How many PPD from this?

My scientific judgment is: lots

:p

I am running a Q6600 at stock and I do an SMP WU in around 12 hours. 3.5-3.8k PPD

An i7 should do a wee bit better I think!
 
If you use the notfred folding + vmware player about 8K I reckon

So about the same as a GTX260
 
OK I'm about ready to start Folding - I'm game for anything in terms of setup, so what will give me maximum PPD and how do I go about doing it? Am I right in thinking it will be to use the Linux SMP client? In which case how many virtual machines should I have, how many CPUs should I allocate each virtual machine, and how many SMP clients should I run?
 
There is a very good guide here: http://www.overclock.net/overclock-net-folding-home-team/607465-guide-vmware-3-0-8-core.html

On how to setup VMWare to fold the bigadv units with which you can get around 25K PPD using an i7 processor.

Looking at the spec of your machine thats the most you will probably get out of it, its not worth folding on the 5850 currently as the folding clients haven't been optimised to use the new technology in the 4xxx and 5xxx cards so you will just waste CPU cycles.

If you can get the above VMWare working then that will max out your performance.

Alternatively you could get around 8-10K PPD running just the notfred vmware image and 8 cores folding in vmware on the regular Linux client.

Why do I even bother mentioning the notfred folding considering its low PPD compared to the bigadv folding? Well its going to be a LOT less demanding on your hardware and will even probably let you use it while folding in the background. Whereas the bigadv units are better suited to dedicated folding machine as the units usually take 2+ days to complete and the only way to get high points is to complete them as quickly as possible. (e.g. don't use your computer for anything else)

Up to you in the end :)
 
The way I see it, if you use your machine on a daily basis then stick with the regular SMP client. The PPD is obviously not as good, but it is much, much safer.

Two or three days is a long time to fold if anything should happen to the WU. If it bombs out on you that's time lost. If it bombs out because your hardware was pushed too hard, then you're affecting the science.

With the regular SMP client it takes hours, rather than days, and you are much less likely to lose WUs [better for FaH] and will actually be able to use your machine [better for you!]. But, if it is a dedicated folder, then BIGWU away! :D
 
I'll give bigadv option a go. I already have an Ubuntu virtual machine setup using VirtualBox that I use for various pieces of work. I've just reconfigured it with 6gB of memory and allocated it 8 CPUs - I'd rather setup FAH in an existing virtual machine than install VMWare. However this means I can't use all the preconfigured stuff. Is there a decent guide for setting up a bigadv client from scratch? If so, could somebody link me to it? Cheers.
 
dhpiggott@mars:~/folding$ ./fah6 -bigadv -smp 8

Note: Please read the license agreement (fah6 -license). Further
use of this software requires that you have read and accepted this agreement.

8 cores detected


--- Opening Log file [December 11 16:38:49 UTC]


# Linux SMP Console Edition ###################################################
###############################################################################

Folding@Home Client Version 6.24R3

http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: /home/dhpiggott/folding
Executable: ./fah6
Arguments: -bigadv -smp 8

[16:38:49] - Ask before connecting: No
[16:38:49] - User name: piggott (Team 10)
[16:38:49] - User ID: 1880165743FA8DE1
[16:38:49] - Machine ID: 1
[16:38:49]
[16:38:49] Loaded queue successfully.
[16:38:49]
[16:38:49] + Processing work unit
[16:38:49] Core required: FahCore_a2.exe
[16:38:49] Core found.
[16:38:49] Working on queue slot 01 [December 11 16:38:49 UTC]
[16:38:49] + Working ...
[16:38:49]
[16:38:49] *------------------------------*
[16:38:49] Folding@Home Gromacs SMP Core
[16:38:49] Version 2.10 (Sun Aug 30 03:43:28 CEST 2009)
[16:38:49]
[16:38:49] Preparing to commence simulation
[16:38:49] - Ensuring status. Please wait.
[16:38:59] - Looking at optimizations...
[16:38:59] - Working with standard loops on this execution.
[16:38:59] - Files status OK
[16:39:01] - Expanded 4193792 -> 117935139 (decompressed 763.8 percent)
[16:39:02] Called DecompressByteArray: compressed_data_size=4193792 data_size=159270593, decompressed_data_size=117935139 diff=41335454
[16:39:02] - Fatal: Could not decompress work unit data
[16:39:02] Error: Could not open work file
[16:39:02]
[16:39:02] Project: 2064252928 (Run 50331648, Clone 318767104, Gen 385875968)
[16:39:02]
[16:39:02] Error: Could not write local file. Exiting.
[16:39:02] - Shutting down core
Click to expand...
Any ideas what's going wrong here?
 
Does this happen every time?

Also, how are you folding, is it through a virtual machine? If so you might need to drop it to 7 cores. I can't find it now, but I read 8 core bigadv in a VM can cause issues.
 
its saying it cant unzip the download so sounds like something s missing from your linux install
 
Did you use sudo to make the directory for folding, if so that may be the issue. Try creating the directory normally.
 
It wasn't a permissions problem, I created it as dhpiggott. Not entirely sure what the cause was, but deleting everything and starting again seemed to fix it. However I then I found I was missing mpiexec so I had to sudo apt-get install mpi-default-bin. Even then it seems that the fah binary tries to do ./mpiexec and as there was no such file in the folding directory so I had to ln /usr/bin/mpiexec. This got it running OK, but then as soon as it started doing any work the whole virtual machine froze up, presumably due to starvation so I had to reboot it. I then tried -smp 7 rather than -smp 8 but had exactly the same problem.

I decided that even if I could get it working on that virtual machine I'd rather not because I do use it for important work, so earlier today I ended up downloading the VMware image from Overclock that you linked to Biffa. This seems to be working quite nicely.

Am I right in getting the impression that I'm meant to do 10 standard smp units for some reason before doing any -bigadv units? I've got the virtual machine working on a standard SMP unit but if this isn't the case I'll let that one finish and then go over to -bigadv straight away.

I've been a bit slow with getting all this together - it's my last week here at uni for the term so I've been quite busy both with work and play (going home tomorrow morning).
 
Mr. Brightside said:
Am I right in getting the impression that I'm meant to do 10 standard smp units for some reason before doing any -bigadv units? I've got the virtual machine working on a standard SMP unit but if this isn't the case I'll let that one finish and then go over to -bigadv straight away.
Click to expand...
No, you have to complete ten WUs with -bigadv set, even if they're standard SMP WUs, before you start receiving bonus points.
 
I did a fair few standard SMP units, but about a week ago or so ended up buying a Corsair H50 because at full load, stock clock, temperatures were much too high for my liking (above 80c).

The H50 is so good in fact that I ended up stopping Folding on this for a few days and "accidentally" overclocking this machine to 4GHz. Took a while to get right, but as of mid-afternoon yesterday it is 24 hour Prime stable - all while staying reasonably cool - so I'm now running bigadv units on it as I reckon it should turn them around quick enough to be worth doing. First one is on 43% right now.

Going to be watching my PPD closely over the next week or so!
 
You must log in or register to reply here.
Back
Top Bottom