Atlas Folding

My new purchase.  Summer restoration fun!

[Show as slideshow]

Atlas Folder has quietly slid past the 300,000,000 point mark.

It’s an interesting time for folding at home, what with the resurgence of the SMP folding client on multi-processor machines.

I’ve been peppered with questions about whether Atlas Folder is going to “modernize” and move to CPU/SMP.  The short answer is “Yes”.  I’m in the process now of experimenting with the technology and devising out a plan to accomplish this both logistically and financially.

Fold on folks.

Jason

Yesterday it was 106 degrees Fahrenheit in Tulsa, Oklahoma.

The weatherman promised “only 94″ and rain on Friday which sounds relatively pleasant, but in Tulsa what that means is that it’ll be 94 degress, 100% humidity and intermittent sunshine making a sauna out of the entire county.

The highest temperature I’ve ever experienced personally was 116 at the Hover Dam in August.  Yeah, the humidity was low but at 116 it just doesn’t matter that much.  The lowest temperature I’ve experienced I’m not too sure of but it’s been -15 or so here in Tulsa that I can recall.

So post your weather extreme comments below if you like, I’m curious what other folks are experiencing around the world.

Yesterday Atlas Folder quietly passed mutsu in total points ever contributed to the Folding@Home project.  This event placed Atlas Folder as the #1 private contributor to F@H in total points and #4 in the world in total points behind only:

1. Anonymous PS3s with no team assignment
2. Anonymous CPUs with no team assignment
3. Purdue Data Center (PDC)

Thanks for the well wishes and pats on the back that I’ve been receiving.  I hope to continue to contribute for a long time to come.

Fold on!

Yeah ok, so I woke up this morning to full farm failure.  Sometime during the night the farm’s file server went down.  That machine hosts all of the data for the farm so that effectively sticks a fork in the entire enchilada.

Not sure why it went down other than we had some storms here and perhaps it experienced a brownout.  The farm isn’t battery backed up as it’s not practical; Atlas Folder draws 13,000 Watts of power continuously.

Anyway, drove out to work from home and spent another three hours on it this morning and afternoon.  Seems to be working at the moment but I figure it’ll die about the time I’m locking the door behind me.  Isn’t that always the way it goes?

All through yesterday and today while at work I’ve had the background task of rolling through the 12 separate computers that comprise my portion of AtlasFolder and performing a series of updates:

1. Allowing Windows XP x64 to fully update
2. Installing the very latest nVidia video/CUDA drivers (258.96)
3. Upgrading to Folding@Home’s GPU3 software

I have been reminded of how the farm doesn’t like to be disturbed.  Almost every time I decide to perform maintenance on it it bucks and kicks like an old stubborn mule and several of the machines will bounce up and down for hours or days before they decide to play nice.  Which is what they’ve been doing…  Three of them were rebooted of their own accord when I came to work this morning.

The problems do seem to be stabilizing a bit this afternoon though, so I am hopeful that my production numbers will be back up to normal starting tomorrow.

On a side note, installing GPU3 required an update to all of the DLL files that are housed locally to each client.  Obviously I did not want to perform a bunch of manual copies to all of the 100+ directories so instead I added a couple of lines to my GPURUN.BAT file which is documented here.  It now has an XCOPY line:

@ECHO OFF
TITLE %USERDOMAIN% - GPU %1
@ECHO Changing to directory F:\MACHINES\%USERDOMAIN%\GPU%1
F:
CD F:\MACHINES\%USERDOMAIN%\GPU%1
@ECHO COPYING NEW DLL FILES (IF ANY)
XCOPY /D /Y F:\GPU\*.DLL F:\MACHINES\%USERDOMAIN%\GPU%1
@ECHO STARTING GPU%1 IN %CD%
F:\GPU\GPU.EXE -gpu %1 -verbosity 9 -advmethods -forceasm -forcegpu nvidia_g80 %2 %3 %4 %5 %6

This allows all clients to check for new DLLs on every boot and copy them over as necessary now.

PS: van_arnam next time you start your clients you’re going to get GPU3 by default!