Trying to pinpoint the heat problems

Case number:699969-997291
Opened by:Timo van der Laan
Opened on:Friday, March 21, 2014 - 00:44
Last modified:Tuesday, March 25, 2014 - 08:04

I just started 3 clients on my laptop, just like I used to do. This laptop has no problems, not with capacity nor with heat. So maybe it is the kind of cpu that has this problem.
So my question to all:
What cpu are you using and do you have heat problems?
Mine on this laptop: AMD A8 3500M, temperature 71%, ventilator at 89% according to the build in toshiba health monitor. Running 3 clients, 1 is design.

(Fri, 03/21/2014 - 00:44  |  14 comments)

Joined: 09/21/2011
Groups: Void Crushers

BTW, if you dont want your info to be public, you can sent me a PM with the info.

spmm's picture
User offline. Last seen 18 weeks 5 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

Is that a new computer, what OS is it running, 32 or 64 bit, how much RAM is it pushing and what type? What type of storage? Foldit seems to do a fair bit of disk I/O which may be the kiss of death for an SSD. Plus the remote server calls, which are frequent. Perhaps why so many people ask to play offline. What is the bus speed? which bus? Is anything being pushed to the GPU? With or without sockets, its an endless tale once we get to the chippery.

And in my very, very humble opinion, I am seeing people come in with new computers, particularly laptops, doing very well and then slowly fading, not that they have dropped in skills as such.
If the hardware is providing any game advantage beyond speed of execution, by that I mean you get to the same place faster, and number of clients etc then we need a spec.

Joined: 09/21/2011

i'm just adding this info for people to compare. my pc is managing for now.

intel core i7-3770 3.4ghz
amd radeon hd6570 3gb
ram 10gb

while running two clients:
core temps 179%(mode) 176-182%(range)
fan one at 2350rpm (2200-2400)
fan two at 1350rpm (not sure the size of fans)

not running any other programs.
not sure this helps without a baseline for this pc when running programs like this. i can only say that no variables were added since 01/13 when i started using this pc (ie, no cat with pounds of fur built up inside the case).

i am well within capacity, but a third client would have my fans roaring. this time last year i would run four clients and you wouldn't even hear the fans because they barely moved.
proteins are running about the same speed as they did in the past with two clients, but adding clients shows a definite slow down compared to last year.

gitwut's picture
User offline. Last seen 3 years 3 weeks ago. Offline
Joined: 05/18/2012
Groups: Contenders

1. Intel Core i7-920 Processor LGA 1366 (8GB RAM)
2. Intel Core i7-2670QM CPU @ 2.20GHz (8GB RAM)
3. Intel Core i5-3470 CPU @ 3.2GHz (8GB RAM)
4. Intel Core 2 CPU 6400 @ 2.13GHz (4GB RAM)

Running more than 2 clients on computers 1-3 results in one client freezing up.

Computer 1 has given heat warnings with 2 clients running.
Computer 2 has given heat warnings with 4 clients running.
Computer 4 is slow, no point in trying to run more than 2 clients.

Joined: 04/19/2009

Computer specs:

Mac 10.7 - 3.1 GHz Intel Core i5 - Memory 8 gb 1333 MHz DDRS - four cores

Almost 2 years old

Please see this feedback for proof of overheating compared to a year ago:

wisky's picture
User offline. Last seen 1 year 18 weeks ago. Offline
Joined: 07/13/2011

Computer specs:

i7-3820 clocked at 4.1ghz - 32gb ram

6 clients running rebuilds the machine at 60°C with CPU fan running between 2700 and 2850 rpm.

MurloW's picture
User offline. Last seen 35 weeks 5 days ago. Offline
Joined: 11/21/2012

Intel i5-2500 quad-core CPU running @ 3.4 GHz
GeForce 9800 GTX+

No heat issues -> custom extra cooling.

860 + 861 + 863 = 65~85% CPU load.

any 4th client will bring CPU load to 90~100% consistently.

I was able to run 8 clients last year; 6-7 if it included large symmetry.

Susume's picture
User offline. Last seen 22 hours 41 min ago. Offline
Joined: 10/02/2011

Timo as you collect the CPU model numbers, you might look for a pattern in CPU cache size. Larger cache will run faster but hotter, as the CPU will not be waiting for the bus to fetch instructions from RAM. It looks to me like your laptop has 4MB cache; I have 3MB, and we are not having new heat problems. Auntdeen with i5-2400 has 6MB, gitwut's machine #1 has 8MB (2 clients = hot) and #2 and #3 have 6MB (4 clients = hot on #2). Karsten has 8MB.

This part is pure speculation: Possibly before Oct 2013 or so (this is when auntdeen thinks the heat problems started) the ratio of calculations to I/O in foldit was low enough that everyone's CPUs were waiting on I/O enough to stay cool. After that point perhaps the proportion of calculations grew so that I/O is no longer a significant limiting factor, and instead it is only waiting for RAM access that keeps the CPU from running at the max % the operating system will allow. Since that time, systems with smaller cache (like mine) would run slower as the calculations get more intense but not hotter, and systems with larger cache (like auntdeen's and karsten's) would run hotter.

There are other changes besides calculations:I/O ratio that could cause the app to run hotter. Ratio of loops short enough to fit in cache vs loops too long to fit in cache could have gone up. Ratio of floating point calculations to other types of instructions could have gone up (though this would affect all machines). Again, pure speculation.

Joined: 12/23/2011
Groups: Gargleblasters

It would help me if I knew what specific parameters you want me to report other that chip set, ram and heat at various numbers of clients running.

wisky's picture
User offline. Last seen 1 year 18 weeks ago. Offline
Joined: 07/13/2011

CPU: i7-3820 (10MB L3 cache) - clocked at 4.1GHz, 32GB memory, GeForce GTX 580 - 1.5GB video memory

Foldit info: Latest devprev update, view options: Score/Hydro/CPK, Cartoon view, Show all sidechains, all other options disabled.

All values are eyeball averages (i.e. if I see the temp fluctuating between 39-41, I say 40 C), unless the fluctuations are rather large

Baseline - 0 clients

-CPU fan: 1850 RPM
-CPU usage: 1.0%
-Temp: 35 C

1 client open, loaded into puzzle 861
Maximized, no recipes running
-CPU fan: 2070 RPM
-CPU usage: 8.5%
-Temp: 40 C

Minimized, no recipes running
-CPU fan: 1950 RPM
-CPU usage: 2.0-4.0%
-Temp: 35 C

Maximized, Rav GAB 2.0.6 running
-CPU fan: 2390 RPM
-CPU usage: 21-23%
-Temp: 50 C

Minimized, Rav GAB 2.0.6 running
-CPU fan: 2300-2360 RPM
-CPU usage: 14%
-Temp: 40-50 C (fluctuates)

After closing client
-CPU fan: 1860 RPM
-CPU usage: 1%
-Temp: 35 C

2 clients open, loaded onto puzzle 861
Both maximized, no recipes running
-CPU fan: 2350 RPM
-CPU usage: 21-24%
-Temp: 45-50 C (fluctuates)

Both minimized, no recipes running
-CPU fan: 1970 RPM
-CPU usage: 2%
-Temp: 35 C

Both maximized, Rav GAB 2.0.6 running
-CPU fan: 2540 RPM
-CPU usage: 42-45%
-Temp: 55 C

Both minimized, Rav GAB 2.0.6 running
-CPU fan: 2540 RPM
-CPU usage: 27%
-Temp: 45-55 C (fluctuates)

After closing both clients
-CPU fan: 1850 RPM
-CPU usage: 1%
-Temp: 35 C

3 clients open, loaded to puzzle 861
All maximized, no recipes running
-CPU fan: 2310 RPM
-CPU usage: 24-27%
-Temp: 45-40 C (fluctuates)

All minimized, no recipes running
-CPU fan: 1940 RPM
-CPU usage: 2-3%
-Temp: 35 C

All maximized, Rav GAB 2.0.6 running
-CPU fan: 2630 RPM
-CPU usage: 60-61%
-Temp: 57-60 C

All minimized, Rav GAB 2.0.6 running
-CPU fan: 2620 RPM
-CPU usage: 40%
-Temp: 50-57 C

After closing all clients
-CPU fan: 1850 RPM
-CPU usage: 1%
-Temp: 35 C

6 clients open (What I use on a regular basis), loaded

to puzzle 861
All maximized, no recipes running
-CPU fan: 2520 RPM
-CPU usage: 49-53%
-Temp: 55 C

All minimized, no recipes running
-CPU fan: 2000 RPM
-CPU usage: 2.5-3.5%
-Temp: 35 C

All maximized, Rav GAB 2.0.6 running
-CPU fan: 2800 RPM
-CPU usage: 99-100%
-Temp: 62 C

All minimized, Rav GAB 2.0.6 running
-CPU fan: 2800 RPM
-CPU usage: 80%
-Temp: 62 C

After closing all clients
-CPU fan: 1850 RPM
-CPU usage: 1%
-Temp: 35 C

Joined: 09/19/2011
Groups: Beta Folders

Potential solution for others that is working for me. Although, the jury is still out on whether or not my clients will be slower.

I spent hours doing a lot of upgrading, including the bios. Nothing changed the temperature problem. I've only been able to run two clients on my laptop if the puzzles were small. Running three took my high tempt to 94 c. That's unacceptable and I just couldn't do it.

Interestingly, the solution for me was in the power settings. My computer was set at "Ultra Performance" and when I switched it to "Cool" it resolved my overheating issue. Running four clients now with the high temp about 35 degrees lower.

Here's a snap of my power options:

Thanks to Karsten for the suggestion to view my Power options.

According to Susume, win7 usually only comes with three options. I have win7 professional on my laptop and have nine options. My desktop has win7 home edition and only has two options.

Those with overheating issues may want to take a look and experiment with these power settings.

I haven't a clue what changed in the recent foldit upgrades, but, my overheating started with the introduction of filters and became worse with the introduction of new chapter.

Hopefully, this information helps others.

Joined: 09/24/2012
Groups: Go Science

MacBook Pro (mid 2012)
2.9 GHz Intel core i7
8 Go 1600 MHz DDR3

3 clients running since 16-82 hours (2 devprev + 1 NC normal)

Now, "CPU usage":
puzzle 862, 70 residues ----- 140%
puzzle 861, 57 residues ------130%
puzzle 863, 40 residues ------100%

Never had heating problem neither before neither after NC (ventilator noise the all time shows that CPU is working hard)

Joined: 09/24/2012
Groups: Go Science

I forgot to mention that this is a maximum: CPU charge = 95-99%

Before NC, I also always used to run 3 clients.
After NC, with a big protein in the list, I can only run 2 clients.

(so yes, NC is more resource consuming, may be something like 33% more)

Joined: 09/24/2012
Groups: Go Science

BTW, I always put performances to maximum available, and I always deactivate all kind of energy savings.


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons