OcUK Folding

Associate
Joined
15 Jun 2008
Posts
143
Just started Folding on my machines and registered for the OcUK team...aaprt from pointing my clients at team 10 is there anything else I need to do?
 
Just started Folding on my machines and registered for the OcUK team...aaprt from pointing my clients at team 10 is there anything else I need to do?

Nope, basically that's it.

It might be worth checking if you are using the optimum clients though.

It's always better to use the CLI clients, rather than the ones with the pretty graphics.

If you have a multi-core processor, then the SMP units give you much better production for your involvement. If you have a supported graphics card, such as an Nvidia 8xxx or AMD 2xxx or better, there are also clients for these that can significantly improve your production.

Welcome to the team and I hope you enjoy your time with us.
 
Thanks...

I'm using the CLI client + 1 GPU client for my 4850.

"SMP units " <--- can you explain this for me please? I guess I need an alternative client for this? Edit: just checked the sticky and it suggests its for *nix boxes? Both of mine are Windows based although I could run a dual boot on one of them if it's worth the effort.
 
Last edited:
OK, I've been reading up about the SMP client and I need help answering a few questions.

1. It seems like it would be best to run on the machine that is also running the GPU client - correct?

2. I should wait for current WUs to complete then uninstall the CLI clients/services?

3. I understand that the GPU needs a CPU core to manage it - so on my quad core I run 3 CLI clients and the GPU client - which gives me 100% CPU utilizaion across all 4 cores. With the SMP client is this managed automatically or do I somehow have to configure it to use only 3 cores to keep one available to manage the GPU client?

Sorry for the Q's, thanks for the help.
 
I'd run one SMP client on both PCs. Without a GPU as well, the X2 should easily be quick enough to complete SMP WUs. The ATI client, as far as I know, usually takes a whole core. All you should need to do on the quad is uninstall the standard clients, install SMP, then set the GPU client to 'slightly higher' priority in the settings screen / config process to make sure that the GPU client always gets the core it needs.

And yep, it's generally best to wait for the current WUs to finish before uninstalling. If you want the client to stop after the current WU and not fetch another, you can run it with the -oneunit flag.
 
Last edited:
Great, thanks. Once my current WU's finish I'll switch to the SMP client on both machines then.

Thanks for the help.
 
It's mildly OC'd and has 2Gb of RAM. It's not a great system and never was as it would never run high FSBs or tight memory timings no matter how many different flavours of RAM I tried with it. Still, it's better than nothing and I guess every WU counts.
 
I think it's all the same as the previous versions? If you're running the systray client, the setting's in the config screen:

http://upload.mattus.co.uk/fahgpu/adv_settings.png
adv_settings.png

If you're running the text-only client, you need to edit your shortcut to add -configonly to the path, and modify it from there.
 
I did the config and got into it txt wise Mattus, but can only enter CPU %, Id rather turn a core off to allow GPU to use a whole core, can this be done? If not what should I put the CPU usage down to?
 
Last edited:
You could always change the core affinity using either task manager or a third party program (i think there is one called FAH affinity changer, although i am not sure...). You could then tell the smp client to only address cores x, y and z, and address the gpu to the other core. If the two clients are fighting one another for cpu time, then you may need to set one to have a slightly higher priority in the config menu (in my experience, the gpu client does not use much cpu power under vista, but it does seem to vary from person to person). In mattus' screenshot above you can see there is the option for idle, or low (at the top). Set one to idle and one to low (gfx to low i imagine would be best), and they should resolve their differences and get along... This is certainly easier than changing core affinity.
 
Last edited:
Back
Top Bottom