A few questions about Throttlestop

[WickedSick277]

Hello everyone!
Recently i used Throttlestop to undervolt, and great results were achieved on CPU(~-15c).

My questions:
1: Why the PROCHOT 92 is turned red while i play games, even though sensors aren't really reporting that high? => 'TS - Gaming' & 'Temps - gaming' screenshots provided below.
2: I know undervolting vary from machine to machine, but could you guys please double-check my options? i sent them below. (Specially iGPU section)
3: why does GPU-Z show thermal resason for Perfcap at gaming?

*: all undervolt options for both plans are same except Turbo cores.
**: Since there are lots of shots, as a recap, my temps on gaming are: 70 CPU, 85-90 GPU on Gaming plan, and 65 CPU, 85-90 GPU on my battery plan but plugged IN. which in both scenarios my GPU is hot!
Thanks in advance!
My rig is: Nitro an515-54, GTX 1650, 9750H intel.

 

[unclewebb]

even though sensors aren't really reporting that high?

PROCHOT is turning red because the sensors are reporting high temperatures. Your computer is set to start thermal throttling at 92°C instead of the Intel recommended 100°C. ThrottleStop shows a maximum temperature of 91°C so yes, your computer really is reaching 92°C. Thermal throttling is designed to keep the CPU just a hair under 92°C which your computer is set to start throttling at. In the Options window you can see that Acer set the PROCHOT Offset value to 8 and they have also locked this setting so it cannot be adjusted. This tells the CPU to thermal throttle 8°C before it reaches 100°C. That is what PROCHOT 92°C on the main screen of ThrottleStop is telling you.

The temperatures that NitroSense are reporting are not accurate. NitroSense is totally missing the peak core temperatures. ThrottleStop uses a few programming tricks to make sure that it can report the peak temperature correctly. You can trust ThrottleStop. Its monitoring abilities are second to none.

Your undervolt settings look OK. -135 mV for the cache is more than most 9750H can run reliably at. Many 9750H start to lose stability at an offset of -120 mV to -125 mV for the cache. When you run the TS Bench test and it is set to 960M, does this test report any errors? If you see even 1 error, I would reduce the cache offset until there are no errors reported.

GPU is hot!

Have you ever disassembled your laptop to clean it out and to replace the thermal paste? That might help. It is possible that the heatsink that Acer is using is barely adequate and not capable of keeping the Nvidia GPU at a reasonable temperature. Problems like this are common in many laptops.

 

[WickedSick277]

The temperatures that NitroSense are reporting are not accurate. NitroSense is totally missing the peak core temperatures. ThrottleStop uses a few programming tricks to make sure that it can report the peak temperature correctly. You can trust ThrottleStop. Its monitoring abilities are second to none.

Hey man, thanks for your reply and devoted time.

In the previous post i was gaming(with low CPU usage) and i double-checked Nitrosense/GPU-Z, they were reporting somewhat of 70°C while TS reported 90°C. though it doesn't really matter though since i still get -5°C to -10°C by undervolting

TS Bench results:
'Threads 12, Priority normal, MHZ fixed - Balanced 230 EPP, Gaming 50 EPP'
On Battery + Balanced plan: 120M: 26s - 960m: 141.8s 50°C
On Battery + Gaming plan: 120M: 16.4s - 960m: 130.8s 70°C, POWER REACHED 25 (Screenshot below)
On DC + Gaming plan: 120m:12.8 - 960m:109.6, PROCHOT REACHED, HOT reached, POWER REACHED 45 (Screenshot below)
On DC + Balanced plan: 120m:30s - 960m:210s 57°C

*:The only errors i had: 2 errors on On Battery + Gaming plan, which i think because of using a Game plan on battery, which i usually never put heavy pressure on machine while on battery.
Should i be worried about this one or leave my undervolt settings be as they are?
**: I see 40°C of cooling down within only 10 seconds after completion the bench! screenshots are below, is that normal?

Have you ever disassembled your laptop to clean it out and to replace the thermal paste? That might help. It is possible that the heatsink that Acer is using is barely adequate and not capable of keeping the Nvidia GPU at a reasonable temperature. Problems like this are common in many laptops.

I thought about it, i had several issues and think it's risky (have to do it on my own).
but if you guys think i 100% need re-pasting and cleaning the dust, i would accept the risk for opening it
i got 57°C CPU, and 52°C GPU on idle.
Thanks in advance!

 

[unclewebb]

i double-checked Nitrosense/GPU-Z, they were reporting somewhat of 70°C while TS reported 90°C

As I said before, trust ThrottleStop for CPU temperature data. ThrottleStop runs at a higher Windows priority so when it queries the CPU for temperature data, the CPU responds immediately. Other monitoring programs have to get in line after ThrottleStop. If the CPU is busy, these other programs might not be getting data that is as accurate as the information that ThrottleStop is getting from the CPU.

I would avoid running the TS Bench or any full load stress testing program when running on battery power. Performance oriented gaming laptops need to be plugged in for maximum performance. Full load stress testing is not good for the long term health of batteries.

I see 40°C of cooling down within only 10 seconds

The instantaneous change in CPU core temperatures seems to surprise a lot of people. I once put my thumb on the bare die of an old Pentium single core CPU, pressed the On button and within seconds, my stupidity was rewarded with a burnt thumb. When pumping 45W through something the size of your small finger nail, it is going to get hot very quickly. As soon as you stop a stress test and reduce the load, the opposite will happen. The CPU cores will cool down almost as quickly. The better the cooler, the faster it will be able to cool the CPU down.

Should i be worried about this one

A TS Bench error is basically the CPU doing the same calculation twice and getting a different result.

2 + 3 = 5
2 + 3 = 6

I want my computer to be 100% stable so even 1 error while running the TS Bench test is too much for me. It is up to you. During normal use when not stress testing, your computer and your undervolt are probably going to be 100% stable as is.

The TS Bench is doing millions of calculations so 1 error in a million calculations might be an acceptable amount of risk for a laptop used mostly for gaming. If you are running a nuclear power plant off of your laptop, better go for zero errors. If you ever have a problem or a blue screen or Windows gets corrupted, give the CPU cache some more voltage.

During stress testing, you have the first core at 92°C. This is causing all of the cores to thermal throttle while the last core in your screenshot is only at 73°C. A temperature variation of 19°C when cores are equally loaded is a pretty good sign that the heatsink is not making even contact with all of the cores. Someone might have done a poor job installing the heatsink where they tightened down one side all of the way before tightening down the other side. It might mean the heatsink is not flat or it might be bent or it might be a sign that the thermal paste was not applied correctly. This temperature variation and the high GPU temperatures are both telling you that your laptop needs some maintenance inside. Watch some YouTube videos. Many recent laptops are fairly easy to disassemble. If you are happy with your laptop as is and cannot risk damaging it, go play a game and leave this cleaning procedure for another day.

 

[WickedSick277]

As I said before, trust ThrottleStop for CPU temperature data. ThrottleStop runs at a higher Windows priority so when it queries the CPU for temperature data, the CPU responds immediately. Other monitoring programs have to get in line after ThrottleStop. If the CPU is busy, these other programs might not be getting data that is as accurate as the information that ThrottleStop is getting from the CPU.

I would avoid running the TS Bench or any full load stress testing program when running on battery power. Performance oriented gaming laptops need to be plugged in for maximum performance. Full load stress testing is not good for the long term health of batteries.


The instantaneous change in CPU core temperatures seems to surprise a lot of people. I once put my thumb on the bare die of an old Pentium single core CPU, pressed the On button and within seconds, my stupidity was rewarded with a burnt thumb. When pumping 45W through something the size of your small finger nail, it is going to get hot very quickly. As soon as you stop a stress test and reduce the load, the opposite will happen. The CPU cores will cool down almost as quickly. The better the cooler, the faster it will be able to cool the CPU down.


A TS Bench error is basically the CPU doing the same calculation twice and getting a different result.

2 + 3 = 5
2 + 3 = 6

I want my computer to be 100% stable so even 1 error while running the TS Bench test is too much for me. It is up to you. During normal use when not stress testing, your computer and your undervolt are probably going to be 100% stable as is.

The TS Bench is doing millions of calculations so 1 error in a million calculations might be an acceptable amount of risk for a laptop used mostly for gaming. If you are running a nuclear power plant off of your laptop, better go for zero errors. If you ever have a problem or a blue screen or Windows gets corrupted, give the CPU cache some more voltage.

During stress testing, you have the first core at 92°C. This is causing all of the cores to thermal throttle while the last core in your screenshot is only at 73°C. A temperature variation of 19°C when cores are equally loaded is a pretty good sign that the heatsink is not making even contact with all of the cores. Someone might have done a poor job installing the heatsink where they tightened down one side all of the way before tightening down the other side. It might mean the heatsink is not flat or it might be bent or it might be a sign that the thermal paste was not applied correctly. This temperature variation and the high GPU temperatures are both telling you that your laptop needs some maintenance inside. Watch some YouTube videos. Many recent laptops are fairly easy to disassemble. If you are happy with your laptop as is and cannot risk damaging it, go play a game and leave this cleaning procedure for another day.

Thank you. learnt a few useful things from your reply.
I actually did another test for double-checking everything and yes, there was at least 10-15°C difference between cores.
So now i concluded they had a good thermal paste (which can reduce ~40°C in 10 second) but pasted in a bad shape (difference between cores). please correct me if i'm wrong.
Seems like i really need to re-paste and accept the risk.

*:When the prochot turns red, my CPU clock goes down all the way from 4.1GHZ to like 3.4GHZ. that means thermal throttling did that, right?
(I even tried benching with no undervolting and EPP=0, the results weren't that different from undervolting due to this core speed fall)

*: I can't reach the promised 4.5 GHZ of 9750h. even with no red Power limit sign in TS. That's due to OEM assembly\limitations, or i did something wrong in my OS?

Thanks in advance.

 

[unclewebb]

i concluded

I would not conclude anything until you disassemble your laptop to have a good look at it. Have a straight edge handy so you can try to determine if your heatsink is flat or not. Sometimes it is obvious and you might see where the thermal paste is poorly distributed.

When the prochot turns red

PROCHOT stands for processor hot. When your computer reaches a peak core temperature of 92°C, the CPU will slow down all of the cores as much as necessary to help prevent the CPU from going over 92°C. Intel thermal throttling works fantastic. The CPU speed is constantly adjusted hundreds of times per second to keep the CPU at a safe temperature. ThrottleStop uses high performance monitoring timers within the CPU which allows it to report a very accurate average speed when thermal throttling is in progress. Many popular monitoring apps choke and are not very accurate when thermal throttling starts.

I can't reach the promised 4.5 GHz

Welcome to the club. The maximum 45 multiplier is only available when a single core is active. Most computers running Windows have hundreds of tasks running in the background all of the time. These tasks are constantly waking up additional cores. That means the 1 core active multiplier is rarely if ever used on most computers. You would need a very lean Windows install with virtually nothing running in the background before you might see evidence of the 45 multiplier. This is not practical.

Start by using the default turbo ratios, 45, 44, 43, 42 for the first four turbo ratios in the FIVR window. You do not have a heating problem when only 1, 2, 3 or 4 cores are active so there is no need to lower these values.

What does ThrottleStop report for C0% when your computer is idle at the desktop?

C0% is an accurate way to measure how much crap is running in the background on your computer. You need C0% to be as low as possible when your computer is idle at the desktop before you will ever see max speed. Try running a 1 Thread TS Bench test and watch for the maximum multiplier that ThrottleStop reports while this test is running.

 

[WickedSick277]

Welcome to the club. The maximum 45 multiplier is only available when a single core is active. Most computers running Windows have hundreds of tasks running in the background all of the time. These tasks are constantly waking up additional cores. That means the 1 core active multiplier is rarely if ever used on most computers. You would need a very lean Windows install with virtually nothing running in the background before you might see evidence of the 45 multiplier. This is not practical.

Start by using the default turbo ratios, 45, 44, 43, 42 for the first four turbo ratios in the FIVR window. You do not have a heating problem when only 1, 2, 3 or 4 cores are active so there is no need to lower these values.

Thank you.
I checked with TS bench on 1 core, and on the mentioned turbo ratios. clock went up to 4.4 GHZ but mostly between 4.2 and 4.35.

What does ThrottleStop report for C0% when your computer is idle at the desktop?

View attachment 261624

C0% is an accurate way to measure how much crap is running in the background on your computer. You need C0% to be as low as possible when your computer is idle at the desktop before you will ever see max speed. Try running a 1 Thread TS Bench test and watch for the maximum multiplier that ThrottleStop reports while this test is running.

C0% is usually 0.6 to 0.5% in the same section you screenshotted.
C0% for the first core playing between 0.5 to 3%, mostly below 1%.
for other cores it is constantly changing, which means all cores constantly waking up. i don't know exactly that's normal or not (i Googled and found it's normal)

*: there are several Thermal YouTube videos, I've seen some methods.
All of them accepted not to tighten up the screws all the way, and also use a cross closing method (like an X, in this video)
But which way you guys actually prefer, dot-method or the line-method?

 

[unclewebb]

Those thermal paste blobs in the video look very generous. I think half that is going to squirt out as soon as the heatsink is put into position and tightened down.

Different pastes recommend different methods for applying them. Read the manufacturer's instructions for whatever paste you are using. For laptop CPUs that do not use a heat spreader over top of the cores, I prefer to spread a thin layer over the entire CPU die. Try this method, try the line or the blob method and find out what works best for your computer. The right way is the one that gives you the best long term results. You might not see a huge difference between any of these methods.

Watch out for thermal paste pump out. This can happen after as little as a week or two.

Keep an eye on your temperatures after you re-paste. They should be reasonably consistent over time.

Cores constantly waking up to process background tasks is normal. Recent Intel CPUs rarely use the highest 1 core active multiplier. This is interesting because I have an older 4th Gen CPU in my laptop and it has no problem using the 1 core active multiplier more than 90% of the time during a 1 Thread TS Bench test. Intel must have changed their algorithm because now the 1 core active multiplier is rarely if ever used. A 9750H with a 45 multiplier makes the Intel marketing team happy but it is useless for consumers as it is never used.

 

[WickedSick277]

Those thermal paste blobs in the video look very generous. I think half that is going to squirt out as soon as the heatsink is put into position and tightened down.

Different pastes recommend different methods for applying them. Read the manufacturer's instructions for whatever paste you are using. For laptop CPUs that do not use a heat spreader over top of the cores, I prefer to spread a thin layer over the entire CPU die. Try this method, try the line or the blob method and find out what works best for your computer. The right way is the one that gives you the best long term results. You might not see a huge difference between any of these methods.

Watch out for thermal paste pump out. This can happen after as little as a week or two.

Keep an eye on your temperatures after you re-paste. They should be reasonably consistent over time.

Cores constantly waking up to process background tasks is normal. Recent Intel CPUs rarely use the highest 1 core active multiplier. This is interesting because I have an older 4th Gen CPU in my laptop and it has no problem using the 1 core active multiplier more than 90% of the time during a 1 Thread TS Bench test. Intel must have changed their algorithm because now the 1 core active multiplier is rarely if ever used. A 9750H with a 45 multiplier makes the Intel marketing team happy but it is useless for consumers as it is never used.

I am working around re-pasting. i wil contact you guys back whenever i decided.
Maybe a dumb question but:
Are you sure my undervolt setting looks Okay? and it is turned on?
I know you said in the first post, but i read somewhere that iGPU and intel GPU should has the same amount in order to TS to work..or something like this.
I had better temps but today i've been playing with undervolting and CPU reached 95 on core #1 and 80 on core #6.
GPU was 88 tho.

 

[unclewebb]

iGPU and intel GPU should has the same amount in order to TS to work.

To undervolt the Intel GPU, both the Intel GPU and the iGPU Unslice need to be undervolted, likely by the same amount. This may not be true on every CPU that Intel has ever made but it is always a good place to start testing. If you have a Nvidia GPU, I do not bother undervolting the Intel GPU. The Intel GPU is rarely significantly loaded in this situation so undervolting the Intel GPU is probably not going to reduce power consumption or temperatures any significant amount.

The Intel CPU undervolt will still work whether you decide to undervolt the Intel GPU or not.

and it is turned on?

Look in the FIVR monitoring table in the top right corner. If you see your offset voltages being reported in the Offset column of that table then that confirms that your undervolt is working.

CPU reached 95 on core #1

Your main problem is likely your thermal paste or a heatsink that is not flat or a heatsink that is not making even contact with the cores. Until you investigate that problem, the sky high temperatures at one end of your CPU are not going to get better. They might be getting worse.