Throttlestop Functionality Suggestions + Inquiries

[8Bit.Pixels]

Hello!
I'm quite new to the forum and the Throttlestop application, but just wanted to say thank you so much for creating this amazing piece of software, it has really allowed me to get the very most out of my CPU and helped me to reduce temps significantly - my Cinebench score has increased from 7000 to 10000, whilst max temps have dropped from 100C to under 90C!
But I was wondering if I could make a few suggestions and inquiries to help improve the user experience and help me to get the most out of Throttlestop's incredibly useful functionality, as well as helping me to understand the software better:
(in terms of system specs, I have a laptop, and you should be able to see the full specs under my profile, as I have filled them out under settings)

1. Would it be possible to implement functionality for more than 4 profiles? I find that some applications are more stressful that others, and cause my CPU to use more power and heat up more. This means, whilst for example in Cinebench I might allow my CPU to draw 55-60W in order to reach sustained boost clocks of ~4.4GHz for under 90C, Minecraft on the other hand may be more unpredictable, drawing the same sort of power but occasionally spiking and causing heating over 90C (due to different core loads maybe? I'm not really sure), which I want to avoid. More than just 4 profiles (which I already use for 'no cooling pad' (for when I'm not at home but plugged in), 'cooling pad' (for when I'm at home and plugged in, using my large cooling pad), 'battery' (for battery power - lower power limits, different undervolt etc) and one for my other, smaller cooling pad I use for travel) would allow me to make application-specific profiles to allow maximum performance whilst minimising temperatures.

2. Leading on from that first point, I was wondering if it is possible to implement a system where the application could monitor the CPU temperature and set the power limits accordingly? What I mean by this is: for example, if my CPU was under 90C, than the application could activate a PL2 of a higher value, say 60W, until the CPU exceeds 90C, causing the application to drop the power limits to a PL1 of say 50-55W until the CPU goes under 85C (these are just example values - the actual CPU power draw at which the CPU hits 90C / reduces under 85C may differ). This would be similar to making multiple profiles, but it could be a more efficient and reliable system from keeping the temperatures under 90C. If this is already possible using other techniques or applications than please let me know, as this would be very helpful.

3. What is the rule for determining when PL1 should turn into PL2? I understand that you can configure the length of time that the CPU's power exceeds PL1, but once that time limit is hit and the CPU power drops back to under PL1, what is the rule for the length of time that it waits until PL2 is activated again? Would it be possible to configure this time delay? Or even better: add temperature triggers as mentioned in the previous point?

4. A rather big enquiry but: would it be possible to develop a Throttlestop for GPUs instead of CPUs? I want to be able to tweak my GPU configuration with the same level of accuracy and detail as I can with my CPU, to edit things such as power limits, undervolts and clock speed adjustments. Currently I feel my GPU is not achieving its full potential, as under max load it draws 75W and only ever hits a max temperature of 65C; I feel like increasing the power limits could allow for higher clock speeds and greater performance, as there is quite a lot of thermal headroom available (thermal limit of my GPU: ~87C). I have already done some basic overclocking using GHelper but I think an application like Throttlestop would give far more control and range, and it would feel like all my GPU settings were in one place. I don't actually know if my GPU supports changing power limits, but maybe an application like Throttlestop could implement workarounds? I remember I had to disable some BIOS settings to get undervolting to work on my CPU. Again, if this is already possible, a run through of how to do it would be greatly appreciated.

5. I also notice in GHelper there is the option to configure different 'CPU Boosts', like 'Aggressive' or 'Balanced' etc. I am aware of what the different levels of boost do, but I was wondering if it could become a part of Throttlestop so I could implement it with existing profiles. Or do the CPU Boosts equate to some settings already available in Throttlestop, in which case, which ones and how do they work?

Thank you so much in advance for taking the time to help me, and I am really grateful for the incredible software you produce - it really is extremely useful!

 

[unclewebb]

Unfortunately there is no money to be made writing free software. I asked the user community a long time ago if they would be interested in financially supporting project ThrottleStop The overwhelming majority said thanks but no thanks. Users are happy using ThrottleStop for free. If you had a job and your boss never paid you, would you keep going to work? Probably not. Definitely not for very long. I still like working on ThrottleStop but I only do it as a hobby now.

1) I only use a single profile. There are no plans to expand the number of profiles in ThrottleStop.

2) You could try setting up two different profiles in ThrottleStop with each profile set to use different power limits. Check the Alarm box In the Options window. Set the DTS value to 20 if you want your computer to change profiles at 80°C. If you want ThrottleStop to switch to Profile 2 at high temperatures, set the Use Profile value to 2.

Hopefully this is good enough. The Alarm feature is rarely if ever used. You have some good ideas but it would take a significant reworking of ThrottleStop. I would rather be outside during the summer.

causing heating over 90C

That is really not a problem. Intel CPUs can run reliably at high temperatures. Intel thermal throttling works smoothly. You can let the CPU reach the thermal throttling temperature when playing Minecraft. You are not going to hurt anything. You do not need to micro manage the power limits to try to keep the CPU at a safe temperature. Any temperature under 100°C is equally safe. That is why Intel decided to set the thermal throttling temperature to 100°C for the vast majority of their CPUs. Some of the newer Intel CPUs can safely run at up to 105°C or even 110°C. Many users worry too much about high temperatures when you really do not have to.

3. What is the rule for determining when PL1 should turn into PL2?

Intel does not publicly document any of their rules. The turbo time limit you can request in ThrottleStop is not a written in stone kind of number. When there is a significant load on the CPU, it usually switches from the PL2 power limit to the PL1 power limit much sooner compared to if the CPU is lightly loaded and running just barely over the PL1 limit. I usually just set both turbo power limits sky high so they do not cause any throttling.

4. A rather big enquiry but: would it be possible to develop a Throttlestop for GPUs instead of CPUs?

I think Nvidia GPUs are more locked down compared to Intel CPUs. I am not sure what is possible. If no one has ever written a ThrottleStop like app for Nvidia GPUs then I doubt it will ever happen. If it was possible, someone would have done this already. No more life consuming freeware projects for me.

thank you so much for creating this amazing piece of software

You are most welcome.

 

[8Bit.Pixels]

Hi unclewebb,
Thanks so much for the info!

You could try setting up two different profiles in ThrottleStop with each profile set to use different power limits.

Ok so I tried creating two different profiles, one for 'Above 90C' and one for 'Under 90C', each with different power limits, and I used the Alarm feature as you suggested. This did work, but it ended up switching quite fast. As it went above 90C, it cut the power so it dropped below 90C, and then subsequently raised the power, causing the temp to exceed 90C once more. So it did work, but it just didn't feel quite right so I think I'm going to stick with your idea of just using a single profile and setting the power limits best I can. If, as you say, it really isn't a problem for my CPU to be averaging in the mid 90s whilst under stress, then this will be fine.
I repasted my CPU about a year ago, but I don't think I did it quite correctly as my CPU was still throttling under load (this was before I discovered ThrottleStop) and constantly thermal throttling when I was playing games - this was prolonged for about 9 months until I got round to repasting again and installing ThrottleStop, and since then my thermals have been much better. However, I've noticed that the temperatures have increased again. I only repasted about 2 months ago but I've had to severely cut the power limits in order to stop thermal throttling - that Cinebench score of 10000 was immediately after repasting with unlimited power limits (CPU would draw ~60W) and I haven't been able to replicate it since - it's dropped back to 7-8k with much lower power limits (35-40W), in order to minimise throttling. I don't know if this is due to heatsink warpage causing bad contact with the CPU, or maybe I put too much paste on causing pump-out and less paste actually remaining on top of the die (which would explain the gradual increase in temps recently). I'm still trying to get to the bottom of that one.
As for power limit tuning, do you recommend setting the power limits as high as possible until thermal throttling occurs? For example, if my CPU started throttling at 95C, should I set my power limits to try and stay below that? Is the temp where thermal throttling starts the temperature I should aim to stay below?

I think Nvidia GPUs are more locked down compared to Intel CPUs.

Yeah I think you're right, I haven't heard much about tuning a GPU with that kind of detail before. The power limits and voltage are probably locked by my BIOS anyways. I'll make do with basic clock speed adjustments. I was just wondering if it was possible really, and of course I wouldn't want you to sacrifice your time developing more stuff. But I am really grateful for the work you've put in already, and the fact you're still actively helping people like me on these forums is amazing. But yeah, of course, go enjoy your summer man!

Thanks again and have a great day!

 

[unclewebb]

I only repasted about 2 months ago but I've had to severely cut the power limits

What thermal paste did you use? Honeywell PTM 7950 provides consistent long term results. Many other popular thermal pastes do not. Some are horrible and within a few weeks or a month there will be little paste left between the CPU and the heatsink. Here's my fav video that shows this.

I am not sure what is better, power limit throttling or thermal throttling. Of course no throttle would be best. You can use ThrottleStop to play around with the power limits but first try replacing the thermal paste.

Thermal throttling is designed to always keep the CPU at a safe operating temperature. The CPU will take care of itself without you having to worry about it. Intel does not expect their customers to constantly monitor the CPU temperature and then vibrate and worry that the end of their CPU is near. I used to worry too about high temperatures but not anymore. I never hear about any CPU degradation because the CPU operating temperature was within the Intel spec. Anything under 100°C is within spec. Intel has more problems with high voltage than high temperatures.

I was never a big fan or user of the ThrottleStop Alarm feature. Wild swings in performance back and forth are not good for smooth game play. I could have put some more effort into this feature but it never seemed that necessary.

 

[8Bit.Pixels]

What thermal paste did you use?

So I've been using Thermal Grizzly Kryonaut, as I've heard it's one of the best. But now, doing some research, it turns out it is prone to drying out under high temperatures, like what my laptop experiences regularly! So turns out that probably wasn't the best choice after all; even though it is very well performing, I think it is more designed for desktop PCs that don't reach nearly as high temperatures. Here's a picture of the Kryonaut paste on my CPU (right) and GPU (left) after 9 months of use:

You can see how most of the paste has ended up around the die! So yeah, looks like the Honeywell pads are much better suited to laptops, so I think that's going to be my next buy. I'm probably also going to do some research into which thermal compound to buy for the VRAMs and other components surrounding my GPU and CPU, as you can see there doesn't seem to be much left. If these Honeywell pads are supposed to last for a long time, then I want to get everything right now so I don't have to open it up for a while.

Thermal throttling is designed to always keep the CPU at a safe operating temperature. The CPU will take care of itself without you having to worry about it.

Ok that's good to know, I will be a bit more lenient with my temperature monitoring from now on. Just out of interest, what kind of temperatures does your CPU reach under load?

Intel has more problems with high voltage than high temperatures.

Thats interesting. I'm glad I've been able to get undervolting working on my CPU, for a while I thought I couldn't due to hard-coded restrictions, but it turns out I just needed to disable virtualisation and a couple other BIOS settings. I've discovered that my CPU is stable at around -90mV, so I'm happy with that.

Thanks again for your help, I really do appreciate it! Have a great day.

 

[unclewebb]

what kind of temperatures does your CPU reach under load?

99.9999°C

Any higher than that and thermal throttling kicks in to save the CPU from any long term damage. I think you are worrying about temperatures again.

When you do a search for thermal paste, almost all of the results you will find are comparisons of how well a thermal paste works when applied to a desktop CPU, usually for about 5 minutes before the next thermal paste is tested. These kind of tests are useless for trying to figure out if a thermal paste is any good for long term use when applied direct die to a mobile CPU.

Kryonaut has an excellent reputation when used on desktop CPUs. One look at the picture you posted and it would be hard for me to recommend using it on a mobile CPU.

Honeywell PTM 7950 has a good reputation amongst laptop owners based on word of mouth. It just seems to work consistently during long term real world use.

Lenovo uses their own version of PTM 7950. I have a 10+ year old Lenovo laptop. I have never once thought about changing the thermal paste. That is a good thing because it was poorly designed for servicing.

https://www.reddit.com/r/LenovoLegion/comments/tfxomi/lenovos_stock_paste_is_actually_ptm7958_a_custom

 

[8Bit.Pixels]

99.9999°C

Ok that makes me feel better

It just seems to work consistently during long term real world use.

Yeah from the research I've done that seems to be the case. I will replace my thermal paste with Honeywell over the summer, and hopefully there will be a big difference! From some performance comparison videos I've watched between Kryonaut and Honeywell in laptops, it seems Honeywell outperforms by a decent margin.

Thanks once more unclewebb, have a good day