Issues with 6.22 SMP client

TheKnat

TheKnat

TheKnat

Associate
Joined
13 May 2007
Posts
1,832
Location
Lancashire, UK
I upgraded my main PC at home to the new 6.22 beta2 SMP client yesterday as it had recentally finished a WU, so I thought it would be a good time.

I stopped both current SMP clients from running, made sure all their cores were killed off, restarted the computer, deleted all the old files then installed the new client. Everything seemed to go OK and I got both clients up and running again.

Fast forward to this morning and neither client has even done 1% of its processing so I start to get suspicious. I killed off both clients again and started runnign just a single SMP instance and within 30 minutes it had finished off 2% of the core.

Anyone got any idea why this might be happening ? (BTW this is still with 2 GPU clients going as well)
 
TheKnat said:
(BTW this is still with 2 GPU clients going as well)
Click to expand...

This gives us a big clue :) Presumably you're on a quad core, in which case you would be trying to run 2 SMP clients on essentially 2 cores (if you're on XP the GPUs use one core each). I would leave it running with just one SMP client for now and see what your PPD is.

You might want to check the priorities of the clients as well. If you have the two GPU clients on the slightly higher priority (as you should ;)) then the SMP clients will have less CPU to work with.

Hope this helps.
 
I've got the GPU clients running at the higher priority and had the SMP clients assigned to all 4 cores. I changed the single SMP client to be on the 2 free cores this morning and it was saying it would crank out about 1500 PPD, which seems to be about right.

The wierd thing is that before the upgrade, both SMP clients said they would be doing about 1500ppd. Would limiting the SMP client to a single core be an idea or is it best to just leave 1 client running over the 2 cores ?
 
I doubt it would like restricting the client to one core, since its an SMP client :) You can try it though.

When it said both SMP clients were on 1500PPD, did you leave it for an hour or so before you had a look at the PPD? It usually takes that long for the true value to show on Fahmon.

To be honest I'd leave it at 1 SMP 2 GPU for the time being, give it a few days to make sure it is settled and is crunching well, then add the second SMP for a day and see what new output you get.

I tried a similar thing with my setup: I had 3 Linux SMP clients and one GPU client running at 9800PPD. I added a fourth Linux SMP client and initially it was working fine at just over 10k PPD, then it went down to about 8000 for no reason (see my recent output in the last few days ;)). I removed the fourth SMP client this morning and its back to normal :)

Don't try to shove too many clients on if it means that the stability/output is reduced is the moral of the story :)
 
Yes it is. One VMWare instance with 2 clients and one with one client, running on Ubuntu. Only thing you need to make sure of with that is that the VMWare has an idle priority and that the GPU client has a low priority so it gets enough CPU. Its a bit fiddly but gets an increase of points over the Windows SMP client - plus its a non beta client :)
 
Might have a look at that then. Its a bit of a PITA to get all the clients running in windows.

BTW: 2nd most points for this week so far :D
 
You must log in or register to reply here.
Back
Top Bottom