Folding Forum

I'm the president of the ACM Student Chapter at my university and our organization has decided to look into getting our university to put Folding@Home on computer around campus. Obviously contributing to F@H is a great thing to do, but in a time when a lot of focus has been on "going green" we don't want to approach the board and say, "It will use more electricity, generate more heat, and cost you more, but it's for a good cause!" without know more precisely how much more. We've search around for some figures, but a lot of what we found is old and out of date. We would like to set up a test lab where we actually monitor the power usages of several computers, some without F@H, some with F@H, but ones that require the user to log in each time the PC is used, so F@H won't be running all the time, and others with F@H running 24/7. We also plan on adjusting the amount of CPU that F@H is allowed to use, to try to find a setup that will not interfere with users but still contribute as much as we can. Before we start this endeavor I felt it would be a good idea to post here and see if other tests like this have been done, and to look for input as far as what we could do to get the most out of this test, or other things we may need to consider when trying to convince the university to invest in this project.

One other issue we may come across is the licensing. According to the License Agreement you can't alter the installation package to install F@H on a distributed system and would have to go to each computer to install it. I've searched Stanford's site for a contact email to ask them about this, but I cannot seem to find one. Who should I contact about this?

At the very least, we have set up a F@H team for our university and we plan to get students involved using their own personal computers for now.

A lot depends on your geographic location. See viewtopic.php?f=38&t=20475&start=15 In our environment, our company found that leaving the lights on all night tended to save money by reducing the heating costs required when the employees came to work. One component of the energy bill was based on total usage and another (very significant) component was based on peak usage. The power utility likes customers who use power uniformly.

FAH can use as little or as much as you use it. There are numerous discussions on this forum where people have calculated the PPD/W for GPUs or for CPUs from various generations of hardware. Most people measure the power with and without FAH using a "kill-a-watt" meter. Obviously running FAH on servers which run 24x7 anyway is less expensive than running it on computers with well-designed power savings setting that shut down or sleep the computer when it's not in use.

We would like to set up a test lab where we actually monitor the power usages of several computers, some without F@H, some with F@H, but ones that require the user to log in each time the PC is used, so F@H won't be running all the time, and others with F@H running 24/7. We also plan on adjusting the amount of CPU that F@H is allowed to use, to try to find a setup that will not interfere with users but still contribute as much as we can.

This setting would vary tremulously with the age of the computer(the CPU), the OS, the usage pattern, etc. etc.

(I am assuming that like many institutions,that hardware is purchased on a rotation, a small percentage replaced periodically rather than every computer on campus replaced at the same time.)

My suggestion,(take it for what its worth) is to find specific hardware that is under-utilized...and look at applying to that specific equipment as a test case.

You will find that asjusting the amount of CPU that FAH is allowed to use will generally not be necessary. For simple tasks like web browsing, Office, etc. the computer users can rarely tell that FAH is running because the OS is very effective at suspending tasks with extremely low priority when a task with normal priority also needs resources. The computer game enthusiast will see a detectable change if FAH is using 100% because as a group, they're just more sensitive to minor changes. Reducing FAH to 95% generally makes them happy.

That's not true If you fold with the GPU -- there is no priority-based task scheduler and screen lag may be noticeable.

Experimentation is good, though, as long as you use blind or double-blind testing methodology, including concealing things like disk activity lights or harddrive sounds. V7 can control FAH's activity remotely.

One thing you might be able to do is use Joulemeter to measure the power usage by a single WU on a single computer (this doesn't measure power required to cool down with the fans, though). It might give you a rough estimate.