Anyone out there Folding@Home?

Got a biggie project in. Delta time between % points is about 50 minutes.
So it took over 4 days !!!

 

My AMD 5000 at stock speeds is taking about 45 minutes for each percentage point in Project 2665 - that's about 30% slower than the Project 2653 workunits I was getting for more than a year, but I've heard that project 2665 is more typical, where 2653 was 'generous' compared to the benchmark used for determining points.

But I remember your machines, even an AMD3800 but overclocked, were all faster than my stock 5000.

I'm a little surprised you're taking 50 miinutes for each percentage - check to see if you're using 100% of your cores. If you're not, switch to the DEINO version instead of the MPICH standard SPM version. I started with the MPICH on this new release and switched to DEINO almost right away and got back up to the speeds I was getting with the older 5.91 SMP version.

But at least you're getting SMP work units worth 1920 points instead of single-core work units worth 47 or 82 points.

BTW, I'm running two copies of SMP on my Intel Quad-core 6600 - and they only take about 23 miinutes per frame. The larger L2 cache on the INTEL compared to the AMD makes a huge difference.

 

New SMP 6.23 executable software is available here: Folding Forum • View topic - new 6.23 SMP public beta clients--improved EUE handling

I've been running this version with Windows Vista 32 bit for three clients for over two weeks, and it really DOES improve Early Unit End handling. While I've only had one EUE in that time, this version was able to send the partial results and download a different work unit and resume without any 'extra steps' on my part.

Most of my EUE's since switching to the 6.22 version have actually been caused by the Windows A1 core that does the folding; Pande Group is working a new core, A2, but they haven't been able to release a Windows version of it yet (a Linux version is out and is reported to be faster than the A1). Once they can release the A2 core for Windows, we won't have to do anything to start using it - it will download automatically as soon as we get a WU that uses it.


If you are already running an SMP 6.22 with -smp, just download the new executable and replace your current executable with it.

 

Quick question for you guys... I can't seem to get a response over on the fah board. I'm leaving my job and won't have access to one of the computers that I use. I want to turn in the results and stop getting new WU's for that computer. I tried the -pause flag but it went ahead and downloaded another WU. So the next time I tried the -oneunit flag and it did the same thing. Can anyone help?

does this info matter.... the server that I need to send the results to is in REJECT mode. (I can see why it won't send the results right now.. but why does it insist on getting a new WU). Also I am using a sneaker-net on this computer. So I shut down the console on the folding computer.. then restart it (with the flag) on the laptop to send the results. Does that matter?

thanks alot. If I can't get it to work I guess there will be 1 or 2 unfinished WU's... not a huge deal.
-Peter

 

You need the -oneunit flag - folding will run until it completes the current work unit, and then it will close itself. I assume you know how to add a flag to your shortcut for running, since you talked about using the - pause flag.

 

Meh! A cheapie video card (30$), the nVidia 8400 GT, downloaded the CUDA driver from nVidia web site.

Then loaded the GPU version for nVidia in the high performance section of Folding @ Home.

Overclocking the card by 10Mhz, which produces 480 points per day (PPD) !

The CPU's are taking 1-2% crunching these 480 points, IOW, no system interference. This is a wonderful find.

Lots of computers, especially in workplaces, have cards than can run the GPU folding, and if all you do is Windows, Internet, Email, Word, Excel, the video card has lots of spare cycles to crunch.

I also got my son an 8800GT (open box) from eBay, so I'll crunch on that next week. Paid 110$ - consider it costs over 300$ last year.

He currently has an 8500 GT, which I'll stick in my other office computer.

I don't take kindly getting kicked out of spot #5 on the team !!!!!!!!!!!!!!

 

Oh, on my laptop, which is turned off twice a day while in transit, I had non-stop problems after crunching 1 unit with Deino.

Switched to MPI, it seems to recover better. 1920 points every 2-3 days!

A lowly P4 2G single cpu, drudges through 2-3 days, to only produce 84 points.

The new nVidia card that came out this week, at 2500$, has like 240 cores...
Wattage-to-points on such a machine must be awesome.

 

This may be old, but considering how cheap these cards are sold now, sometimes used, can be a folding godsend. I got a used 8800GT for 110$. I saw for <100$ some 9600 GT's on eBay.



Running GPU only for points means the computer can produce more points and use less wattage than running CPU based with SMP.
Or run both - if the computer contributes useful heat.

 

I have no objection at all to someone passing my #3 spot on the team- I'm folding steadily and not intending to do any upgrades in the near future, even though graphics cards rack up the points and are getting cheaper and cheaper.

 

You should see me in your stats update within a short time. Running 3 computers.:rockon:

 

There's a RUSH to the #5 spot on the team !!!

I'm seeing multiple varieties of nVidia 8800GT cards on the market below 100$, there is a rush from the gamers to get the newer 9000 series.

My local Kijiji has about five people selling theirs. I haven't yet installed mine, picking it up wednesday, so in two days from now, should start seeing a spike in PPD.
#5 here I come.

 

Back at #5 !!!

In spite of having three (3!) clients break down in the upper 90's and 100%...all worth 1920 points each.

Well, the nVidia 8800GT is crunching wonders. :cheer2:
That single GPU client is more than 50% of my entire production with 4 CPU client computers.

 

Losing # 5 :sniff: !

OTOH, way to go Trevor NYC ! Whizzing by me to the #2 spot...

 

It's getting dramatic!

Now, how do we get some other new people onto our team and, more importantly KEEP then on our team?

There are a lot of members who stay folding - some for the entire two years the team has existed (waving to Wildkow, Daveinolywa, et al) - some who get 'hooked' enough on the folding that they buy more computer equipment to get more points/more science, some who set it up and let run in 'service' mode for the most part unattended, and some who try it and leave.

Is there somethihng we could do here in Priuschat to attract more people to try Folding@Home, and then to keep them on the team? Ideas, please!

 

First, I'm having a blast doing it, and am so happy that this thread existed as I had not previously heard of the project.

My initial thought is to send PMs to those people who were folding but no longer are and encourage them to restart. Since those folks took the effort to be involved at one point, it's worth checking in and pointing out that the research is really bearing fruit and their continued assistance would be appreciated.

My second thought is to let people know in this thread (and maybe other FHOP threads?) about the research that is coming out of the project. Using your excess processor capacity for a good cause seems like such an easy and painless way (well, apart from checking the stats pages!) to make a positive contribution - especially at a time when many people find themselves having to cut back on other charitable donations.

 

Advertising FAH in your signature helps.

 

Go with the progeny of the original:

BOINC

Berkeley invented distributed computing with its SETI@home, and IMO they still do it best.

 

cool.
[05:40:34] Protein: Calmodulin in water

 

Thanks. I hadn't been paying attention to my clients for a while, so the sudden burst was just making sure that my boxes were actually folding. One slip from my and you're back in #5. We'll both be in the million club next year!

 

:welcome:You're both more than 75% of the way there!:cheer2: