Hardware Secrets Forums


Go Back   Hardware Secrets Forums > Miscellaneous > Content Comments



Reply
 
Thread Tools Search this Thread Display Modes
Old 10-29-2008, 06:03 PM   #1
Hardware Secrets Team
Administrator
 
Join Date Nov 2004
Posts: 5,598
Hardware Secrets Team is on a distinguished road

Default Seeking The Best Performance per Watt for Folding@Home

There has been a new article posted.

Title: Seeking The Best Performance per Watt for Folding@Home
URL: http://www.hardwaresecrets.com/article/640

Here is a snippet:
"We got so excited in participating in the Folding@Home project that we built as many high performance systems we could running both the SMP and GPU clients. We were very happy with the results until w..."

Comments on this article are welcome.

Best regards,
Hardware Secrets Team
http://www.hardwaresecrets.com
Hardware Secrets Team is offline   Reply With Quote
new Sponsored Links

Old 10-30-2008, 02:11 AM   #2
Olle P
Senior Member
 
Olle P's Avatar
 
Join Date Oct 2008
Sweden
Posts: 1,736
Olle P is on a distinguished road

Default

Interesting article!

A few comments and questions though:

You recommend the use of a wattmeter. Always nice if you have one, but they're considerably more expensive than a vanilla multimeter that will do the job good enough.
Power (P) = voltage (U) * current (I)
The voltage is as good as constant, so it's sufficient to measure that once, before plugging in the computer, and then track the current while the computer is running.
Multiply the current with the voltage and you get the power. (Actually, you get the apparent power, a combination of effective and reactive power, which is marginally larger than the effective power which is what you have to pay for, but that doesn't really matter.)
______________________________________

The second test range, where you just switch graphics cards, what folding client was used? GPU alone or both GPU and SMP? Tried both varieties to find the best?
(With a low end G-card and fairly fast CPU running both clients might be better than any one alone.)
______________________________________

When you ran both clients on a dual core CPU with one G-card it seems strange that the clients didn't use one core each, while the SMP client did use the fourth core in system 5.
And what about system 4, did the SMP client use a third core there leaving one core idle or did it use both the spare cores?
______________________________________

If building a computer with the sole purpose of running the Folding GPU client, then perhaps one consideration should be to use a single core CPU? (But then again, new CPUs that invariably are multi-core might be more energy efficient than older single core CPUs. Are they?)
______________________________________

Suggestions for further tests, if desired:
How does variations in CPU effect the performance of the GPU client for a specific G-card? (Changing the CPU clock speed should suffice for this test.)
I guess the GPU client actually use the CPU for part of the calculations, then with a faster CPU core the calculations should speed up as well.

What's the "Performance per Watt" when running Folding on a computer while using it for regular work?
The average performance is less, since the CPU and GPU is often busy doing higher priority tasks. At the same time the average power consumption should only be slightly higher than when the Folding clients are off.
Perhaps you get the best performance per watt by keeping the computer off while you're not busy using it (at the cost of a considerable drop in actual "folding" process being performed)...

Cheers
Olle

Last edited by Olle P; 10-30-2008 at 02:17 AM.
Olle P is offline   Reply With Quote
Old 10-30-2008, 04:22 AM   #3
Xajel
Junior Member
 
Join Date Oct 2008
Posts: 1
Xajel is on a distinguished road

Default

You might find this intersting

about HD 4xxx Folding performance

Quote:
Folding client not use all shaders it use amount of shaders 3xxx series new supported client on its way but when i dont know maybe mhouston know they got all shaders working on 4xxx series but it unstable ATI.AMD choose Stability in first place.
http://foldingforum.org/viewtopic.php?f=51&t=6165
Xajel is offline   Reply With Quote
Old 10-30-2008, 12:18 PM   #4
SirkoZ
Senior Member
 
SirkoZ's Avatar
 
Join Date Sep 2008
Posts: 25
SirkoZ is on a distinguished road

Default

I'm sure that HD4670 would be a great Folding card, it's got computing power higher and consumption lower than a GFFX 8800GT.
However - the client might not be optimised for it yet...
SirkoZ is offline   Reply With Quote
Old 10-30-2008, 01:51 PM   #5
Gabriel Torres
Administrator
 
Gabriel Torres's Avatar
 
Join Date Oct 2004
USA
Posts: 4,436
Gabriel Torres is on a distinguished road

Default

Quote:
Originally Posted by Olle P View Post
Interesting article!

A few comments and questions though:

You recommend the use of a wattmeter. Always nice if you have one, but they're considerably more expensive than a vanilla multimeter that will do the job good enough.
Power (P) = voltage (U) * current (I)
The voltage is as good as constant, so it's sufficient to measure that once, before plugging in the computer, and then track the current while the computer is running.
Multiply the current with the voltage and you get the power. (Actually, you get the apparent power, a combination of effective and reactive power, which is marginally larger than the effective power which is what you have to pay for, but that doesn't really matter.)
Vanilla multimeters can't read AC current, only DC current. You would need a clamp multimeter connected to the main AC cable:
http://www.multimeterwarehouse.com/266f.htm

Then by measuring the AC voltage (e.g. 115 V) you can get the wattage as you described.

So you would need to buy another tool anyway.

I will reply to your other questions later, because I am quite busy right now.

Cheers,
Gabriel.
Gabriel Torres is offline   Reply With Quote
Old 10-31-2008, 12:14 AM   #6
Olle P
Senior Member
 
Olle P's Avatar
 
Join Date Oct 2008
Sweden
Posts: 1,736
Olle P is on a distinguished road

Default

Quote:
Originally Posted by Gabriel Torres View Post
Vanilla multimeters can't read AC current, only DC current.
I beg to differ. Cheap multimeters might not be able to measure AC current, but every single one of the mid range (~$100) meters I was able to find in a quick survey around my work place had the ability to measure at least up to 10A AC through them.
- FLUKE models 23, 85 and 87
- SOAR 4050
- Mastech MY-64
(A quick survey at the standard electronics retailer I use also proved the point; all multimeters >400SEK are able to measure AC.)

You don't need a clamp either. Just a way to attach one lead from the mains cable to (through) the multimeter.

I've attached an image of my test cable.
A standard power cable with the outer shield split open a bit in the middle and one of the leads cut off. Then I attached a safety "banana" pin to each end of the cut lead, and taped up the other leads.
For testing I just plug the banana pins into the multimeters corresponding receptacles.

Cheers
Olle
Attached Thumbnails
Click image for larger version

Name:	Testkabel.JPG
Views:	151
Size:	118.5 KB
ID:	98  
Olle P is offline   Reply With Quote
Old 10-31-2008, 12:19 PM   #7
Gabriel Torres
Administrator
 
Gabriel Torres's Avatar
 
Join Date Oct 2004
USA
Posts: 4,436
Gabriel Torres is on a distinguished road

Default

Quote:
Originally Posted by Olle P View Post
I beg to differ. Cheap multimeters might not be able to measure AC current, but every single one of the mid range (~$100) meters I was able to find in a quick survey around my work place had the ability to measure at least up to 10A AC through them.
- FLUKE models 23, 85 and 87
- SOAR 4050
- Mastech MY-64
(A quick survey at the standard electronics retailer I use also proved the point; all multimeters >400SEK are able to measure AC.)
In your original message your clearly said VANILLA MULTIMETER. A USD 100 tool isn't a vanilla multimeter. A Fluke isn't a vanilla multimeter.

You message is right. However using this kind of multimeter you need to install it in series with the AC power cord, which requires a certain dose of knowledge. With a digital wattmeter you don't need to do that -- and it is way cheaper than a Fluke -- and that is why we recommend it over a expensive multimeter.

Cheers,
Gabriel Torres
Gabriel Torres is offline   Reply With Quote
Old 10-31-2008, 12:35 PM   #8
Gabriel Torres
Administrator
 
Gabriel Torres's Avatar
 
Join Date Oct 2004
USA
Posts: 4,436
Gabriel Torres is on a distinguished road

Default

Quote:
Originally Posted by Olle P View Post
The second test range, where you just switch graphics cards, what folding client was used? GPU alone or both GPU and SMP? Tried both varieties to find the best?
(With a low end G-card and fairly fast CPU running both clients might be better than any one alone.)
It is explained on the tables. "SMP" = SMP client only, "GPU" = GPU client only, "GPU + SMP" = both clients.

Quote:
Originally Posted by Olle P View Post
When you ran both clients on a dual core CPU with one G-card it seems strange that the clients didn't use one core each, while the SMP client did use the fourth core in system 5.
When I ran both clients on a dual-core CPU the load was at 100% and they were using both cores. Please clarify where you read otherwise.

Maybe you got confused somehow. When you run the SMP client all CPU cores are used and put at 100% load.

Quote:
Originally Posted by Olle P View Post
And what about system 4, did the SMP client use a third core there leaving one core idle or did it use both the spare cores?
Same as above. Running SMP client makes all cores to be used.

Quote:
Originally Posted by Olle P View Post
If building a computer with the sole purpose of running the Folding GPU client, then perhaps one consideration should be to use a single core CPU? (But then again, new CPUs that invariably are multi-core might be more energy efficient than older single core CPUs. Are they?)
Yes, if you are going to use only one video card and run only the GPU client. Notice that I downgraded the Phenom 9700 to an Athlon X2 4600+ because of that to save some watts. I couldn't use a single-core CPU because at this moment I don't have any single-core CPU available.

Quote:
Originally Posted by Olle P View Post
Suggestions for further tests, if desired:
How does variations in CPU effect the performance of the GPU client for a specific G-card? (Changing the CPU clock speed should suffice for this test.)
I guess the GPU client actually use the CPU for part of the calculations, then with a faster CPU core the calculations should speed up as well.

What's the "Performance per Watt" when running Folding on a computer while using it for regular work?
The average performance is less, since the CPU and GPU is often busy doing higher priority tasks. At the same time the average power consumption should only be slightly higher than when the Folding clients are off.
Perhaps you get the best performance per watt by keeping the computer off while you're not busy using it (at the cost of a considerable drop in actual "folding" process being performed)...
Thanks for the suggestions. Keep in mind that when you run the advanced configurations you can set the "idle level" (i.e. when to run the client) and "CPU load" (i.e. how much CPU load the client will use).

Cheers,
Gabriel.
Gabriel Torres is offline   Reply With Quote
Old 10-31-2008, 12:35 PM   #9
Gabriel Torres
Administrator
 
Gabriel Torres's Avatar
 
Join Date Oct 2004
USA
Posts: 4,436
Gabriel Torres is on a distinguished road

Default

Quote:
Originally Posted by Xajel View Post
You might find this intersting

about HD 4xxx Folding performance



http://foldingforum.org/viewtopic.php?f=51&t=6165
I didn't know that. Thanks.
Gabriel Torres is offline   Reply With Quote
Old 10-31-2008, 12:36 PM   #10
Gabriel Torres
Administrator
 
Gabriel Torres's Avatar
 
Join Date Oct 2004
USA
Posts: 4,436
Gabriel Torres is on a distinguished road

Default

Quote:
Originally Posted by SirkoZ View Post
I'm sure that HD4670 would be a great Folding card, it's got computing power higher and consumption lower than a GFFX 8800GT.
However - the client might not be optimised for it yet...
According to our tests, all ATI cards consume a lot of power and do not provide a high score compared to nVidia. We cannot talk about the future, we can only talk about the situation right now.

If you want to run a high-end Folding client, forget ATI. Go with nVidia.

Gabriel.
Gabriel Torres is offline   Reply With Quote
new Sponsored Links

Reply

Share This Thread

Bookmarks

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Forum Jump


All times are GMT -8. The time now is 01:09 AM.


vBulletin Security provided by vBSecurity (Lite) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
vBulletin Optimisation provided by vB Optimise (Lite) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd. ()
2004-12, Hardware Secrets, LLC. All rights reserved.