Folding Forum

This is my 3rd work unit but I noticed that for this work unit, my core temp reaches 90 C. What is going on? How can I reduce my core temp?

You can open the case of your PC and/or otherwise improve airflow.
And you can lower the # of threads used by Folding - in Advanced Control -> Configure -> Slots -> click cpu then Edit, read the descriptions, enter the number of threads you want to use, OK, Save, done.

F@h does work the hardware pretty hard. There are the basics, cleaning out any build up of dust and lint, improving air flow over the CPU cooler, etc.

From the client side you can manually set the number of CPU cores used. If you have an Intel processor with HT, using just the main cores can reduce the temperature for instance.

To do so, using the advanced control, FAHControl, open the Configure function. Select Slots, and click on the one named CPU and Edit down in the bottom right of the window. You set the number of CPU threads by changing the CPU thread setting from the default of '-1' which leaves control to the software and the slider position. The number can be from 1 to the total number of CPU cores/threads available. If you have a GPU folding slot, that requires one thread for each GPU.

Not knowing which CPU you have, suggestions for the number to set are multiples of 2, 3, and sometimes 5. Once you have set a number, OK the change and Save it. The client will pause the processing and restart at the new thread setting.

thanks... its doing better now

Hi,
I had similar issue on my gamer laptop (core i7, 7th gen) few weeks ago, when I started to fold. Actually, I had to reduce the 4 cpu threads to 2 threads in order to keep cpu from melting.
( and with 2 threads it was still 75-80 C on a cooling pad.) (then I got a GPU work unit too but then removed the gpu slot to cancel that.)

So, then I realized that folding (sort of) requires a desktop pc with good airflow. I installed it to my workstation, and it can handle that.
(Evga SR2, 2x Xeon x5670 (95W tdp) with "Arctic cooling freezer xtreme" huge cpu coolers, and dust filters on the case, and good airflow. Now, even with some overclock, cpu is 62C only with most of cores running. Without overclock, it was 55C at full load. )
these two cpus and 23 threads give 126k ppd at 3.66GHz overclock and it was running overnight, when no CPU WU then I run Rosetta@home.

Also, GPU needs good cooling (big cooler, 2 big fans on gpu, the previous one was running way hotter, but the current one is never hotter than 65C)

To choose the number of CPU Treads to run:

In the taskbar to the lower right of the screen, you should see a F@H molecule icon, click it (you may need to click an Up Arrow to see it ^)

The second item in this menu is Advanced Control, click it

On this screen to the left is a Configure button, click it

Now you get a screen with a Slots tab, click it

On this white field should be a cpu item, click it and then click edit

By default F@H set the number of CPUs to -1 meaning let the software decide.
You can enter any number from 1 to the number of threads you CPU supports.
If you have GPUs, F@H reserves one CPU per GPU to feed it data across the PCIE bus.
F@H has difficulty with large primes and their multiples number of CPUs.
7 is always large, 5 is sometimes large, and 3 is never large. Try to choose a number that is a multiple of 2 and/or 3.
2, 3, 4, 6, 8, 9, 12, 16, 18, 24, 27, etc. are good numbers of CPUs to choose.
5. 10. 15, 20 etc may work most of the time. Other numbers will bite you.

Choose the number of threads you wish to run and click save.

Having said that, the new 7.51 client tries to automatically shift down to a safe number.

JimboPalmer wrote:To choose the number of CPU Treads to run:

In the taskbar to the lower right of the screen, you should see a F@H molecule icon, click it (you may need to click an Up Arrow to see it ^)

The second item in this menu is Advanced Control, click it

On this screen to the left is a Configure button, click it

Now you get a screen with a Slots tab, click it

On this white field should be a cpu item, click it and then click edit

By default F@H set the number of CPUs to -1 meaning let the software decide.
You can enter any number from 1 to the number of threads you CPU supports.
If you have GPUs, F@H reserves one CPU per GPU to feed it data across the PCIE bus.
F@H has difficulty with large primes and their multiples number of CPUs.
7 is always large, 5 is sometimes large, and 3 is never large. Try to choose a number that is a multiple of 2 and/or 3.
2, 3, 4, 6, 8, 9, 12, 16, 18, 24, 27, etc. are good numbers of CPUs to choose.
5. 10. 15, 20 etc may work most of the time. Other numbers will bite you.

Choose the number of threads you wish to run and click save.

Having said that, the new 7.51 client tries to automatically shift down to a safe number.

On mine it says:
CPU
A CPU slot uses 1 or more CPU's. More CPU's on one slot can complete simulation trajectories faster than many single CPU slots.
CPU's - The number of CPU threads this slot should use. -1 lets the client choose.
-1

I have an i7 7700K that gets very high temperatures when folding (near 90 degrees). I'd prefer to keep it down to below 75, do you know about how many threads I should set it to by any chance (best guess)?

I don't want to cook the cpu, but I also don't want to go so slowly that I may as well not bother.

Thanks for any help you can give

I personally would try 4 so the HT core threads aren't being used. See if that is enough, otherwise 3 or 2 should reduce the temperature.

I have an i7-7700k in a PC that I assembled to be a Hackintosh, and also boot up into Linux. I have run into the same issue, there are reports that Intel did a poor job of packaging many of the chips in that series, and there is not good thermal transfer between the chip and its "lid". That said, I have run my machine for days at a time without running into problems, the thermal management does a good job of keeping the temperature within limits.

Some people have fixed the problem by de-lidding their CPU, applying new thermal transfer paste, and replacing the lid. I may do that at some point in the future, but tearing apart the machine to get the CPU out is for another day.

Joe_H wrote:I personally would try 4 so the HT core threads aren't being used. See if that is enough, otherwise 3 or 2 should reduce the temperature.

I have an i7-7700k in a PC that I assembled to be a Hackintosh, and also boot up into Linux. I have run into the same issue, there are reports that Intel did a poor job of packaging many of the chips in that series, and there is not good thermal transfer between the chip and its "lid". That said, I have run my machine for days at a time without running into problems, the thermal management does a good job of keeping the temperature within limits.

Some people have fixed the problem by de-lidding their CPU, applying new thermal transfer paste, and replacing the lid. I may do that at some point in the future, but tearing apart the machine to get the CPU out is for another day.

Thank you so much for the reply. Whilst servers are down I'll progress some of my rosetta jobs, and then switch back to fah.

I managed to tear apart and rebuild my 2013 laptop which is just doing rosseta and folding, but I'm certainly no where near experienced enough to attempt a delidding.

Oh I got a work assignment, I have gone to 75 degrees, perfect.

If I just run this computer on >80 degrees cpu and around 85 degrees gpu (1080) then that shouldn't cause any long term problems should it (looking for guidance, I know there are no guarntuees and no-one is liable for their response not aligning with reality!)?

vegancookies wrote: If I just run this computer on >80 degrees cpu and around 85 degrees gpu (1080) then that shouldn't cause any long term problems should it?

That looks good. If your CPU or GPU reaches near 100°C then it will throttle and then shutdown. I would not let CPU or GPU run hotter than 90°C.

Although the necessary level of experience with your machine would be greater, I'm surprised no one mentioned setting your UEFI (which has replaced what we used to think of as the computer's BIOS... accessible by special key combination when you're booting a PC, before the OS starts to load and run) for slightly safer values of several parameters.

The simplest change would be just to restrict the maximum processor speed; try 100MHz lower than you were running before, and voilà, at full 100% load the CPU sensors should show a few degrees C lower.

You can also make a similar restriction right in Windows' power settings -- there's a setting for the maximum loading Windows will allow on each core. That's more about cycling work on and off of them rather than clock frequency so it's not as direct of an approach to a temperature problem but it should be sufficient to get things under thermal control easily enough.

For those willing to go further in their UEFI, particularly if they may have forgotten that they left very optimistic overclock settings in place due to years of gaming that likely did not nearly full load the CPU for long stretches of time... if you really want to fold and are willing to sacrifice a bit of game smoothness and frame rate during the current pandemic, this would be a good time to return to stock voltage or only slightly boosted voltage.

To give a concrete example from my own 8-core i7-9900X system, I returned to auto voltage, temporarily removed most of my overclock related settings, undid my per-core multipliers and just told the thing to run at the max all-cores speed that I know keeps the rig under 85C at stock voltage all day. This means I won't get 4.5GHz performance on single-threaded tasks the way I like to in Photoshop etc, or around 4.3 for several cores and multithreaded software, I'm just running the whole machine at a lowly 3.8. Even a manufacturer stock overclock of 3.9 was still hitting 92C but 3.8 and optimized-default mfr voltage regulator settings stays below 85C. I can be glad it's still well above the CPU list base speed of 3 (and listed turbo to 3.6). Anything to help out the project is fine, I can get back to being excited about overclocking at some future date.

For owners of similar rigs, if my quoted temperatures seem high, it's true that I need to perform a five year internal deep clean soon and that should be worth a few °C too.

Automatic CPU core voltage can sometimes become a problem. My last two Ryzen CPUs (1800X and 3900X) have both been using a much higher automatic voltage than a stress-tested stable manual one is. I guess that is due to the manufacturers adding some headroom to ensure that the system will be stable no matter what.

heated_snail wrote:The simplest change would be just to restrict the maximum processor speed; try 100MHz lower than you were running before, and voilà, at full 100% load the CPU sensors should show a few degrees C lower.

You can also make a similar restriction right in Windows' power settings -- there's a setting for the maximum loading Windows will allow on each core.

I kinda did just this, I set 'maximum processor state' in the power options to 99%, thus disabling Turbo. I have a i7 3930k with stock cooling, not OC'd. My rig is a Frankenstein, parts largely scavenged or gifts from friends, so I have to... make decisions that'd leave others scratching their heads, let's put it like this. Anyway, 12 threads @ 3.8 GHz with Turbo = 90+ C. Without it, it's 3.2 GHz at much less concerning temps of 60-65 C. I should also note that when idling, temps sit at ~30 C.

I'm not sure how much time is lost this way? 600 MHz sounds very significant to me. But on the other hand, I imagine increasing the lifespan of my components = more folding in the long run!

For a different perspective:

I run an enterprise grade server (2x Xeon E5-2697 v3 @2.60GHz 512GB) configured for performance CPU loads, which pretty much self manages and it only spins its (5x) fans up as it needs to, to cool the various components (FRU/SDR/BMC setup correctly for performance).

It is really noticeable that the server balances its fans to ensure that the core temps are kept a few degrees below Intel TCC/Tjmax values (for these CPUs that is 93C) and so I never see max values over time above 90C ... This is allowing the server to throttle up the cores some 15% even under full load (2x27/28core WUs) and still maintain an acceptable (although it may seem hot to some people) temp ... At this temperature the fans are only run up to just below 50% - the server could spin these higher and keep temperatures lower but it doesn't need to.

BTW I also have a workstation laptop (Xeon E3-1505M @ 2.8GHz) which runs in much the same manner (TCC/Tjmax for its CPU is 100C) and it spins its fans up and keeps a maximum of less than 93C.

Caveat: I will only run the server and laptop at these temperatures on a sustained basis (oh for a constant stream of "14 day" WUs for both slots (even thought they take server less than 4 hrs to process) - I wouldn't want to be heating them up to this level and cooling down "quickly" (spikey loading ramping temperatures up/down every few seconds/minutes) - A discussion many years ago with an HPC builder with regards to "heat cycling" being far worse for kit than sustained high temperatures.

I am lucky in that I know my kit very well (having specified components and configuration) ... Most modern kit will self manage temperatures (if setup correctly) but it is worth being cautious if you have any doubts at all - Laptops will in general run hotter and struggle more with cooling especially under 100% CPU and GPU loads (their cooling pipes are in general not designed/configured to cope with this type of work - mainly aimed at supporting light browsing and business loads) but some of the gaming /workstation laptops cope better.

Take care of your kit - although it sounds like you are doing so