need help with setting up offline folding - need smp! got 64 x7560 cores sat idle :-)

n31l

Weaksauce
Joined
May 6, 2011
Messages
108
Found this monster to play with so thought I'd do folding for team 33 while I can but dont have internet access on the machine - access is via VPN :mad:

lotsofidlethreads.jpg


I've downloaded the FAH GPU client, started the SMP client let it download the WU and start on my laptop then killed the client - copied the SMP folder to the remote machine, client runs and looks like it's working but doesn't kick the cores into life :-(

any suggestions?
 
Also, do you have a proxy you can use to access the internet from your VPN?
 
that thing is a beast... would that now be the highest-threadcount single rig on [H] if we get it up?
 
Hmmm, but running windows client might be an issue - doesn't the 32-bit client run out of memory and crash nearing the 100% mark on a bigadv on 60-64 cores... safield noted this somewhere...
 
Hmmm, but running windows client might be an issue - doesn't the 32-bit client run out of memory and crash nearing the 100% mark on a bigadv on 60-64 cores... safield noted this somewhere...
ahh, very good catch. this is likely why it isn't running.
 
I think over 32 cores or something you have to manually specify smp 64 or something too. - 10e had something to say about it.
 
Morning, cheers for the replies, I connect into a corporate VPN via my laptop and then remote desktop into server, not sure if I can pass internet connection through - I'll try reading up on it a bit more.

I'll manually install the client as recommended and see how things go from there.

I just missed out on folding on this 8 Tesla beasty, maybe next time :)

1teslaor8sir.jpg
 
o_O what's that tesla box going to be used for now that it's (presumably) gone into a production environment?
 
The Tesla box was a 'test' build so it got dissasembled afterwards, I just got chance to take photo's - got some more pics if you'd like :)

Tried tweaking the proxy on the machine to use my laptop internet connection and now I've lost connectivity to it - hopefully I can VPN into another machine on the same local network, get back in and reset the gateway - networking isn't my strong point :)

Here's another pic while I'm trying get to get it back online, "why is only 0.98TB useable by windows" ;-)

18tbbutonly098useable.jpg
 
ok I can remote desktop back into the server again now :)

It would be good to get this offline folding working as I should have an 8x8 core AMD box to play with next week, unfortunately it's not Bulldozer, maybe next month ;-)
 
Here's another pic while I'm trying get to get it back online, "why is only 0.98TB useable by windows" ;-)[/IMG]

Your OS should be able to use all 2TB. There's a hotfix that might help; otherwise check your BIOS settings and/or contact your hardware vendor. If you upgraded from 2008, you may need to do a clean OS install. If you've got two different types of RAM, like you added an extra 1TB for testing, that could also be the issue.
 
Your OS should be able to use all 2TB. There's a hotfix

Thanks, It was a fresh 2008 R2 install yesterday, I didn't have any issues installing or starting, I'll hunt around for an alternative hotfix if I manage to get f@h working.

I'll have to 'make do' with 0.98TB for now :D
 
Time for breakfast, I tried installing 'Microsoft Loopback' as I read somewhere it might help get smp working - unfortuntely it wasn't a 'silver bullet' solution.. I'll try again after I've had some 'food' for thought :)
 
So I managed to get 1 thread working on 1 WU - just need to get the other 127 going now :D

got1coreworkingonawu.jpg



** work type 78 not eligible for variable processors - the search continues :-(
 
Last edited:
Sounds like you forgot to use the -smp flag, that's an uni-processor unit. Can you post the contents of FAHlog.txt and embed it here using the
Code:
 tags?
 
ok, got 64 working now - half way there :)

got64workingis128toomuc.jpg



Time for a beer.. while I sit back and reflect on my c64 days :)
 
I'd still like to see your FAHlog.txt file please.
 
I just messed it up trying to stop it from auto connecting, it started updating WU list and wont resume WU now.

I'm going to reboot it, wPrime yesterday ran on all 128 threads then suddenly had a funny few minutes and then would only run 64 - this thing takes 7 minutes just to flash up the initial BIOS screen :eek:

Code:
Launch directory: C:\Users\Administrator\Desktop\FAH GPU Tracker V2\SMP
Executable: C:\Users\Administrator\Desktop\FAH GPU Tracker V2\FAH.exe
Arguments: -oneunit -forceasm -smp 128 -bigadv 

[08:30:38] - Ask before connecting: No
[08:30:38] - User name: n31l (Team 33)
[08:30:38] - User ID: ***me***
[08:30:38] - Machine ID: 2
[08:30:38] 
[08:30:38] Loaded queue successfully.
[08:30:38] 
[08:30:38] + Processing work unit
[08:30:38] Core required: FahCore_a3.exe
[08:30:38] Core found.
[08:30:38] Working on queue slot 01 [July 31 08:30:38 UTC]
[08:30:38] + Working ...
[08:30:39] 
[08:30:39] *------------------------------*
[08:30:39] Folding@Home Gromacs SMP Core
[08:30:39] Version 2.27 (Dec. 15, 2010)
[08:30:39] 
[08:30:39] Preparing to commence simulation
[08:30:39] - Ensuring status. Please wait.

I went back to using FAH GPU Client as when I used the other client I got the "work type 78 not eligible for variable processors" even with -smp. I'll try the normal client again later, I was just happy to see half the cores working.
 
Last edited:
as I mentioned before, you'd be much better off ditching the GPU Tracker and running the client manually. It's pretty easy to install and I gave you the link to the guide a little earlier in this thread.

Also, you'd be better off removing the -bigadv flag for now as it will be a bit easier to trouble shoot.
 
as I mentioned before, you'd be much better off ditching the GPU Tracker and running the client manually. It's pretty easy to install and I gave you the link to the guide a little earlier in this thread.

Also, you'd be better off removing the -bigadv flag for now as it will be a bit easier to trouble shoot.


Ok, I'll try again when it's back.. I installed the client as you suggested but when it ran it gave that earlier error message and the GPU client didn't and I just had to post :)


It's back online, wish me luck ;)
 
I think I might have to settle for only 64 threads - I'm not sure the client can see the other 4 CPU's, HWinfo and CPUz only see 4 processors so I'm guessing the client will have the same issue too..

27408344.jpg



I'll let it finish it's current WU and try uploading it - then I'll try tweaking it a little more :)
 
It's time to put Linux on that bad boy.

but yeah, if the OS is only seeing four CPUs, the client is only going to see four.
 
It's strange as Windows sees all CPU's and threads :confused:

Looks like CPU 0 & 1 need reseating, compared to the others there a bit toasty :)

tempswhilefolding.jpg



I'll try installing ubuntu on Monday if I get a chance, might also try installing a Ubuntu in a VM on it tomorrow - if windows sees all CPU's then maybe ubunut will to?
 
Well Windows is definitely seeing all 8 procs

CPUz and the like were never designed with type of system in mind, so its really no surprise that it cannot see all 8 procs.

My only question is, if the box doesnt have internet, how is it going to upload the WU, when it completes?
You said you are connecting to VPN to a box on the remote network, then RDP from the remote box to the "Monster"....right? If this is the case what OS is the box you are connecting to VPN running and does it have internet access?
 
quick little update, the F@H client sees all 64 cores and threads;

clientsees64cores.jpg


I've just downloaded ubuntu 11.04 server, I'll copy it onto the server, install and see what happens - hopefully I'll have a chance later today but it's time for some food and a beer or 2 :)
 
Last edited:
quick little update, the F@H client sees all 64 cores and threads
That's GPU Tracker doing that detection, GPU Tracker is NOT the F@H Client.. it's a 3rd party application designed mostly for people with GPUs... hence the name.
 
I'd love to see more logfile detail of FAHlog.txt so I can see how gromacs is being launched.

Also, make sure you are using a passkey so you can get bonus points after you upload 10 successful SMP results.
 
Seriously reccomend not using GPUTracker....I know I made that thread here, but its for people with GPUs or that have consumer systems.
 
Back
Top