University Computer Labs
 

I was looking, my University has over 100 new P4's at 3.06 GHz in the library alone. I was wondering if anyone had any suggestions for approaching and convincing the University of running Prime95. That would be some phenomenal power, at least compared to little ol me.

You can try it, don't know exactly how you should do it, but I asked the same at my university and they told me that when the computers in any room already do research of the university itself in their spare time.

So I didn't bother to go to the big bosses here.

PS : You might want to check these things with the local technician of one building first.

I find the best way to talk someone into a DC project is to be casual and to tell them about it in bits and pieces. My local High School has about 52 P4's 1.5-2.4ghz and 13 AMD Athlon 1500-2000+'s spread across the campus (I wont even mention the PII's :? ). I have the system admin trying Prime95 out on a hand-full of the machines, but it took a good week just to get him around to look at it (because of a busy schedule). We will approach the school board about running it on the higher end PC across the school district if all goes well :D .

You should draw their attention to other examples of Universities/Institutions that are running P95 without any problems (see for example [b]curtisc [/b]or [b]ganser[/b]). The idea that they would be contributing to new math discoveries and that the discovery of a new prime would be a good form of advertising the University shall also be considered. But I think the bottomline is to clearly show they have nothing to loose in running the program. Start by running it on a trial basis on a couple of machines, to demonstrate it is really "harmless".
Man, 100 3.06G P4s is really too much power to go to waste :shock:. It is at least worth trying...

There are *advantages* to running mprime!
 

One argument for running mprime/prime95 is the early detection of hardware problems.

While I didn't have to convince anyone else of the benifits of running mprime, I did have to convince myself. My conclustion is that there's no downside, and a rather important upside -- mprime/prime95 can detect faulty hardware where no other testing system does.

I would advise, however, that the distributed systems run only a DC or LL test -- the P-1 test is rather heavy in the memory usage, and this can cause paging/thrashing when the user is using other memory intensive apps. What I do is have a half dozen non-user machines which do any required P-1 testing, and deployed machines run the LL/DC tests.

I've written a small bit of Perl code which handles all this for me. My machines check in with a proxy, which allocates work as appropriate. (Be careful though -- my system recently caused me a bit of embarrasment when it grabbed a large number of early exponents when George released them just before my reservation script ran...)

My system reports (in flashing red) any returned results which had errors during the LL/DC process. I've only had one machine come back with errors, but it died two weeks after the first reported error. When everything is good (black), you (or the sysadmin) can be fairly confident that the system(s) are solid.

Rather than trying to justify the benign nature of mprime/prime95, you may want to also argue this point.

[quote="lycorn"]But I think the bottomline is to clearly show they have nothing to loose in running the program. [/quote]

Many universities are on tight budgets. You should be upfront that running prime95 on a computer probably costs 50 watts of power. Multiply that by 100 machines running 24 hours a day, 365 days a year and you've just spent some real money. Correct me if I'm wrong...

Yes, you are right BUT ...

our university runs their computers (about 50 P4's in our faculty alone) from 9am till 11PM (which is 14 hours a day) anyway, while there are only about 10 constantly being used. So in terms of electric power, what difference does it make that they are just sitting there with no one working on them or they are running Prime95 ?

[quote]So in terms of electric power, what difference does it make that they are just sitting there with no one working on them or they are running Prime95 ?[/quote]

Running prime95 draws an [b]extra[/b] 50 Watt compared to an idle pc.

Can you show us the numbers?

I dont know the exact number but I know my laptop vent is ice cold when cpu is not busy. Once prime95 starts, the fan never stops blowing hot air.

It has a P4M 2.0G.

Actually, I'd say that 50 watts is probably a little on the high side for most machines.

My P4 (Shuttle SS51G) uses 20-25 more watts of power when running Prime95 than when idle - very similar increase for any of my Athlon XP1800+ systems.
The systems I have doing TF work tend to increase less - but they're lower-power systems to start with.


All figures *measured* on a Brand 1850 wattmeter - NOT extrapolated or estimated.


On the other hand, pretty much ALL of my machines use more power once Prime fires off than while they are going through the bootup process....