just aquired a p4 3 ghz prescott core machine

[jheil]

through the death of windows that is. I installed ubuntu 7.10 on it. Right now I'm having a hell of time getting the wireless setup, screwing around NDISwrapper. I got that installed using synaptec, actually getting it to install the windows drivers is tough for a linux virgin like me. Anyway, this isn't for linux support. I will figure it out sooner or later. I think it should do ok once it is running. Maybe in a month. lol

[RAH]

Makes a small SMP folder. My 631 does about 750PPD. Though you are always better off with wired to me.
Good luck.

And get to folding.

[v00d00]

If you know what the wireless network card is (make, model, etc), could you post it on this thread or pm me it, and i'll see if there is anything available outside NDIS (which is a POS to use, native is generally way better).

[jheil]

I've had it working with the wireless for a week now.

I used a automated installer some guy made that sets up finstall and everything. It installs two clients on because hyperthreading makes it think it has two CPUs.

I'm getting projects worth around 250-350 points on the regular linux clients. Running two clients I'm getting 1% done every 40 mins to an hour depending on the projects running.

I run it around 12 hrs a day, it might finish an SMP unit if I ran it 24 hrs a day, but at the current up time it wouldn't.

[bruce]

I've had it working with the wireless for a week now.

I used a automated installer some guy made that sets up finstall and everything. It installs two clients on because hyperthreading makes it think it has two CPUs.

I'm getting projects worth around 250-350 points on the regular linux clients. Running two clients I'm getting 1% done every 40 mins to an hour depending on the projects running.

I run it around 12 hrs a day, it might finish an SMP unit if I ran it 24 hrs a day, but at the current up time it wouldn't.

I've got a 2.8 GHz P4. The software will allow it to run SMP, but it's nowhere near fast enough. The best choices are (A) disable HT in the bios or (B) Run uniprocessor client(s).

Running a single client will turn in WU faster, which is better for Science. Running two clients earns a few more points but each WU takes almost twice as long since you still have only one piece of hardware that runs floating point / SSE, not two.

[alancabler]

I've got a 2.8 GHz P4. The software will allow it to run SMP, but it's nowhere near fast enough.

Northwood P4s will definitely not complete an SMP WU within deadlines- as a result of their small 512K cache- while running at any stable clock speed... there might be some OC'd extreme-cooled machine that would disprove this, but even then, it would be close.
Socket 478 Prescott P4s with 1M cache will complete an SMP WU, but only if dedicated and running at an "advanced" clock speed. 3.0 GHz won't do. Socket 775 P4s w/2MB cache will have an easier time of it, but look to clocks> 3.5GHz.

Of course, running a capable P4s as an SMP machine may earn you a philosophical argument from some quarters that SMP should be run on only multi-core boxes (even though a fast Prescott is faster than some of the slower dual-core SMP machines).

[7im]

Here's to having a points system adjustment to help curb the behavior.

Here's to having the CPU detection in the v6 client improved to resolve the behavior.