Yeah, I stopped the GPU client and now my SMP client shot back up to 5k ppd. Seems the GPU leak is still pretty bad. I guess I'll wait til the 11.4 drivers are released so I can follow that other optimization guide. Not wanting to run beta drivers on my desktop.
So, I got a 6970 yesterday and installed drivers etc. Definitely a lot of processor leak to work on and the a16 core is NOT designed to use environment variables to improve performance (my computer froze up after I tried to put some in). In theory, the a16 core would not have noticed them (as the a11 core is all that should be affected), but that was not my result. Sometime next week I'm going to have to rebuild my server box and put by Nvidia cards back in action.
They definitely have a lot of stuff to work out for Radeon GPUs still. Not being able to play some games or watch a video while folding is very discouraging, in addition to the still-extant large processor leak. Occupying an entire core compared to <1% utilization for Nvidia is sad.
Beta 5 dropped on the 10th for those who are using it or want to see what the next interface will look like
"Some of the highlights of this beta are log filtering, significant Linux and OSX package improvements and improved networking. This release brings us very close to the next phase where we will start to recommend the v7 client on the http://folding.stanford.edu/ front page. Only a few issues remain before we commit to this next step. The most important of which is improved ETA/PPD calculation, which I am working on right now."
Source and download (remember, this is available for Windows, OSX, Mint/Debian/Ubuntu, and Red Hat/Fedora/CentOS)
If your current setup is working without issues, there is no advantage points wise. The advantage for new users will be a highly automated setup process (i.e. none of the command line stuff I asked you to do), with automatic detection of GFX cards and settings. There is a segmentation of what is available to change depending on the "expertise" level you set, making it harder to mess up settings via obscurity.
In theory, monitoring can be done within the FAHControl GUI of remote clients, but I still vastly prefer the HFM.NET interface at this time for that functionality (no PPD total for all monitored machines is one example, better PPD calculation on HFM is another). HFM will be updated to work with v7 clients, but I have not seen any beta version of this thus far (I just checked their website and it indicated that they are working on it, with active changes in the last 48 hours towards this goal, but have not released anything publicly)
just installed v7 beta on the i7 machine for shits and giggles and I can now run SMP on that machine because it makes it easy to limit the number of cores being used therefore bringing my temps into reasonable range until I can get a new HSF
just installed v7 beta on the i7 machine for shits and giggles and I can now run SMP on that machine because it makes it easy to limit the number of cores being used therefore bringing my temps into reasonable range until I can get a new HSF
For future reference, the cmd line way of limiting cores is to change the "advanced options" line from "-smp -verbosity 9" to "-smp # -verbosity 9" (which I believe works with odd numbers, but I have never set that up personally)
This i7 950 is blowing through the first unit in the queue with only 3 cores set to fold, can't wait to see what is will be once I can run full bore and not be in the mid 80s C or higher range. Dumb stock cooler.
I get a corrupted client box when I fold w/ V7 on my GPU. If I just fold with the SMP alone on the CPU everything is fine. If I fold with both it will run fine all night but sometime during the day the F@h display box with get corrupt with little boxes in place of the text. Both clients at that point will start turning EUE.
Anyone else have this issue ...should I try the V6 client?
What GPU? I didn't have any issues other than seemingly slower response times to turn it off and not liking the way you did it (right-click menus vs buttons in FAH GPU Tracker). You can either just fold SMP on v7 and GPUs using systray client or something like FAH GPU Tracker (what I'm currently using on two machines).
@csimon I've been using v7 to fold on my GPU, haven't had any issues.
Idunno what's the deal *shrugs*. I've been using it for SMP but when I combine it with GPU it craps. I haven't tried GPU alone yet. I may try that next. Maybe the case interior is overheating or something ...who knows. It gets wicked hot in there even with all of the ventilation I have.
It certainly might be artifacting due to heat, but try this. Similar interface and equally easy to setup. I like having the one client that runs everything rather than multiple things hiding in systray.
It will set most of the stuff for you, but for a 570, you just need to make sure it is indicated as fermi and you can setup big packet size and turn advanced on. If you get EUEs, try turning off advanced, then changing packet size (much less likely unless there is actually a problem with your card). You could also do some temperature logging and use either built-in fan profile or something like MSI Afterburner to make your cooling more aggressive. I currently have one 180 (maybe smaller) in front, a 200 on top, 120 on back, and another 120 in the case just blowing air directly over the two cards I have folding. Massive temp drop from the directed air.
I folded only GPU last night and it went well but that is typical. The tell tale is during the day while I am at work ...it's usually not corrupt until I get home. This inclines me to believe it's a thermal issue. But I'm wondering if it's the GPU or the memory. The gpu fan only turns at 50% when folding and the temps are around 80c. The case temps get as hot as 45c when both cpu & gpu fold.
So what I'm thinking is that if may try moving an internal direct fan from mosfets to ram. It's worth a shot possibly.
At the moment I am running stock clients w/ no flags set ...no advanced, just to ensure that's not the cause.
If you are letting it go during the day while away, have a fan profile for that time and let it run way higher. I noticed EUEs starting as low as 70c on my 570
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
edited January 2012
No dust in case or on case fans and/or filters for them??? I am hoping not, but feel all bases need covering. Even if this helps only others, I think it is worth a mention.
Note, nights tend to be cooler than days. Cooler room at night with fans on blowing air through case will mean cooler case temp then and better folding.
No dust in case or on case fans and/or filters for them??? I am hoping not, but feel all bases need covering. Even if this helps only others, I think it is worth a mention.
Note, nights tend to be cooler than days. Cooler room at night with fans on blowing air through case will mean cooler case temp then and better folding.
No dust whatsoever. The build is 1 month old and has been cleaned again recently. You're right though ...worth the mention. And yes, nights are quite cooler than days when no one is home.
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
45C is high, I try for an average of 35C max even in Florida, so here is something else unmentioned -- fan volume is a factor of combined fan blade rotation speed, shape of fan blades (scoop blades can grab and push more air than flat blades if just semi-scoop (best)), and size of fan. I use 4-5 blade fans with ball bearing or liquid suspension and have 5 or so per build down in Florida (and buy cases accoringly).
Another thing, I think it is more likely the GPU cooling or GRAM temp that is bad/excessive(too extreme such that it is heating case). Can you control a case fan that blows on the video card with a conductive type thermal temp sensor placed right on the GPU heatsink?? The sets that do this are cheap now that enthusiasts have a lot of liquid cooling available economically more (volume production always lowers cost per unit.
Don't mind me if you've checked and/or tried all this, I tend to worry a system problem until it is fixed as cheaply as possible Been doiing this for 30+ years, so know environment factors play big into temp.
Comments
"Some of the highlights of this beta are log filtering, significant Linux and OSX package improvements and improved networking. This release brings us very close to the next phase where we will start to recommend the v7 client on the http://folding.stanford.edu/ front page. Only a few issues remain before we commit to this next step. The most important of which is improved ETA/PPD calculation, which I am working on right now."
Source and download (remember, this is available for Windows, OSX, Mint/Debian/Ubuntu, and Red Hat/Fedora/CentOS)
In theory, monitoring can be done within the FAHControl GUI of remote clients, but I still vastly prefer the HFM.NET interface at this time for that functionality (no PPD total for all monitored machines is one example, better PPD calculation on HFM is another). HFM will be updated to work with v7 clients, but I have not seen any beta version of this thus far (I just checked their website and it indicated that they are working on it, with active changes in the last 48 hours towards this goal, but have not released anything publicly)
Anyone else have this issue ...should I try the V6 client?
What's FAH Tracker?
It will set most of the stuff for you, but for a 570, you just need to make sure it is indicated as fermi and you can setup big packet size and turn advanced on. If you get EUEs, try turning off advanced, then changing packet size (much less likely unless there is actually a problem with your card). You could also do some temperature logging and use either built-in fan profile or something like MSI Afterburner to make your cooling more aggressive. I currently have one 180 (maybe smaller) in front, a 200 on top, 120 on back, and another 120 in the case just blowing air directly over the two cards I have folding. Massive temp drop from the directed air.
So what I'm thinking is that if may try moving an internal direct fan from mosfets to ram. It's worth a shot possibly.
At the moment I am running stock clients w/ no flags set ...no advanced, just to ensure that's not the cause.
Note, nights tend to be cooler than days. Cooler room at night with fans on blowing air through case will mean cooler case temp then and better folding.
Another thing, I think it is more likely the GPU cooling or GRAM temp that is bad/excessive(too extreme such that it is heating case). Can you control a case fan that blows on the video card with a conductive type thermal temp sensor placed right on the GPU heatsink?? The sets that do this are cheap now that enthusiasts have a lot of liquid cooling available economically more (volume production always lowers cost per unit.
Don't mind me if you've checked and/or tried all this, I tend to worry a system problem until it is fixed as cheaply as possible