Page 3 of 3

Re: Issue with CPU usage and overheating

Posted: Wed Oct 30, 2013 8:47 pm
by Joe_H
If you can tolerate the running two cores with the CPU slider set at 100% when not using Excel and other apps, then I would recommend that. Running this project on a single core is not an option, as an A3 WU it requires SMP processing on at least 2 cores. With two cores in use, that still leaves two virtual cores through HT for other apps.

Re: Issue with CPU usage and overheating

Posted: Wed Oct 30, 2013 8:48 pm
by bruce
I have experience with BOINC (and can tell you why I no longer run it ;) but that's not your question).

BOINC supports a variety of projects and you'll find very significant differences between them. FAH basically supports a variety of projects but they're all related to protein folding. By far the most important difference, from your perspective, is deadlines.

The original BOINC project was seti@home so I'll use it as an example. Data collected from radio-telescope signals was gathered and distributed for analysis. Each segment of the sky was independent of some other segment so the order that donors returned results was meaningless. Nobody really cared when a WU was returned as long as everything was analyzed. Deadlines were assigned but were pretty much meaningless except as a means to uncover lost WUs. The data analyzed by most BOINC projects share this characteristic.

For FAH projects, deadlines are critical. Each WU is a member of a time-series and future times cannot be analyzed until the previous time segment has been analyzed. Delaying any WU actually delays all of the WUs that follow in that same series. We're not analyzing pre-recorded data, we're predicting the future direction that the protein will take. While there are enough different series of WUs that everybody can be working on a different series, the particular WU you're assigned has a unique relationship with every WU that follows it.

Most BOINC WUs are 100% parallel (except for a couple of projects). FAH WUs are both serial and parallel, and it's the serial nature that forces FAH's deadlines to become critical. Bottom line: FAH's math is different than BOINC math.

Re: Issue with CPU usage and overheating

Posted: Wed Oct 30, 2013 10:12 pm
by Ivkosky
Joe_H wrote:If you can tolerate the running two cores with the CPU slider set at 100% when not using Excel and other apps, then I would recommend that. Running this project on a single core is not an option, as an A3 WU it requires SMP processing on at least 2 cores. With two cores in use, that still leaves two virtual cores through HT for other apps.
I am afraid this is not feasible. I have tried it now for a couple of minutes with all the other applications closed (except antivirus), having just 65 processes together with FAH turned on (which I think is pretty good for a Windows 7 laptop...), and the CPU temperature jumped from 75C to 83C. The motherboard temp is still 80C, but I guess this is only thanks to vigorous spinning of the ventilator, which I guess is on its max now. I just wonder, why the manufacturers even offer laptops with such a strong configuration, if users cannot use even a half of it?! Ridiculous! But, that is again off topic...

As I cannot afford to risk a damage to this laptop now (it is my work laptop as well), I am afraid I would have to back off... I really wanted to help and I am most grateful for all your help and time spent on replying to my questions. Thanks.

P.S.: I would be still grateful if bruce and 7im can share their views on BOINC - please PM me.

Re: Issue with CPU usage and overheating

Posted: Thu Oct 31, 2013 6:56 am
by bruce
True story:
I once purchased a name-brand desktop computer. It was a long time ago and probably contained an AMD dual-core CPU. I started FAH and it ran a few minutes and shut down. I repeated this scenario and then called the new-computer-support-desk. After several calls it became obvious that the support technician had no idea what was wrong. (Neither of us measured the temperature.) The model was new enough that the technician had never seen one. After many fruitless calls, I returned the computer to the store and got a different model which worked faithfully for many years.

From what I know today, I'm certain that the thermal compound was poorly applied and/or there were other cooling issues. Maybe the thermal design (case / air path / heat sink design / fan size / fan setting, etc.) had never been qualified.

I'm not suggesting that you should ignore the issue and intentionally overheat it -- but do look at it from the perspective of a reputable computer manufacturer. They are responsible for designing the cooling system to operate properly in the hottest possible configuration including a scorching desert running the heaviest processing that sofware can deliver. If it overheats under those conditions, they will have to replace it under warranty and they certainly do not want to do that.

If a computer is assembled by your neighbor, it probably was NOT tested for overheating under the ultimate worst conditions (or even if somebody just adds a GPU) so you can't depend on this always be true, but since you're talking about a laptop, that doesn't apply. A little judgment should apply here, though. Nobody will promise how much noise the fan will make -- only that the hardware shouldn't be able to damage itself, and if it does, the warranty should cover it, even it it's a problem with thermal grease -- and I would have that checked before giving up.

I respect your choice to limit the fan noise, though, so do whatever you think best.

Re: Issue with CPU usage and overheating

Posted: Fri Nov 01, 2013 11:32 am
by Ivkosky
Bruce, thanks for your response. I am, however, not concerned about the noise of the fan. What I am concerned about is its lifespan and potential burn down of the CPU/motherboard... As I wrote before, I had Dell XPS 13" a couple of years ago. I ran BOINC on it and didn't care about the temperatures too much. The fan was on 100% all the time, which resulted in a physical damage of the fan. This obviously made my laptop unusable for a couple of days. I cannot afford that to happen now, as it is my only laptop and, I am sorry, but I really don't intend to buy a separate PC just for FAH. :) It even wouldn't be possible, because where I live at the moment, there is no option for me to connect another PC to Internet...

As suggested by Joe_h in his post above, I have been running FAH on two cores now, with a slider for the CPU at 100%. It takes 50% of my overall CPU, but increases the temperatures for the CPU to 87C and motherboard to 90C. Is this sustainable? It obviously increases the performance and I am now easily able to finish off the WU on time...

Re: Issue with CPU usage and overheating

Posted: Fri Nov 01, 2013 2:50 pm
by bruce
I do know what you're talking about. I'd call that a Dell "quality" issue.

Many years ago, we bought a bunch of cheap PCs for our office. For some reason, the fans ran 24x7. After a few years, the fan lubrication failed and started killing computers. Fans with ball-bearings can last forever. Fans with sleeve-bearings are cheap and are designed to last until the day after the warranty expires. ;)

Re: Issue with CPU usage and overheating

Posted: Sat Nov 02, 2013 6:49 pm
by PantherX
Ivkosky wrote:...As suggested by Joe_h in his post above, I have been running FAH on two cores now, with a slider for the CPU at 100%. It takes 50% of my overall CPU, but increases the temperatures for the CPU to 87C and motherboard to 90C. Is this sustainable? It obviously increases the performance and I am now easily able to finish off the WU on time...
Generally speaking, laptop parts have a higher temperature endurance rating then their desktop counterparts. However, your definition of high temperature could be different from mine. I would be fine with those temperatures. Also, what is your ambient temperature? Mine is 25C - 35C and my laptop (Sager) operates from 75C to 90C folding on the CPU and GPU 24/7 for almost a year without any issues. I do clean my fans on a regular basis and have removed the rear panel and elevated it with bottle caps. My previous laptop (ASUS) ran F@H 24/7 on the CPU and GPU with temperatures from 75C to 100C without any issues for around 4 years before I retired it.

P.S. -> Welcome to the F@H Forum Ivkosky, just realized that you weren't greeted so better late than never

Re: Issue with CPU usage and overheating

Posted: Mon Nov 04, 2013 7:23 am
by Ivkosky
Hi PantherX

Thanks for your greeting - I am happy to be a part of the community! :)

My ambient temperature is almost always 23C (indoors, air conditioned). I do have the back of the laptop elevated (with my wallet for now :D ), but I am afraid it would not be practical for me to remove the rear panel, as I travel with it almost every day. My GPU (Intel) is not supported, so I am folding only on CPU.

I have been testing various settings these days and it seems that the best way for me would be to use two cores, but have the slider for CPU usage on 80%. I am not sure what this setting actually does in terms of the usage of the two cores (one of the earlier replies suggests that not having the slider on 100% means that one thread always has to wait for the other to sync, but I am not sure what this really means...). Anyway, by doing this, I am still able to conveniently meet the deadlines while the CPU temperature dropped from 87C to 83C and the motherboard temperature from 90C to 80C comparing to having the slider on 100%! Also, the fan is not on its 100% all the time, which is another good sign, I guess.

I have read somewhere that the "meltdown" temperature for Dell Vostro is around 100C. I would therefore like to have some buffer (as other applications can drive the temperatures up as well, e.g. Excel) and I think running FAH on 90C would be right on the edge... I am therefore much more comfortable with 80-85C.

Re: Issue with CPU usage and overheating

Posted: Mon Nov 04, 2013 3:47 pm
by PantherX
Regarding the slider, if you see FahCore_a* using 50% of your CPU in the Task Manager, then it is safe to say that the Slider isn't having any effect on your system. I would personally remove it since it doesn't work with the newer FahCores. The temperature variation of 4C is marginal and can be related to a possible difference in the ambient temperature, different WU (some WUs can be computationally intensive than others), etc.

Regarding the "temperature buffer", the hardware has it built-in. Once it reaches a critical temperature, the first thing it does is to downclock itself which can drastically reduce the temperature (on my laptop, an automatic downclock from 3.6 to 2.8 can reduce the temperature by 3C - 10C, depending on the ambient temperature) while keeping the same CPU Usage. The downclocking occurs in steps and once it reaches the lowest possible value and the temperature is still critical, it will automatically shut itself to prevent permanent hardware damage. Moreover, the heat generating part of the CPU is mostly the FPU (Floating Point Unit) which F@H stresses out than most applications. Unless you run intensive functions on a very large Excel Worksheet, I wouldn't worry about additional load from Excel.

Re: Issue with CPU usage and overheating

Posted: Mon Nov 04, 2013 4:15 pm
by Ivkosky
PantherX, thanks for your reply. I indeed see a drop in the CPU load from 50% to 45% when I moved the slider from 100% to 80%. The drop in temperature was only 5C for CPU, but as much as 10C for the motherboard. Anyway, I will test it with various WUs as they come and I will see. In any case, I will definitely try to stick with the FAH! :)

Not related to this topic, but just a quick question: While playing with the settings, I have accidentally deleted what was supposed to be written in the "Command" box in the Preferences popup window. Could you please tell me what to write there? :lol: Thanks!

Re: Issue with CPU usage and overheating

Posted: Mon Nov 04, 2013 4:27 pm
by bollix47
Preferences Command box: FAHViewer

Re: Issue with CPU usage and overheating

Posted: Mon Nov 04, 2013 5:00 pm
by Joe_H
Ivkosky wrote: have been testing various settings these days and it seems that the best way for me would be to use two cores, but have the slider for CPU usage on 80%. I am not sure what this setting actually does in terms of the usage of the two cores (one of the earlier replies suggests that not having the slider on 100% means that one thread always has to wait for the other to sync, but I am not sure what this really means...).
The CPU slider is still affecting processing on one core, the thread on the other core is having to wait. In effect you will be using somewhere between 80-90% of the two cores capacity which is not as bad as when you had the slider set way down. I do not have access to the code at the level needed to determine the exact Wait state the second core would be in, but from your observation it is one that uses less power. It does disrupt the SMP processing in a non-linear manner, but not as badly as you would see in higher core counts where multiple threads would be waiting to synch with the slow one.

Re: Issue with CPU usage and overheating

Posted: Wed Nov 06, 2013 3:10 pm
by codysluder
I agree that you should not need to become an IT expert to Fold (or BOINC) but you would need to be satisfied with the default settings. Stanford (and probably Berkeley) establishes conservative defaults. They may not give you the maximum number of points but they should give you a reasonably stable system that produces a reasonable amount of work.

The moment that somebody shows up on the forum saying (in effect) what settings can I change to get more points, they need to understand that they're claiming to be an expert and they'll understand that changing any setting will involve trade-offs. They need to be enough of an expert to understand all of the possible eventualities to reconfiguring their system.

Re: Issue with CPU usage and overheating

Posted: Wed Nov 06, 2013 6:01 pm
by 7im
There is a bug ticket to remove the CPU usage slider until the fahcores are updated to where that feature is well supported.