Does Linux generate more heat?

Moderators: Site Moderators, PandeGroup

Re: Does Linux generate more heat?

Postby Stonecold » Thu Feb 16, 2012 2:36 am

k1wi wrote:
Grandpa_01 wrote:
k1wi wrote:I'm planning to give running F@H in a VM another shot as I am aware that the linux core appears to provide better performance. I will have to keep an eye on the thermals of my machine.


You will not be able to tell anything using a VM when you run a VM the VM is dependent upon the OS it is running on. I run a vm on different rigs at different times for testing while Linux reports 100% cpu usage the Windows host will show between 95% and 100%. So the VM uses 100% of what windows provides to it. Not 100% of actual cpu. :wink:
I am aware that running a VM results in a decrease in performance relative to running on a native OS, but for me the only thing I am interested in is whether my TPF decreases (and PPD as a result increases) as a result of running the VM... The reason I am planning on running a VM instead of in native Linux is because running native Linux on that computer is not an option for me.

Therefore, what I am testing is whether FAH on Linux in a VM is faster or slower than natively on Windows... If it is faster it means the Linux FAH core is more efficient than the Windows FAH despite the VM overhead, if it is not then any greater efficiency of the Linux core does not exceed the overhead of running in a VM...

If it does run faster then I want to keep an eye on how much hotter my CPU runs.

I think I heard somewhere that it was 15-20% faster.
Stonecold
 
Posts: 392
Joined: Sun Dec 25, 2011 9:20 pm

Re: Does Linux generate more heat?

Postby PantherX » Thu Feb 16, 2012 2:42 am

From what I have read, with native Linux installation, it would be ~20% faster than Windows. With VM, it would still be ~15% faster than Windows. The temperature difference for the CPU would be interesting and am looking forward to it.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Chrome Folding App (Beta) Ӂ Troubleshooting "Bad WUs" Ӂ Troubleshooting Server Connectivity Issues
User avatar
PantherX
Site Moderator
 
Posts: 6321
Joined: Wed Dec 23, 2009 9:33 am

Re: Does Linux generate more heat?

Postby Stonecold » Thu Feb 16, 2012 2:44 am

PantherX wrote:From what I have read, with native Linux installation, it would be ~20% faster than Windows. With VM, it would still be ~15% faster than Windows. The temperature difference for the CPU would be interesting and am looking forward to it.

So by more efficient it means that it pushes the CPU harder and not that there is less overhead?
Stonecold
 
Posts: 392
Joined: Sun Dec 25, 2011 9:20 pm

Re: Does Linux generate more heat?

Postby Grandpa_01 » Thu Feb 16, 2012 2:45 am

F@H on a VM on windows is faster than Windows but not much. My wife's rig is a 970@ 4.3Ghz it runs a VM and it is a little faster than native windows. It does create a lag in Windows though. The only reason I run the VM on her rig is to run Bigadv. When it is running smp I let it run on windows there just is not a big enough difference in PPD to listen to my wife's griping about the lag. :ewink:
Image
2 - SM H8QGi-F AMD 6xxx=112 cores @ 3.2 & 3.9Ghz
5 - SM X9QRI-f+ Intel 4650 = 320 cores @ 3.15Ghz
2 - I7 980X 4.4Ghz 2-GTX680
1 - 2700k 4.4Ghz GTX680
Total = 464 cores folding
User avatar
Grandpa_01
 
Posts: 1757
Joined: Wed Mar 04, 2009 7:36 am

Re: Does Linux generate more heat?

Postby Grandpa_01 » Thu Feb 16, 2012 2:47 am

Stonecold wrote:
PantherX wrote:From what I have read, with native Linux installation, it would be ~20% faster than Windows. With VM, it would still be ~15% faster than Windows. The temperature difference for the CPU would be interesting and am looking forward to it.

So by more efficient it means that it pushes the CPU harder and not that there is less overhead?


No it means Linux is more optimised for F@H it has nothing to do with the CPU.
User avatar
Grandpa_01
 
Posts: 1757
Joined: Wed Mar 04, 2009 7:36 am

Re: Does Linux generate more heat?

Postby 7im » Thu Feb 16, 2012 3:51 am

No!!! It means the fahcore is more optimized for Linux, and has everything to do with FAH, and nothing to do with any overhead in the OS. 8-)
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Does Linux generate more heat?

Postby bruce » Thu Feb 16, 2012 8:24 pm

You have not mentioned what project(s) are being assigned. If Windows was running project X and Linux is running project Y there is likely a difference in performance that absolutely nothing to do with either the FahCore or the OS.

There are also subtle differences between Linux and Windows when you're trying to get maximum performance of a workload that combines FAH with other processes. There was a complaint recently about FAH still using a significant percentage of CPUs even though the FAH processes were set for a maximum value of nice (minimum priority). Obviously how the OS allocates the resources depends on MANY factors ... but mostly by whatever else is running on the system.
bruce
 
Posts: 20837
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Does Linux generate more heat?

Postby Stonecold » Thu Feb 16, 2012 8:46 pm

bruce wrote:You have not mentioned what project(s) are being assigned. If Windows was running project X and Linux is running project Y there is likely a difference in performance that absolutely nothing to do with either the FahCore or the OS.

Yes, the projects are different. The last projects Linux ran were (from most to least recent) 6097, 6098, 6098, 6099, 6097, and 6097. The last projects Windows ran were 7610, 8001, 8001, 8004, 8001, and 8004. Do the former projects generate more heat?
Stonecold
 
Posts: 392
Joined: Sun Dec 25, 2011 9:20 pm

Re: Does Linux generate more heat?

Postby k1wi » Thu Feb 16, 2012 10:28 pm

Regarding relative 800* vs 609* temperatures, I just happened to complete an 8004 and start on an 6096. The graphs at the end of this post (click to open 100% large copies) illustrate that there is little or no variation in CPU temperature as a result from the WUs, once ambient temperature is accounted for.

On both graphs, core temps of my CPU are measured on the left hand axis, while ambient temperature is measured on the right hand axis, both in degrees C. Graph one shows the 24 hour variation in computer temperatures, while Graph one two shows the temperatures across the week. Sharp troughs in the data indicates F@H sending and receiving a work unit at the point of logging. Long troughs with sharp sides indicate folding was shut down and the computer was used for other processes. Generally speaking, FAH is only run when the computer is not being used for other tasks and likewise, other tasks are only run when FAH has been shut down temporarily. Unfortunately, the x-axis does not display the time so events and raw data are used to identify points of time in the graph.

According to my FAH log, two P8004 projects and one P6906 projects have run this morning (local time). The first P8004 ran for one hour, between 17:38UTC and 18:42UTC. On graph 1 18:43UTC is be marked by the rapid increase from idle to 100% CPU utilisation and temperature increase from ~30DegC to ~65DegC, (x = ~245) followed by a roughly 25 minute period where CPU temperature stabilises @ just under 70DegC (Core 0). This 25 minute period reflects the amount of time it takes to heat the water in my water cooling loop (8m of 19mm tubing and a 2litre reservoir).

The second P8004 started immediately after the first P8004 finished uploading, at 18:43UTC and completed at 19:46UTC. By the completion of this work unit, Core 0 has stabilised at 71DegC. At 19:47UTC the P6096 project began and temperatures remained at ~70-72Degrees for about 55 minutes, until 20:45UTC. Between 20:45UTC and 21:00UTC, temperatures increased from 73DegC to 75DegC..

A complicating factor in this analysis is ambient temperature. My watercooling radiator is located underneath my house, with access to a mix of outside air and air from my garage (where my computer is located). Long-term analysis by myself suggests that there is a roughly one hour lag between ambient temperature increase and CPU temperature increase - Core temperatures tend to peak roughly an hour after the hottest temperature of the day, and start increasing roughly an hour after temperatures increase for the day. Furthermore, regression analysis (yes I am that obsessed with stats) indicates that a 1 degree ambient temperature increase results in a 0.8 degree increase in CPU temperature.

18:25UTC represents the minimum ambient temperature of 8.2DegC, from 19:00UTC until 21:00UTC, ambient temperature increases approx. 3.7DegC/hour. Therefore, given that core temperatures remained at a relatively stable during the first 55 minutes of P6096, despite an increase in outside ambient temperature, my conclusion is that the increase in temperature after 20:45UTC is associated with ambient warming. This indicates, that in the Windows environment, there is no increase in heat output between P8004 and P6096 WUs and any difference between projects folded under Windows and Linux is therefore likely to be due to either the FAH Core or OS.

Further supporting this result is the temperatures over the previous day, when only P800* projects were being folded (as they had not yet been released to the public). Core temperatures while folding a p8004 WU at 18 DegC (x = ~50-55) are inline with those found this morning.

Image
graph 1
Image
graph 2

[ADD]Some additional information: The computer is a i7 920 D stepping CPU @4009 MHz, running on a P6t with 12GB of ram @ 1600 MHz, Windows 7 64bit is the operating system and the FAH client is currently V6.34.

Core temperatures and utilisation are measured and logged every 30 seconds by CoreTemp. Every five minutes the latest row of logging is collated by a custom script and appended to a .csv file for Gnuplot to graph and also exported to a mysql database. The custom script also logs a range of other data, including local ambient temperature, humidity etc.
k1wi
 
Posts: 1152
Joined: Tue Sep 22, 2009 10:48 pm

Re: Does Linux generate more heat?

Postby Stonecold » Thu Feb 16, 2012 10:38 pm

Wow, that's a lot of data! It looks like you really took your time looking this up! So in summary, WUs from p8004 to p6906 don't have a significant difference in heat production when running under Windows? Okay, so the extra heat is probably due to the OS (every project I've run with Linux has generated extra heat). Hopefully this means it's working harder and pumping WUs out faster. What program did you use to capture that data, by the way?
Stonecold
 
Posts: 392
Joined: Sun Dec 25, 2011 9:20 pm

Re: Does Linux generate more heat?

Postby k1wi » Thu Feb 16, 2012 10:46 pm

I think it is too early to say it is due to the OS... Rather, it is due to either the OS or the FAH Core being more efficient in utilising the available hardware. You can test 'working harder and pumping WUs out faster' by comparing TPF/PPD from one project under windows to the same project under Linux, but it would be my expectation that harder does = faster.


I added some information on how I capture the data to my previous post.
k1wi
 
Posts: 1152
Joined: Tue Sep 22, 2009 10:48 pm

Re: Does Linux generate more heat?

Postby Stonecold » Thu Feb 16, 2012 10:48 pm

k1wi wrote:I think it is too early to say it is due to the OS... Rather, it is due to either the OS or the FAH Core being more efficient in utilising the available hardware. You can test 'working harder and pumping WUs out faster' by comparing TPF/PPD from one project under windows to the same project under Linux, but it would be my expectation that harder does = faster.


I added some information on how I capture the data to my previous post.

How do I test whether it's the OS or the FAH core that's contributing to the extra heat production?
Stonecold
 
Posts: 392
Joined: Sun Dec 25, 2011 9:20 pm

Re: Does Linux generate more heat?

Postby k1wi » Thu Feb 16, 2012 11:03 pm

Also, the graphs show why any increased temperature, even small, is of importance to me. The red line @ Y1 = 80 in graph 1 represents the peak core temp I am comfortable with and should only be exceeded temporarily and infrequently. Analysis shows that, for my set up, when folding natively on the Windows client, when the ambient temperature is roughly 24 degrees, it is highly probable that this core temp will be exceeded.

Fortunately, my location is particularly mild and there are very few instances where the temperature is below 0 DegC or above 24 DegC.

In response to your question whether it is OS or FAH core related... The short answer is I don't believe you can, as you cannot test windows FAH on Linux or the Linux FAH core on Windows... But I also don't believe it is nessessary to do so. What is important is whether, at this given point in time, FAH on Linux generates more heat (and whether this is reflected in higher productivity or PPD) than FAH on Windows. Knowing whether it is the FAH core or the OS is not really important (in my opinion).

My own personal opinion is that 7im's conclusion that it is the efficiency of the FAH core rather than the OS is the correct hypothesis. This has been shown by changes in efficiency across revisions of the FAH core on the same OS.
k1wi
 
Posts: 1152
Joined: Tue Sep 22, 2009 10:48 pm

Re: Does Linux generate more heat?

Postby 7im » Thu Feb 16, 2012 11:53 pm

Stonecold wrote:
k1wi wrote:I think it is too early to say it is due to the OS... Rather, it is due to either the OS or the FAH Core being more efficient in utilising the available hardware. You can test 'working harder and pumping WUs out faster' by comparing TPF/PPD from one project under windows to the same project under Linux, but it would be my expectation that harder does = faster.


I added some information on how I capture the data to my previous post.

How do I test whether it's the OS or the FAH core that's contributing to the extra heat production?


Take temp readings from Windows, then run fah with a project 8xxx. How much does heat increase?

Take temp readings from Linux, then run fah with a project 8xxx. How much does heat increase?

Which one had the higher increase as a percentage of the total?
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: Does Linux generate more heat?

Postby bruce » Fri Feb 17, 2012 1:32 am

Does the FahCore report the same version number?
bruce
 
Posts: 20837
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

PreviousNext

Return to Linux CPU V6 Client

Who is online

Users browsing this forum: No registered users and 1 guest

cron