortunately I wasn't logging anything in TS.
I raised the cache a tad and tweeked the core as well, and since then no more blue screens.

I am however hitting some limits. Thought maybe I'd share and see if you guys have any insights.

Settings, and log is attached (I didn't want to fiddle too much with ingame settings before actually posting here, but naturally I'll do that too after some discussion). I did however try changing a minor detail. I changed the fps limiter in game, from 144 to 60. Naturally temperatures were better on 60 but I was still hitting the limiters. However, from what I could tell from the 1 hour session, neither setting affected my gameplay that much to be honest. Like I said, the only thing affected was the temperatures and I guess that is to be expected.

(Side note, I don't really need to play a game like Death Stranding at 144 fps, it does just fine at 60 fps)

1.png
2.png
3.png
4.png
 

Attachments

  • 2020-07-18.txt
    127.2 KB · Views: 204

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,273 (1.26/day)
Your CPU and GPU are running fine when gaming. Temperatures are within spec and both are running at a consistent speed.

I raised the cache a tad and tweeked the core as well, and since then no more blue screens.
That is all you can do. A blue screen when under volting is almost always because your CPU is not getting enough voltage. Keep giving your CPU more voltage until it is 100% game stable.

Maybe -100 mV for the cache and -180 mV for the core will be 100% stable. Some users are always chasing after the bare minimum voltage that they can get away with. I like using a little bit of extra voltage (10 mV or 20 mV) when I reach the bare minimum point to help ensure 100% stability.
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
Your CPU and GPU are running fine when gaming. Temperatures are within spec and both are running at a consistent speed.


That is all you can do. A blue screen when under volting is almost always because your CPU is not getting enough voltage. Keep giving your CPU more voltage until it is 100% game stable.

Maybe -100 mV for the cache and -180 mV for the core will be 100% stable. Some users are always chasing after the bare minimum voltage that they can get away with. I like using a little bit of extra voltage (10 mV or 20 mV) when I reach the bare minimum point to help ensure 100% stability.

That makes a lot of sense. That's pretty much what I wanted to hear as well
Thank you, once again :)
 
Joined
May 10, 2020
Messages
168 (0.12/day)
A Cinebench R20 score of 3048 puts you tied for 6th best in the world at HWBot.


Not bad at all. It is always fun to watch someone's scores go from nothing special to something very special. The guys at the top of the charts might have faster (more expensive) memory or running Cinebench at a higher Windows priority. There are lots of tricks to get a few more points.

Your log file shows that your heatsink and fan are having no problem keeping your CPU at a reasonable temperature with power consumption hitting 76W. That means you can bump the long term power limit up to 75W or 80W without any worries.

No need to use the 38 multiplier to get some steady MHz. The 39 multiplier is also steady as long as the power limits are set appropriately.

Your laptop probably enables Speed Shift in the BIOS. If you like using Speed Shift, it will not hurt anything to use the option, Enable Speed Shift when ThrottleStop Starts. This will just double check to make sure Speed Shift is enabled. This was included for older laptops that do not enable Speed Shift in the BIOS. No problem checking this on newer laptops.

With Speed Shift EPP unchecked on the main TS window, open the FIVR window and look at the monitoring table data at the top right. The last line in that table constantly shows what EPP value the CPU is using. Switch Windows power profiles and you should see that number change. If you can get the EPP value you want by adjusting Windows, there is no need to check the Speed Shift EPP option on the main screen of ThrottleStop. Letting Windows control it is OK.

On older computers or older versions of Windows, changing power profiles did not change the Speed Shift EPP value that the CPU was using. It those situations, you can check the Speed Shift EPP option on the main TS screen to control EPP. Making sense? ThrottleStop has a lot of options to cover a lot of unique situations. No need to use any of these options if you do not need to.
You seem to be discovering the important options. :toast:
I would really like to know how Mr Fox did that 3342 with 9750h. :D I can pay 20€ for just knowing.
 

Attachments

  • lastrun.PNG
    lastrun.PNG
    284.4 KB · Views: 352
Last edited:
Joined
May 7, 2020
Messages
244 (0.17/day)
I would really like to know how Mr Fox did that 3342 with 9750h. :D I can pay 20€ for just knowing.
I would like to know too. After updating to Windows 2004 yesterday, mine reached 3141 (before was 1119). Will split the 20€ with you :)
 
Joined
Jul 18, 2020
Messages
4 (0.00/day)
Location
Bengaluru
System Name ACER PREDATOR PH317-53
Processor i7-9750
Motherboard CFL
Cooling Aero Blade 3D
Memory 16 GB (DDR4)
Video Card(s) NVIDIA RTX 2070 with Max-Q Design (8GB)
Storage 2 TB
Display(s) LCD
Audio Device(s) WNX
Power Supply 180W
Mouse HP Omen Mouse x9000
Keyboard ANT Esports Mechanical
Software Cinebench
Benchmark Scores 3005
Hi UncleWebb,

First and foremost, your in-depth knowledge is commendable. I expected no less from the creator of ThrottleStop. I have seen excellent performance improvement in my system as well. You are the one person who gave the usage importance on TTL (Turbo Time Limit). Believe me, every video I have ever watched, they have either set to 3670016 or they have kept it untouched (in both cases saying they don't know what the setting does). However, here in the forums, only because of you (and thanks to the other person whom you had replied to in your post), I was able to know what that Limit was and why it should be set to 28 seconds. It seems, however, that I have let to learn. Thanks to you, I developed a keen interest in TS. I analyze TS Logs (thanks to your insight on some of the replies to our friends). My thirst for knowledge and curiosity on TS has exponentially increased! I have come to the conclusion I want to learn it all. I am 99% sure and there is a hunch that, I might actually be doing the wrong way. We experiment, bring voltages down, and then crash the System until we find a stable voltage. But, as much as I would love to say that I know something, I believe that there is more to it than just bringing the voltages down. And only you could enlighten me more about it.
One thing which I got to know for certain is the equation which is in play (or the reason for this beautiful Application):
ΣVi (Input Voltages) = ΣVo (Output Voltages; required to run the System) + Heat (H)
Hence,
↑ Input Voltages (Vi) will have ↑ Heat (H) leading to ↑Input Power.
Conversely, ↓ Vi will ultimately lead to ↓H leading to ↓Power consumption.

Kevin aka. UncleWebb, I am really fascinated with your application and heartily thank you for saving millions of machines out there from getting fried. Statistics suggest that there are approximately 2.5 billion gamers out there in the entire world. Apart from being highly lightweight, your application renders all that and even more than what XTU has to offer!

WHAT I WANT
I have just begun my journey in ThrottleStop. I request your expertise on all the necessary settings which need to be checked, their brief description (and the reason why we are making such a change) and how do I go about making changes (i.e. what changes would you suggest to me to make first: TPL window/ Fully Integrated Voltage Regulator (FIVR)? How to do an in-depth analysis of log files? I would like you to explain the various error codes which we stumble upon when analyzing the logs/ the parameters which I need to carefully look upon? I do not believe in frying my Laptop just to run games in Ultra High settings. I believe in creating my own presets (where it is not too shabby or too extraordinarily beautiful but optimal enough to get the best results for suitable environments).
I would also like you to stress on the main window (Speedshift, Speedstep, BiDirectional Processor Hot) of ThrottleStop and Turbo Power Limit (TPL) window, the CPU core, CPU Cache, and Intel GPU options, Turbo Ratio Limit, Non-Turbo Ratio in FIVR, the various things we change there and why do we change it in the first place, its significance. If there is something else I have missed please enlighten me with that.

I have recorded various stable profiles changing the Turbo Ratio Limits, changing the CPU Core/Cache voltages, and the Speed Shift limits. In all the cases I have kept the Turbo boost Long Power Max and Short Power Max at 75 and 100 respectively. I would like you to stress more on these two, when and under what case do we make changes in them. Why do we have to change it in the first place and what are their significances.

Lastly, I have attached a screenshot of my "Benchmark" Profile. I get consistent 3.993GHz (for a limited time) then, as the temps go high or the Pkg power goes high there is a Throttle and it falls to 3.2GHz. In the Cinebench benchmark window, as you can see the maximum I have touched in my i7-9750H is 3005. I am 100% sure if I get that 40.00 FID (Frequency Identifier) continuously, I can easily cross the 3005 mark. Is there a way I can achieve that consistent result? Point to note is for that high cinebench score I used EXP PROF 6 from the TS profiles.xlsx file. There was a consistent performance at first (FID: 40.00) then dropped to 34.00, then 32.00). Request your valuable insight on the same. I really want to learn and be your student. I can assure you, I shall not disappoint you in any manner. I am a quick learner and a curious cat.
Hope you are safe in this COVID-19 pandemic situation and looking forward to hearing from you soon.

Your admirer!
Regards,
Saomya

Attachments:
(1) A couple of TS Log files. (it will be great if you could explain to me about the terminologies like MULTI, C0% (I believe as you said in one of your posts it is the application memory consumed when the system is idle), BAT_mW (after careful observation I found out that in most cases it is 0 however, in many cases it is in-between range -42874 to 42258). What is that any why is that?
(2) Associated image files of TS profiles.
(3) Self-created TS profiles.xlsx
 

Attachments

  • TS profiles.xlsx
    25.4 KB · Views: 229
  • 2020-07-17.txt
    1,003.5 KB · Views: 103
  • 2020-07-18.txt
    950.7 KB · Views: 125
  • 2020-07-19.txt
    933.7 KB · Views: 132
  • how do I maintain that frequency.png
    how do I maintain that frequency.png
    540.3 KB · Views: 292
Joined
May 10, 2020
Messages
168 (0.12/day)
Hi UncleWebb,

First and foremost, your in-depth knowledge is commendable. I expected no less from the creator of ThrottleStop. I have seen excellent performance improvement in my system as well. You are the one person who gave the usage importance on TTL (Turbo Time Limit). Believe me, every video I have ever watched, they have either set to 3670016 or they have kept it untouched (in both cases saying they don't know what the setting does). However, here in the forums, only because of you (and thanks to the other person whom you had replied to in your post), I was able to know what that Limit was and why it should be set to 28 seconds. It seems, however, that I have let to learn. Thanks to you, I developed a keen interest in TS. I analyze TS Logs (thanks to your insight on some of the replies to our friends). My thirst for knowledge and curiosity on TS has exponentially increased! I have come to the conclusion I want to learn it all. I am 99% sure and there is a hunch that, I might actually be doing the wrong way. We experiment, bring voltages down, and then crash the System until we find a stable voltage. But, as much as I would love to say that I know something, I believe that there is more to it than just bringing the voltages down. And only you could enlighten me more about it.
One thing which I got to know for certain is the equation which is in play (or the reason for this beautiful Application):
ΣVi (Input Voltages) = ΣVo (Output Voltages; required to run the System) + Heat (H)
Hence,
↑ Input Voltages (Vi) will have ↑ Heat (H) leading to ↑Input Power.
Conversely, ↓ Vi will ultimately lead to ↓H leading to ↓Power consumption.

Kevin aka. UncleWebb, I am really fascinated with your application and heartily thank you for saving millions of machines out there from getting fried. Statistics suggest that there are approximately 2.5 billion gamers out there in the entire world. Apart from being highly lightweight, your application renders all that and even more than what XTU has to offer!

WHAT I WANT
I have just begun my journey in ThrottleStop. I request your expertise on all the necessary settings which need to be checked, their brief description (and the reason why we are making such a change) and how do I go about making changes (i.e. what changes would you suggest to me to make first: TPL window/ Fully Integrated Voltage Regulator (FIVR)? How to do an in-depth analysis of log files? I would like you to explain the various error codes which we stumble upon when analyzing the logs/ the parameters which I need to carefully look upon? I do not believe in frying my Laptop just to run games in Ultra High settings. I believe in creating my own presets (where it is not too shabby or too extraordinarily beautiful but optimal enough to get the best results for suitable environments).
I would also like you to stress on the main window (Speedshift, Speedstep, BiDirectional Processor Hot) of ThrottleStop and Turbo Power Limit (TPL) window, the CPU core, CPU Cache, and Intel GPU options, Turbo Ratio Limit, Non-Turbo Ratio in FIVR, the various things we change there and why do we change it in the first place, its significance. If there is something else I have missed please enlighten me with that.

I have recorded various stable profiles changing the Turbo Ratio Limits, changing the CPU Core/Cache voltages, and the Speed Shift limits. In all the cases I have kept the Turbo boost Long Power Max and Short Power Max at 75 and 100 respectively. I would like you to stress more on these two, when and under what case do we make changes in them. Why do we have to change it in the first place and what are their significances.

Lastly, I have attached a screenshot of my "Benchmark" Profile. I get consistent 3.993GHz (for a limited time) then, as the temps go high or the Pkg power goes high there is a Throttle and it falls to 3.2GHz. In the Cinebench benchmark window, as you can see the maximum I have touched in my i7-9750H is 3005. I am 100% sure if I get that 40.00 FID (Frequency Identifier) continuously, I can easily cross the 3005 mark. Is there a way I can achieve that consistent result? Point to note is for that high cinebench score I used EXP PROF 6 from the TS profiles.xlsx file. There was a consistent performance at first (FID: 40.00) then dropped to 34.00, then 32.00). Request your valuable insight on the same. I really want to learn and be your student. I can assure you, I shall not disappoint you in any manner. I am a quick learner and a curious cat.
Hope you are safe in this COVID-19 pandemic situation and looking forward to hearing from you soon.

Your admirer!
Regards,
Saomya

Attachments:
(1) A couple of TS Log files. (it will be great if you could explain to me about the terminologies like MULTI, C0% (I believe as you said in one of your posts it is the application memory consumed when the system is idle), BAT_mW (after careful observation I found out that in most cases it is 0 however, in many cases it is in-between range -42874 to 42258). What is that any why is that?
(2) Associated image files of TS profiles.
(3) Self-created TS profiles.xlsx
Your cpu is running good. But why you have speedsift at 120 if you making benchmarks? Set it to 0 and try again. That fid is also normal. It doesnt matter what you set to TPL because your laptop seems to be like mine. You can set it to 200 or maximum. It just doesnt change anything. You have to just find best undervolt settings for cpu and cache.
 
Joined
Jul 18, 2020
Messages
4 (0.00/day)
Location
Bengaluru
System Name ACER PREDATOR PH317-53
Processor i7-9750
Motherboard CFL
Cooling Aero Blade 3D
Memory 16 GB (DDR4)
Video Card(s) NVIDIA RTX 2070 with Max-Q Design (8GB)
Storage 2 TB
Display(s) LCD
Audio Device(s) WNX
Power Supply 180W
Mouse HP Omen Mouse x9000
Keyboard ANT Esports Mechanical
Software Cinebench
Benchmark Scores 3005
Your cpu is running good. But why you have speedsift at 120 if you making benchmarks? Set it to 0 and try again. That fid is also normal. It doesnt matter what you set to TPL because your laptop seems to be like mine. You can set it to 200 or maximum. It just doesnt change anything. You have to just find best undervolt settings for cpu and cache.
Thank you ereko for the suggestion. I am looking more into doing deep analysis on the application itself.
 
Joined
Jul 10, 2020
Messages
25 (0.02/day)
Thank you ereko for the suggestion. I am looking more into doing deep analysis on the application itself.

He made a write-up here, http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/ detailing out information about the options. This is another one that contains information that might be useful, https://www.ultrabookreview.com/31385-the-throttlestop-guide/

I'm not sure if they contain what you're looking for but they might be useful resources nonetheless :)
 
Joined
Jul 18, 2020
Messages
4 (0.00/day)
Location
Bengaluru
System Name ACER PREDATOR PH317-53
Processor i7-9750
Motherboard CFL
Cooling Aero Blade 3D
Memory 16 GB (DDR4)
Video Card(s) NVIDIA RTX 2070 with Max-Q Design (8GB)
Storage 2 TB
Display(s) LCD
Audio Device(s) WNX
Power Supply 180W
Mouse HP Omen Mouse x9000
Keyboard ANT Esports Mechanical
Software Cinebench
Benchmark Scores 3005
Joined
Jul 10, 2020
Messages
25 (0.02/day)
Hello!

I've ran into a bit of an odd one. This is my initial thread (https://www.techpowerup.com/forums/...rvolting-thermal-throttling-questions.269674/) when I first caught the undervolting bug. I'm using the same laptop, a Clevo PB71EF-G. A couple of days ago I tried moving over to Linux, I used this tool to undervolt under Linux (https://github.com/georgewhewell/undervolt) and the settings would stick, no problem.

Now when I'm back on Windows I can't seem to make the settings change. The FIVR doesn't appear to be locked (which would be strange anyway since it never was locked). I've tried downgrading the BIOS, but still no luck. See the attached picture below for FIVR settings.

[Edit]

Another interesting thing is that this BIOS comes with a modest undervolt already pre-applied. And even though I can see that setting being applied in the BIOS, ThrottleStop doesn't seem to be able to see it. So normally, if I installed ThrottleStop on a fresh Windows installation, I would see that undervolt in the FIVR window.

Hello!

I've ran into a bit of an odd one. This is my initial thread (https://www.techpowerup.com/forums/...rvolting-thermal-throttling-questions.269674/) when I first caught the undervolting bug. I'm using the same laptop, a Clevo PB71EF-G. A couple of days ago I tried moving over to Linux, I used this tool to undervolt under Linux (https://github.com/georgewhewell/undervolt) and the settings would stick, no problem.

Now when I'm back on Windows I can't seem to make the settings change. The FIVR doesn't appear to be locked (which would be strange anyway since it never was locked). I've tried downgrading the BIOS, but still no luck. See the attached picture below for FIVR settings.

[Edit]

Another interesting thing is that this BIOS comes with a modest undervolt already pre-applied. And even though I can see that setting being applied in the BIOS, ThrottleStop doesn't seem to be able to see it. So normally, if I installed ThrottleStop on a fresh Windows installation, I would see that undervolt in the FIVR window.

Alright. I've solved this, by just stumbling upon a solution. I tried installing Intel XTU just to see if the issue was reflected there as well. It wouldn't even start, complaining about an incompatible OS feature being enabled 'Virtual Machine System'. The first thing I did on this fresh install was activate WSL2 so I could mount one of the ext4 drives and gain access to the data. So apparently MS has restricted access to the 0x150 MSR registers when WSL2 is active... I have no clue why this is.

Seeing how I have zero use for it atm, it's an easy "fix" tho. @unclewebb, maybe ThrottleStop can have a small check during startup and return a reasonable error/warning message? :)
 

Attachments

  • FIVR.png
    FIVR.png
    52.5 KB · Views: 65
Last edited:

eidairaman1

The Exiled Airman
Joined
Jul 2, 2007
Messages
40,435 (6.61/day)
Location
Republic of Texas (True Patriot)
System Name PCGOD
Processor AMD FX 8350@ 5.0GHz
Motherboard Asus TUF 990FX Sabertooth R2 2901 Bios
Cooling Scythe Ashura, 2×BitFenix 230mm Spectre Pro LED (Blue,Green), 2x BitFenix 140mm Spectre Pro LED
Memory 16 GB Gskill Ripjaws X 2133 (2400 OC, 10-10-12-20-20, 1T, 1.65V)
Video Card(s) AMD Radeon 290 Sapphire Vapor-X
Storage Samsung 840 Pro 256GB, WD Velociraptor 1TB
Display(s) NEC Multisync LCD 1700V (Display Port Adapter)
Case AeroCool Xpredator Evil Blue Edition
Audio Device(s) Creative Labs Sound Blaster ZxR
Power Supply Seasonic 1250 XM2 Series (XP3)
Mouse Roccat Kone XTD
Keyboard Roccat Ryos MK Pro
Software Windows 7 Pro 64
So why not update the old thread?
 

Tatty_Two

Gone Fishing
Joined
Jan 18, 2006
Messages
25,793 (3.88/day)
Location
Worcestershire, UK
Processor Rocket Lake Core i5 11600K @ 5 Ghz with PL tweaks
Motherboard MSI MAG Z490 TOMAHAWK
Cooling Thermalright Peerless Assassin 120SE + 4 Phanteks 140mm case fans
Memory 32GB (4 x 8GB SR) Patriot Viper Steel 4133Mhz DDR4 @ 3600Mhz CL14@1.45v Gear 1
Video Card(s) Asus Dual RTX 4070 OC
Storage WD Blue SN550 1TB M.2 NVME//Crucial MX500 500GB SSD (OS)
Display(s) AOC Q2781PQ 27 inch Ultra Slim 2560 x 1440 IPS
Case Phanteks Enthoo Pro M Windowed - Gunmetal
Audio Device(s) Onboard Realtek ALC1200/SPDIF to Sony AVR @ 5.1
Power Supply Seasonic CORE GM650w Gold Semi modular
Mouse Coolermaster Storm Octane wired
Keyboard Element Gaming Carbon Mk2 Tournament Mech
Software Win 10 Home x64
I have merged these posts into the old thread ....... carry on.
 
Top