TR Forums

ChrisDTC · Mon Nov 19, 2007 5:48 pm

I installed a Windows v6 console client on a machine and the process is only running at 50%, and has its priority set to low, and it wont let me change the priority through the task manager

Ive gone through using the -configonly flag several times but I cant get it to run at any more than 50%

flybywire · Mon Nov 19, 2007 5:56 pm

The v6 client is for single core system.. is your computer dual-core?

ChrisDTC · Mon Nov 19, 2007 5:58 pm

negative, 2.8 P4

flybywire · Mon Nov 19, 2007 5:59 pm

Is Hyperthreading enabled on that chip?

Firestarter · Mon Nov 19, 2007 6:09 pm

flybywire wrote:
Is Hyperthreading enabled on that chip?

That would explain 50% load.

ChrisDTC · Mon Nov 19, 2007 6:14 pm

http://www.pittspeed.com/uploaded/p4.JPG

ChrisDTC · Mon Nov 19, 2007 6:17 pm

checked in BIOS, it said HT was enabled, I disabled it and will try to re-enable it

ChrisDTC · Mon Nov 19, 2007 6:20 pm

well, now its working with HT disabled, who knows

whatever Im pretty sure its going now

thanks guys

crazybus · Mon Nov 19, 2007 7:17 pm

With a HyperThreading cpu, 50% cpu usage when running a single threaded application actually means that the cpu is running at 100%. Disabling HyperThreading [i]may[/url] speed up folding slightly but will probably give you a net performance decrease in overall system speed and responsiveness.

PRIME1 · Mon Nov 19, 2007 7:36 pm

Enable hyperthreading and run 2 copies of the console. One with the -local switch and change the machine number. That's what I do and you will get more PPD.

Nitrodist · Mon Nov 19, 2007 9:40 pm

PRIME1 wrote:
Enable hyperthreading and run 2 copies of the console. One with the -local switch and change the machine number. That's what I do and you will get more PPD.

IIRC, Stanford does not want you to run HT. You will earn more points with the HT on but overall the work units do not complete as quickly.

Ragnar Dan · Mon Nov 19, 2007 11:05 pm

Your machine, your electricity, your decision. Stanford can wish all they want. More PPD means more overall throughput for them, no matter their complaints.

Nitrodist · Mon Nov 19, 2007 11:14 pm

Hyperthreading CPUs

Running two instances of F@H will cause both WUs will be completed more slowly than running them one at a time, but this will generate approximately 15-30% more points than running a single instance. On the surface, this sounds great, but this is not as beneficial to the overall folding effort as it sounds on the surface. Projects are released in generations, and all of the WUs in one generation must be completed prior to the start of the next generation. By slowing down individual work units, it slows the completion of one generation and progression on to the next generation. Generations can number in the hundreds, so a delay of even a few hours can cause significant delays in the completion of a project. So, even though you're cranking out more points, you may actually be slowing things down.

Some site.

jss21382 · Tue Nov 20, 2007 12:13 am

Your machine, your electricity, your decision. Stanford can wish all they want. More PPD means more overall throughput for them, no matter their complaints.

Guess it's a matter of whether you're doing it for the competition or to contribute to science

Tue Nov 20, 2007 12:44 am

jss21382 wrote:
Your machine, your electricity, your decision. Stanford can wish all they want. More PPD means more overall throughput for them, no matter their complaints.

Guess it's a matter of whether you're doing it for the competition or to contribute to science

The two aren't necessarily mutually exclusive; it's a matter of the relative value you place on each.

While I personally refuse to run multiple clients on hyperthreaded machines for the reasons stated -- and urge others to do likewise -- I also understand that it's a personal decision. And at the end of the day, I'm sure Stanford would still rather have the computing cycles than not, as long as the WUs are turned in before the deadline.

One could also argue that the point system is flawed if it encourages behavior that is considered non-optimal. If Stanford really values quick turnaround, then perhaps they should award a small additional point bonus for WUs that are returned quickly.

PRIME1 · Tue Nov 20, 2007 12:56 pm

I run 2. They both make it in well under deadline. I really don't see the problem. How is it slowing their research down if my WUs make it in well before the deadline?

How is the research slower if they are actually rewarding more points?

Not to mention that one copy can keep folding while the other is sending\receiving or if it hits an error.

Saying it slows their already slow research makes very little sense as there are people using slower CPUs that probably don't even send one in the amount of time mine can send 2.

EDIT: I just checked my WU is due January 10th. :lol:

ChrisDTC · Tue Nov 20, 2007 1:47 pm

since I have several machines folding, Ill just run 1 client with HT enabled

41.60l23 · Wed Nov 21, 2007 6:38 pm

I am actually running the SMP client on my P4 HT @ 3.52 GHz. It's doing about 45 min/frame, enough to make deadlines.

41.60l23 · Thu Nov 22, 2007 11:21 am

Hah. Finished an SMP unit on a Pentium 4 HT.. who would've thought?

Thu Nov 22, 2007 6:37 pm

Congrats! About 590 PPD from a P4 is pretty tasty. Makes me wonder if my 3.0 can make deadlines.

41.60l23 · Thu Nov 22, 2007 11:15 pm

I'm sure it can. It was running at 3.2 most of the time, and was turned off multiple times.. overclock that sucker and there's no question.

TR Forums

Windows v6 Console client only running at 50%

Windows v6 Console client only running at 50%

Who is online