Folding Forum

I've been folding in the past with nvidia GPUs and always went the systray route due to the ease of use. Also I've only used a single GPU card in the past as well. I've upgraded some hardware over the past couple years and I am now using an ATI HD5970, I've not really folded with this machine since the card swap. Wasn't sure how easy it would be but I did some reading tonight and found out the ATI cards can use the V7, you guys should update the download page to reflect this...

Ok, so I have installed and running V7 (I believe) on my HD5970. Seems to be doing ok, I'm monitoring it with Fahmon at the moment, if there's something better please let me know. Now the HD5970 is a dual GPU card. I'd like some help setting up F@H to run on both GPUs. Can/will anyone help me do this? I've attempted it earlier with the V6 systray client with unreliable results so I quit and tonight decided to start over...

My issue is that I'm not familiar with the console client at all and I have no idea how to make changes to it, I'm not sure if it is running as a service and if not how to make it run as a service. I'm definitely in the dark when it comes to running on the second GPU on my card.

I have a quad core cpu as well and it seems to be using a lot of CPU cycles as well... I'm under the impression this is not right.

Well that's about it, I know, It's enough. Any help would surely be appreciated.

Welcome to the forum, tdwhite.

First off, GPU clients can't be run as a service due to Windows security restrictions. The best you could do is have it autostart on system boot.

Can I ask why you're running v7 in console mode? Most people opt for the GUI since it was designed to be easy to use.

In v7, to use the other GPU you'd add a second GPU slot. The FAHClient Guide covers how to do that using the GUI; basically, change the client to Advanced or Expert Mode (next to the big yellow About star), then follow the instructions about adding slots. This guide shows how to change the configuration without the GUI.

The AMD core does use a non-negligible amount of CPU time. When running alongside and SMP client, it's recommended to set the GPUs to run at Slightly Higher priority to make sure they're not bottlenecked. Some people choose to run SMP on n-1 cores, leaving one free to handle the GPU. I believe one CPU core is enough to handle 2 AMD GPUs, but I may be wrong.

The download page will be updated once v7 is considered more ready. While it's pretty stable and most of the existing bugs aren't very serious, the Pande Group doesn't feel it's polished enough to recommend to everyone.

Welcome to foldingforum.org, tdwhite.

In addition to the guides mentioned by Zagen30, be sure to check the forum for the V7 FAHClient Open Beta. Granted, most people will not encounter many of the bugs reported there, but it's simply not ready for a general release, hence it's not on the download page yet. If you do encounter problems, that's the place to discuss them.

Also, third party programs such as Fahmon have not yet been updated by their developers so I really doubt that you're monitoring V7 with Fahmon at the moment.

Any estimates as to when it will make it to the High Performance download page?

Zagen30 wrote:Welcome to the forum, tdwhite.

The AMD core does use a non-negligible amount of CPU time. When running alongside and SMP client, it's recommended to set the GPUs to run at Slightly Higher priority to make sure they're not bottlenecked. Some people choose to run SMP on n-1 cores, leaving one free to handle the GPU. I believe one CPU core is enough to handle 2 AMD GPUs, but I may be wrong.

How do I go about setting the cores to only utilize 3 cores? New to this v7.

Ok guys, turns out I'm mistaken, I'm not running V7 I'm running GPU3 from the download page. So can I upgrade to V7 from GPU3 and not lose my work? If so, what is the process?

I'd definitely be interested in V7 if it can run ATI cards in the GUI mode.

Thanks for the help.

Wow, so I went ahead and installed v7. It automatically setup the client to fold with both GPUs. This is nice although it's still causing my CPU to run anywhere from 40% to 75% on all four cores. The CPU is not folding.

So I set the affinity for each of the GPU cores to only use one CPU core. Now each GPU folding client pegs the assigned CPU core to 100% but the other two CPU cores are back to normal/idle.

Is this type of setup ideal? It does make the computer more usable while folding. I don't have a slouch machine either but with affinity set to use all cores, I could barely type in a web browser...

You can check whether the GPU is being starved of stuff in needs from the CPU by checking the sensors on GPU-Z. In the ideal situation the Load should be 100%. If its not see if allocating an additional core lets it reach 100% load.

chilihotorcold wrote:How do I go about setting the cores to only utilize 3 cores? New to this v7.

Welcome to the forum, chilihotorcold.

Follow the instructions in the guide for Editing Options, and create a new one. You'll want the Name to be smp and the Value to be 3.

Spitfire19 wrote:You can check whether the GPU is being starved of stuff in needs from the CPU by checking the sensors on GPU-Z. In the ideal situation the Load should be 100%. If its not see if allocating an additional core lets it reach 100% load.

I'm using MSI afterburner which shows both GPU cores simultaneously. GPU1 is low 90s load and GPU2 is high 90s load.

I can't stand the default MSI afterburner skin and I suggest getting the Aero Ultra-Compact Dark by crow Trobot skin if you plan to use MSI afterburner for anything.

If I'd like to reboot my machine will F@H client lose it's place or will it pick up where it left off before you rebooted?

tdwhite wrote:

Spitfire19 wrote:You can check whether the GPU is being starved of stuff in needs from the CPU by checking the sensors on GPU-Z. In the ideal situation the Load should be 100%. If its not see if allocating an additional core lets it reach 100% load.

I'm using MSI afterburner which shows both GPU cores simultaneously. GPU1 is low 90s load and GPU2 is high 90s load.

I can't stand the default MSI afterburner skin and I suggest getting the Aero Ultra-Compact Dark by crow Trobot skin if you plan to use MSI afterburner for anything.

If I'd like to reboot my machine will F@H client lose it's place or will it pick up where it left off before you rebooted?

F@H should start back up from last checkpoint. (Which by default is within the last 15min.) with most cores.

tdwhite wrote:If I'd like to reboot my machine will F@H client lose it's place or will it pick up where it left off before you rebooted?

It's NOT! proper procedure to pause.

Zagen30 wrote:

tdwhite wrote:If I'd like to reboot my machine will F@H client lose it's place or will it pick up where it left off before you rebooted?

Proper procedure is to hit the big Pause button and wait for the slots to show "Ready" (as opposed to running), as this means they've been stopped properly. Even if you didn't do that they'd almost certainly pick back up at the last checkpoint, but it's safest to pause before reboot.

Gosh you guys are all conflicting. 7im says never to pause before quitting.

Apparently I'm wrong. I must have missed the instructions to not pause, and I just assumed we should, like how it was recommended to quit before rebooting with v6.

No, you don't need to quit before rebooting. We don't normally recommend that either.

It's only a work around on some systems that don't seem to shutdown gracefully.

It's possible to edit the client.cfg file manually, but we don't recommend that either, because it's too often done incorrectly. Run the -config switch instead. That never breaks.

You can over clock your system, and produce more science, but we don't recommend that either. It's too often done without proper testing, causing the clients to corrupt work units and burn through many units potentially emptying work servers.

Beta work units do not get higher points. It's often a much lower PPD, because the WUs error out more frequently.

I should start a whole new forum section to dispell folding myths.

So what is the proper way to shut down v7? just quit or pause then quit? Got confused reading the earlier posts.

Also looking at my work queue, I see that my 2 GPU it don't show how much credit I will receive is that normal?