Puppy Linux Discussion Forum

lakedude · Joined: 21 Oct 2008 Posts: 69

Thanks! Works perfect with 4.3.1

I was wondering if anybody uses the high performance GPU client with Puppy?

My skill level is not so hot, the PET packages have me spoiled. Don't know how to do anything the hard way cause so far I've always been able to find a PET that works.

Anybody know where to get a GPU PET?

or how to set up a GPU client the hard way?

There is help for Ubuntu and Arch but I was thinking Puppy would be awesome since Puppy is so fast.

TIA!

starhawk · Posted: Sat 12 Feb 2011, 15:08 Post subject:

Does anyone still have this *.pet? The original link is way broken.

DPUP5520 · Joined: 16 Feb 2011 Posts: 758

Seconding that request here, tried compiling my own pet for FAH and failed miserably and can't find original pet anywhere.

maxpro4u · Posted: Tue 01 Mar 2011, 19:24 Post subject:

http://cid-92fde94f233d97a1.skydrive.live.com/self.aspx/WHERE%20MINE%20WENT/boinc-6.2.15.pet

found this one

DPUP5520 · Joined: 16 Feb 2011 Posts: 758

Nice, Berkeley's Version of FAH thanks alot

maxpro4u · Posted: Wed 02 Mar 2011, 20:03 Post subject:

DPUP5520 · Joined: 16 Feb 2011 Posts: 758

Found it, don't know if it's the right one but here's what I found haven't tried it yet

Sunny · Joined: 20 Oct 2009 Posts: 12

Just set up an old computer (Pentium D) with FAH SMP and I'm about 50% through my very first Work Unit. (I used Team Puppy Linux - User Sunny).

Frugal install of FatDog-521 on a USB Stick. That fired right up and is running very stable on my machine -- nice job on that distribution guys. I've been using Puppy since version 2 ... this is the first 64-bit puppy I've tried.

I put the following in rc.local to help minimize power consumption:
swapoff -a
hdparm -S 24 /dev/sda
echo "1" > /sys/devices/system/cpu/cpufreq/ondemand/ignore_nice_load

The first two lines are used to put the hard drive into standby. The third line is used to throttle back the cpu so that FAH does not run it at 100% ... those Pentium D processors run really hot.

I run this computer headless using x11vnc-server & have turned off the "Save Session" in Puppy Event Manager.

I installed FAH using the instructions on the site for Linux SMP. I had to compile Curl, but that went without a hitch. (Note that I had to use the fd64-32bit-libs-6.sfs & fd64-devx_521.sfs files to do the compile.)

I would try the FAH GPU program using Wine, but I've only got a 7600gt card in this machine.

Anyway, FatDog-64 & the FAH program have been running really stable for about a day now as I been tweaking it.

Haut_Dawg · Joined: 25 Nov 2012 Posts: 27 Location: Portland OR

My name is Haut_Dawg. I just joined the team. Just wanted to introduce myself.
Cheers!

Haut_Dawg · Joined: 25 Nov 2012 Posts: 27 Location: Portland OR

My First Certificate!! Woo Hoo!!

DPUP5520 · Joined: 16 Feb 2011 Posts: 758

Decided to contribute 2 of my machines to the cause , an old emachines T1090 900mhz celeron and a server I don't use anymore running an 8320 eight-core AMD and Nvidia GTX 650 TI(Kelper) with all three running projects the two machines are getting about 17000 PPD.

Sunny · Joined: 20 Oct 2009 Posts: 12

I acquired an old Asus en9600gt Silent graphics card and decided to install it in the PC I was using for SMP Folding with FatDog-521. This is the procedure I used in case anyone else is interested in doing GPU folding with FatDog. Note that the SMP client runs in linux, but the GPU client must be run from FatDog using Wine.

I followed much of the guide provided here:
http://linuxfah.info/mediawiki/index.php?title=Main_Page

Step 1) I started with a clean frugal installation of FatDog-521. I installed the nvidia-280.13-for-520-final.pet for my 64-bit nvidia driver. I also loaded the fd64-dev_521.sfs and fd64-32bit-libs-6.sfs packages so I could compile a couple of programs.

Step 2) I downloaded the cudatoolkit_3.0_linux_32_rhel4.8.run program from nivida. This is a 32-bit cuda tool kit package which is needed as indicated in the reference guide. I quit the X-server and executed this program using the sh command. When this is done, it provides a couple of PATH entries that I added to /etc/profile for PATH & LD_LIBRARY_PATH.

Step 3) This step was optional. I wanted to use the FahMon tool and this pkg requires wxGTK. There is a pet package available for wxwidgets that may work, but I decided to compile wxGTK myself. I downloaded the source for wxGTK-2.8.12 and compiled/installed it using "./configure --prefix=/usr --enable-unicode --enable-shared". I then downloaded the source for FahMon-2.3.99.3 and compiled/installed it using "./configure --prefix=/usr"

Step 4) This step was optional. At this point my puppy save file is pretty big because the nvidia driver and cuda toolkit are huge ... so I decided to remaster my FatDog CD. I unloaded all sfs packages and used the "Remaster Puppy live-CD" program to do a quick remaster.

Step 5) I booted my remastered FatDog which now includes the nvidia 64-bit driver, the 32-bit cuda toolkit, the wxGTK package and FahMon. At this point, it is necessary to load the fd64-32bit-libs-6.sfs and wine-1.3.25-i486.sfs packages at bootup. In order to setup the wine directories, I ran the "wine notepad" command in a terminal. Using the reference guide, I then downloaded and saved the cuda wrappers cudart.dll.so & cufft.dll.so into the ~/.wine/drive_c/windows/system32 directory. Next, I created the symbolic links required as shown in the reference guide.

Step 6) Finally, I installed the Folding@home-Win32-GPU_XP-641.zip console client from the Stanford site. I extracted this and removed the dll files from the directory. I execute this program with the "nice -n 19 wine Folding@home-Win32-GPU.exe -forcegpu nvidia_g80" command.

I am currently running this FatDog from a USB stick with frugal install. I turn off the swap and shut down the hard drive using a "hdparm -S" command. I used nvidia-settings to overclock my GPU and the SMP/GPU clients run concurrently. With an old Pentium D and 9600gt, I can get around 4000 ppd. CPU and GPU both run under 55 degrees C.