I don't know. The literature specifies a quad is as high as SLI goes, but I do not know if that is 4 cards or 4 GPU's. That being said, folding does not care about the SLI status.
Folding@Home
- Thread starter The JoJo
- Start date
I don't know. The literature specifies a quad is as high as SLI goes, but I do not know if that is 4 cards or 4 GPU's. That being said, folding does not care about the SLI status.
ddrueding
Fixture
Thanks. And that was a typo...they are GTX690s (basically a pair of 680s each) so two cards is 4 GPUs. I don't think I'm into this enough to drop another $1k just for folding...
CougTek
Hairy Aussie
Apparently, the secret sauce doesn't taste so good, or so do your stats say. Did they simply stop distributing beta GPU units or do you have problems with the core updating itself automatically?
CougTek
Hairy Aussie
I'm doing a unit 8057 on my GTX 570 and I'm only using FAH Core 2.25, not the 2.22. So I guess I won't need to change the file and play with the hex editor.
CougTek
Hairy Aussie
My client hangs when I enable GPU folding on my Windows 8 system. It worked fine when I was only using the SMP client and it also worked for about a week with the GPU client, but now it no longer does. It stays at the updating phase after a few hours of work and cannot be restarted unless I remove the FAH software completely and then reinstall it. The card is brand new and the power supply is a 850W unit with the 80Plus Gold certification. I would be very surprised if it can't feed the card.
I'll work on this issue this week-end, but for now, no more GPU folding.
I'll work on this issue this week-end, but for now, no more GPU folding.
The beta server VPS12 has been having fits recently and has been regularly going down. When that is happening the transfer to the normal work servers has not been smooth. It has been issuing GPU2 WU's that cannot be completed by Nvidia GPU's. The symptom is that the client gets stuck repeatedly trying to download FAHCore_11 and failing. The solution is to delete the WU rather than reinstalling the FAH software and hope that you get a normal WU. Is that the problem you are seeing?
To delete a WU: pause the client. Remember the work queue number for your GPU. Then go to the work folder (Start->All programs->FAHClient->Data Directory->Work) and delete the folder with the same name as the work queue number. Un-pause the client and it should automatically download another WU and clean up the one you just dumped.
I should not recommend dumping WU's but this technique is also quite convenient for getting rid of those pesky low scoring p76xx or p8018's and 8054's.
I need to see the log (including the system portion) to see what is happening with your Win8 for a proper diagnosis. Folding seems to be quite picky about the Win8 video driver version -- Get the most recent from Nvidia rather than relying on the default Windows drivers (which don't work). Also don't run folding as a service if you are GPU folding. Microsoft, in its infinite wisdom, considers a service that accesses the video subsystem to be a security breach and prevents it.
To delete a WU: pause the client. Remember the work queue number for your GPU. Then go to the work folder (Start->All programs->FAHClient->Data Directory->Work) and delete the folder with the same name as the work queue number. Un-pause the client and it should automatically download another WU and clean up the one you just dumped.
I should not recommend dumping WU's but this technique is also quite convenient for getting rid of those pesky low scoring p76xx or p8018's and 8054's.
I need to see the log (including the system portion) to see what is happening with your Win8 for a proper diagnosis. Folding seems to be quite picky about the Win8 video driver version -- Get the most recent from Nvidia rather than relying on the default Windows drivers (which don't work). Also don't run folding as a service if you are GPU folding. Microsoft, in its infinite wisdom, considers a service that accesses the video subsystem to be a security breach and prevents it.
CougTek
Hairy Aussie
Sounds about like what happens. I'll try to modify the client's settings and delete the GPU work unit next time. I have another GTX570 coming in next week (a GigaByte instead of an EVGA this time). I hope both will work fine together, with the modifications you suggested. I still won't quite match your production, but I'll be close.
CougTek
Hairy Aussie
Looking at the stats, I understand why it's so critical that I increase my production since a team is less than a week from passing us. I'm trully sorry I'm only the 671th most productive F@h user out of 17110 active users. I'll do my best to bring less shame to our team.
CougTek
Hairy Aussie
So now Kepler GPU work for the F@h client? I thought you wrote it didn't. Sdbardwick would be another one we should focus on convincing to re-join the team. Handruin too, if he can manage to configure his client for his GTX 560 Ti.
CougTek
Hairy Aussie
I restarted it 25 minutes ago and it seems fine. Let's hope it'll hold.
CougTek
Hairy Aussie
Oh, another out-of-nowhere unit from Santilli. Always unexpected, always appreciated. It's like finding a dollar bill while walking the street.
Tea is probably due to send a unit this week too.
Tea is probably due to send a unit this week too.
CougTek
Hairy Aussie
And he just did.
mubs
Storage? I am Storage!
Is Tea back in town from her wanderings?
I initially looked into it but gave up due to time and the complication in getting it to work. At this point what do I need to install and configure properly to get the appropriate work units?
CougTek
Hairy Aussie
No one knows about Tea. It would only make sense for Haussies to verify. I think the closest member to Red Hill is LiamC and it's still quite a ride. And LiamC doesn't seem to visit very often these days.
Regarding the GPU client setup, Mark is really the expert. I think I did configure it properly, but I won't know for sure until a batch of beta units is sent my way, which did not happen since I restarted it early yesterday. If I find out it's ok, I'll try to make an illustrated how-to guide.
A non-optimised Fermi GPU like yours will do about 12K to 15K ppd. It's not that bad, just not a spectacular amount either. Way better than nothing.
Regarding the GPU client setup, Mark is really the expert. I think I did configure it properly, but I won't know for sure until a batch of beta units is sent my way, which did not happen since I restarted it early yesterday. If I find out it's ok, I'll try to make an illustrated how-to guide.
A non-optimised Fermi GPU like yours will do about 12K to 15K ppd. It's not that bad, just not a spectacular amount either. Way better than nothing.
ddrueding
Fixture
With only the 4 GPUs folding (CPU OC isn't stable enough) the system is drawing 700W. I imagine that is some points in there.
WUs currently in the box are: 2x 8054, 8018, and 8057. Based on the "Credit", it looks like the 8057 is the place to be? 16520 credits in about 4 hours?
WUs currently in the box are: 2x 8054, 8018, and 8057. Based on the "Credit", it looks like the 8057 is the place to be? 16520 credits in about 4 hours?
P8057's are the only WU's that are testing the new GPU bonus points. 16520 in 4 hours is no better than a single GTX 460's. I would have expected more than 2x that. P.S. The PPD (lower left) is probably better for comparison purposes than credit column
ddrueding
Fixture
Ah, PPD is currently 171613.
ddrueding
Fixture
Now over 275k PPD and climbing. Seems I was having some cooling issues causing one of the cards to throttle. Really should plumb though the wall and watercool with some massive passive radiator, the office is getting a bit warm.
I'll go talk to the facilities guy
I'll go talk to the facilities guy
That 275K Total PPD is a good solid number but I was expecting much more: Like 900K PPD peak. Your previous 16500 credit number with one slot would be a little less than 100K PPD. Your computer has 4 GPU slots each running a different WU. I do not know if that 275K PPD is coming from a single slot or three slots running p8057's. If it is one then you are doing better than I expected and if it is three slots then you are doing much worse. That is my quandary concerning your number.
My earlier expectations for your 2x GTX 690's was that each slot that is running a P8057 should be generating 200K+ PPD each for a maximum total of around 900K PPD.
As a comparison: A stock 580's are getting 225K PPD per slot. You have 2x GTX 690's which is equivalent to four 680's. Hence my estimate of 4x 225K PPD for a total of 900K PPD A single GTX 460 (OC'ed to 825MHz) on my machines produces 100K PPD.
Yes, adding 700W is a significant heater if you don't have an independent thermostat for the room.
My earlier expectations for your 2x GTX 690's was that each slot that is running a P8057 should be generating 200K+ PPD each for a maximum total of around 900K PPD.
As a comparison: A stock 580's are getting 225K PPD per slot. You have 2x GTX 690's which is equivalent to four 680's. Hence my estimate of 4x 225K PPD for a total of 900K PPD A single GTX 460 (OC'ed to 825MHz) on my machines produces 100K PPD.
Yes, adding 700W is a significant heater if you don't have an independent thermostat for the room.
Last edited:
CougTek
Hairy Aussie
Indeed, comparing what I did get last time I was lucky enough to receive a 8057 unit with the one I've just received, there's a 30% improvement with the core 2.22 versus the 2.25 I used last time.
BTW, Gigabyte also offers a huge discount on its GTX570 these days, not just EVGA. ~195$ shipped for the V570OC-13I at amazon.ca. Can't let this one pass. Make sure your enclosure has enough space for it though as it is quite a bit longer than the EVGA, thanks to the improved heatsink. It barely fits in my Cooler Master HAF912.
Last edited:
ddrueding
Fixture
You're going to have to get geared up to hold me off this time, Coug. *
Just turned the setting to "Prefer Maximum Performance". The hottest GPU is 88C-90C at this point. Not sure when they throttle?
Currently three of the GPUs are on 8057s (the other is an 8054) and the PPD is at 414313. I don't think we'll get reliable numbers until it chews on them for a few days.
*Just turned the setting to "Prefer Maximum Performance". The hottest GPU is 88C-90C at this point. Not sure when they throttle?
Currently three of the GPUs are on 8057s (the other is an 8054) and the PPD is at 414313. I don't think we'll get reliable numbers until it chews on them for a few days.
CougTek
Hairy Aussie
I'm gearing up slowly, but surely. Only one of the two GTX570 is doing a 8057 unit. The other one is doing a 8044, which is worth much less. Together with the SMP unit that's also being processed, that system is supposed to do ~160000ppd. add another three 4c/8t Sandy and Ivy Bridge doing SMP units and I must be doing between 200K and 250K ppd currently.
I'd need to get two more GTX570 to keep you at bay. I'll have to assemble another box for that. I currently only have the enclosure, the motherboard, the RAM, heatsink and PSU. I'm sure I can free a small SATA hard drive, but I'll still need to buy an expensive LGA1366 CPU and at least one GTX570. It won't happen before Christmas. My 2.5 million points advance won't last that long. I'll probably bo third again in about two weeks.
Unless I get a few customer calls and ends up with a few extra bucks...
I'd need to get two more GTX570 to keep you at bay. I'll have to assemble another box for that. I currently only have the enclosure, the motherboard, the RAM, heatsink and PSU. I'm sure I can free a small SATA hard drive, but I'll still need to buy an expensive LGA1366 CPU and at least one GTX570. It won't happen before Christmas. My 2.5 million points advance won't last that long. I'll probably bo third again in about two weeks.
Unless I get a few customer calls and ends up with a few extra bucks...
CougTek
Hairy Aussie
F-word. The discount is gone. It last all of last week, but it disappeared less than an hour after I post about it. Now it is up to ~365$. This is probably the best GTX570 on the market. Too bad.
You're going to have to get geared up to hold me off this time, Coug.
Just turned the setting to "Prefer Maximum Performance". The hottest GPU is 88C-90C at this point. Not sure when they throttle?
Currently three of the GPUs are on 8057s (the other is an 8054) and the PPD is at 414313. I don't think we'll get reliable numbers until it chews on them for a few days.
It generally shouldn't take more than 5-6 frames to get an accurate PPD as long as it is in continious operation (pausing or stopping will initially mess up the calcs but they should self correct within 30 minutes).
The prefer maximum performance disabled is the forcing the driver to determine if the video card is playing a 3-D game and if so then it ramps up the clocks from a 2-D clock rate to a 3-D clock rate. Folding does not set off the 3-D game ramp up. If the setting is enabled then the driver maintains 3-D clock rates at all times.
At 3 GPU's getting aprox. 400K + 1 GPU producing 14K you are getting significantly lower PPD then I thought you would. That being said it is still a lot of points.
See: http://foldingforum.org/viewtopic.php?f=66&t=22808 for different PPD's for different cards.
Personally I'd look to Ebay or the like for used with return rights.
CougTek
Hairy Aussie
We have a clown who went Handruin hunting.
What?
CougTek
Hairy Aussie
Bozo resumed his participation and he's close to you. About 200K behind you I think.
ddrueding
Fixture
PPD is back down to 174k. I'm not seeing very many 8057 units at all.
p8057's come and they go, typically in streaks. I was quite annoyed when after the first few days of continuous p8057's, I had an 11 day dry streak while others were busy sending in reports. Now I'm getting them intermittently.
CougTek
Hairy Aussie
Congratulations to Bozo for passing the 5 million points milestone.
Bozo
Storage? I am Storage!
Thanks! :-D