TPU's F@H Team

[Squirrely]

I shall join when I get my desktop up and running again.

But, will the client only fold when my desktop is idle? Also, will it do it automatically (start folding after 5 mins of a desktop being idle for example)? But if I go to my desktop again, will it stop? (If it says it does, must of missed it )

Or is it pretty much similar to BONIC with seti@home?

 

[sneekypeet]

does fahmon work with the gpu2 client?

yes sir...Infrared walked me through it and got it to function 100%

A squirrely. F@H can start when the PC boots, it will not however turn on and off automatically. If you are folding and tick open a game it will fold while gaming. If you right click in the taskbar you can pause or stop it manually!

 

[dadi_oh]

does fahmon work with the gpu2 client?

Yup. You just need to point FAHMon to the rigth directory. For the GPU2 client it creates a folder in the "My Documents" somewhere so make sure you point to the right place.

 

[infrared]

The gpu2 client stores the files here:

C:\documents and settings\user name\application data\Folding@Home-gpu\

 

[Random Murderer]

thanks guys.

 

[irongun324]

Yes, the multi-core client is more efficient at utilizing multiple cores, and does more advanced work than the single core client.

I'm running the SMP client, but CPU usage is only hitting 50% (near 100% on one core)... does it take a while to kick in or does it only spike above that when needed? seems to fluctuate between 49-51%.

 

[dadi_oh]

I am going to be installing a Q6600 soon on my main rig and was looking for help and advice on a few things related to folding.

Right now I run the GPU2 folding client pretty much fulltime on the 8800GT in this rig. I sometimes run the CPU client (non-SMP) overnight only and leave it off during the day when my family is using the machine (so I don't get any complaints of slow response). With the GPU client running, Core1 loads to 100% and Core0 seems to be left untouched which makes it available for other tasks on the machine. Therefore, with the GPU client running I don't really notice any performance hit in non-3D tasks.

With the replacement of my E8400 dual core with the Q6600 Quad core I was thinking I could get the extra 2 cores folding full-time without really impacting anything (including gaming).

So given this background what would be the best implementation of folding 24/7 that still allows general use of the machine, including gaming?

For example, can I somehow force the affinity on a non-SMP client to core2 and run another instance on Core3, leaving core0 and core1 to deal with the GPU client?

BTW I am most of the way through building a completely dedicated folding rig made up of an E1200 dual core on a Nforce 610 mobo with 2GB RAM and the 8800GT from my main rig (once my 260GTX arrives). I was thinking I would run the GPU2 client on the 8800GT and a non-SMP client on the processor, 24/7.

 

[ShadowFold]

Can I fold with my 4850? Folding with my E7200 @ 3.8 doesn't seem to be going very fast..

And its only making Core0 go up, how do I get it to do multicore?

 

[mrhuggles]

how do i get my gpu to stay in 3d mode so the gpu client doesnt go slow?

 

[dadi_oh]

Can I fold with my 4850? Folding with my E7200 @ 3.8 doesn't seem to be going very fast..

And its only making Core0 go up, how do I get it to do multicore?

You need to use the SMP client. It is in the "High Performance" section rather than the generic download section.

http://folding.stanford.edu/English/DownloadWinOther

And yes, you can fold with the 4850. The GPU client is also in the high performance download section.

 

[ShadowFold]

Think my 4850 would fold better?

 

[dadi_oh]

how do i get my gpu to stay in 3d mode so the gpu client doesnt go slow?

I believe that the GPU client is accessing your shaders directly so there is no need to worry about 2D vs 3D mode. I would think that the video card knows it's shaders are active and switches to the 3D clocks automatically (I would hope).

 

[dadi_oh]

Think my 4850 would fold better?

The GPU client hands out points about 10X faster than the non-SMP mode on my 3 rigs. The GPU is much better at certain calculations than the CPU. I run a single instance of the CPU folding and the GPU folding at the same time on my rigs.

 

[Random Murderer]

has anyone else seen a huge hit in their ppd in the last few days? we were hitting 6k+ per day and now we're struggling to hit 2k, and i've even upped my cpu usage and thrown my 3870's core in as well...

 

[sneekypeet]

Mine goes up and down regulary, not like it was b4 where they seemed pretty similar.

see here http://folding.extremeoverclocking.com/user_summary.php?s=&u=262286

Most all of those days were damn close in actual hours folding everyday!

 

[Kursah]

They're using different WU's that aren't PPD heavy, GPU and CPU from what I've heard...though the hardest hit is the GPU clients...my GTX get's around 6.5-7.2kPPD on 480PPD WU's, and drops to around 4-5k on 430PPD WU's.

 

[WhiteLotus]

im folding as fast as i can! 1790 points in a day and a bit - i cant go any faster!

 

[BUCK NASTY]

has anyone else seen a huge hit in their ppd in the last few days? we were hitting 6k+ per day and now we're struggling to hit 2k, and i've even upped my cpu usage and thrown my 3870's core in as well...

new public NV projects 5508-5513, 5017-5022

Postby DanEnsign on Mon Sep 08, 2008 6:56 pm
Hi,

I'm warning you, you will start to see a new series of projects coming out this morning. The proteins we're studying here are larger and in some sense more relevant to biology than anything we've run on GPU2 to date.

However, due to the larger size of the protein, you may see a PPD drop putting NV cards more in line with ATI production. This isn't due to different benchmarking or a different way to assign points, but rather we anticipate NV to be slightly less efficient with larger proteins than it has been with the smaller ones.

Dan

User avatar
DanEnsign
Pande Group Member

Posts: 318
Joined: Fri Nov 30, 2007 10:41 pm
Location: Stanford, CA

 

[Random Murderer]

new public NV projects 5508-5513, 5017-5022

Postby DanEnsign on Mon Sep 08, 2008 6:56 pm
Hi,

I'm warning you, you will start to see a new series of projects coming out this morning. The proteins we're studying here are larger and in some sense more relevant to biology than anything we've run on GPU2 to date.

However, due to the larger size of the protein, you may see a PPD drop putting NV cards more in line with ATI production. This isn't due to different benchmarking or a different way to assign points, but rather we anticipate NV to be slightly less efficient with larger proteins than it has been with the smaller ones.

Dan

User avatar
DanEnsign
Pande Group Member

Posts: 318
Joined: Fri Nov 30, 2007 10:41 pm
Location: Stanford, CA

yes, i knew about that, but that only applies to gpu2 clients, and before the drop i wasn't even using the gpu2 client... so wtf?

to sneeky, yes we were seeing a similar fluctuation, but all of a sudden you could see where it took a complete nose dive.
http://folding.extremeoverclocking.com/user_summary.php?s=&u=355822
note: 8/27 is when i dedicated my cores to my dad.

 

[sneekypeet]

yes, i knew about that, but that only applies to gpu2 clients, and before the drop i wasn't even using the gpu2 client... so wtf?

to sneeky, yes we were seeing a similar fluctuation, but all of a sudden you could see where it took a complete nose dive.
http://folding.extremeoverclocking.com/user_summary.php?s=&u=355822
note: 8/27 is when i dedicated my cores to my dad.

Well I wasnt folding my GPU at all this past weekend, all points were from just her rig and one client on the E6400....funny thing its I did better those days with just her on the F@H.

It has to be just a server delay of when the work is posted vs what was actually folded that day!

 

[dadi_oh]

im folding as fast as i can! 1790 points in a day and a bit - i cant go any faster!

That's the attitude...

Go Gadget Go....

 

[irongun324]

I'm now folding on all my PC's -

Athlon 64 - 1 client
X2 4400+ - 2 clients
Laptop (in specs) - 3 clients, 1 gpu, 2 cpu...

FAHMon is showing 5200 PPD... guess thats good?

 

[newtekie1]

I'm running the SMP client, but CPU usage is only hitting 50% (near 100% on one core)... does it take a while to kick in or does it only spike above that when needed? seems to fluctuate between 49-51%.

You need to start the client with the -smp argument, or set the -smp option in the advanced options. Otherwise it will only use one core, you also have to go into your install directory and run the install.bat file.

Can I fold with my 4850? Folding with my E7200 @ 3.8 doesn't seem to be going very fast..

And its only making Core0 go up, how do I get it to do multicore?

Same as above, use the smp argument. You can fold on your 4850 using the gpu2 client.

 

[dadi_oh]

I am trying to find the right mix of OC settings that avoid getting the "mdrun_gpu returned -1" message. See below. The funny thing is, I can OC this 8800GT stable to 750 core and 2100 memory and test til the cows come home with furmark, various benchmarks, gaming, etc... The GPU is watercooled and hits in the 48C range in a warm (26C) room.

But when I leave it folding then about 1 run out of 5 or so I get the unstable machine message. The funny thing is it ALWAYS happens at the very end of the run, right after the "Completed 100%" message. I have backed off the OC all the way to 700 core and 1900 memory and I still get the occasional error. I went down to 680 core and 1900 memory and it seemed to run a few days without errors. It just seems to me that is a LOOOONG way from the 750/2100 settings that I can game at with confidence.

I have 2 other overclocked cards that run fine without these errors. A 8800GT @ 700 core, 2000 memory and a 9600GT @ 740 core, 2140 memory.

Is it possible this can be the OC on my CPU causing the issue? E8400 @ 3.88GHz. Maybe I am chasing the wrong problem?

==================================================================

[13:07:39] Completed 92%
[13:08:58] Completed 93%
[13:10:18] Completed 94%
[13:11:37] Completed 95%
[13:12:57] Completed 96%
[13:14:17] Completed 97%
[13:15:36] Completed 98%
[13:16:56] Completed 99%
[13:18:15] Completed 100%
[13:18:15] mdrun_gpu returned -1
[13:18:15] Going to send back what have done.
[13:18:16] logfile size: 127747 info=127747 bed=0 hdr=25
[13:18:16] - Writing 128285 bytes of core data to disk...
[13:18:16] Done: 127773 -> 12800 (compressed to 10.0 percent)
[13:18:16] ... Done.
[13:18:16]
[13:18:16] Folding@home Core Shutdown: UNSTABLE_MACHINE
[13:18:20] CoreStatus = 7A (122)
[13:18:20] Sending work to server
[13:18:20] Project: 5015 (Run 6, Clone 456, Gen 52)
[13:18:20] - Read packet limit of 540015616... Set to 524286976.

 

[BUCK NASTY]

I am trying to find the right mix of OC settings that avoid getting the "mdrun_gpu returned -1" message. See below. The funny thing is, I can OC this 8800GT stable to 750 core and 2100 memory and test til the cows come home with furmark, various benchmarks, gaming, etc... The GPU is watercooled and hits in the 48C range in a warm (26C) room.

But when I leave it folding then about 1 run out of 5 or so I get the unstable machine message. The funny thing is it ALWAYS happens at the very end of the run, right after the "Completed 100%" message. I have backed off the OC all the way to 700 core and 1900 memory and I still get the occasional error. I went down to 680 core and 1900 memory and it seemed to run a few days without errors. It just seems to me that is a LOOOONG way from the 750/2100 settings that I can game at with confidence.

I have 2 other overclocked cards that run fine without these errors. A 8800GT @ 700 core, 2000 memory and a 9600GT @ 740 core, 2140 memory.

Is it possible this can be the OC on my CPU causing the issue? E8400 @ 3.88GHz. Maybe I am chasing the wrong problem?

==================================================================

[13:07:39] Completed 92%
[13:08:58] Completed 93%
[13:10:18] Completed 94%
[13:11:37] Completed 95%
[13:12:57] Completed 96%
[13:14:17] Completed 97%
[13:15:36] Completed 98%
[13:16:56] Completed 99%
[13:18:15] Completed 100%
[13:18:15] mdrun_gpu returned -1
[13:18:15] Going to send back what have done.
[13:18:16] logfile size: 127747 info=127747 bed=0 hdr=25
[13:18:16] - Writing 128285 bytes of core data to disk...
[13:18:16] Done: 127773 -> 12800 (compressed to 10.0 percent)
[13:18:16] ... Done.
[13:18:16]
[13:18:16] Folding@home Core Shutdown: UNSTABLE_MACHINE
[13:18:20] CoreStatus = 7A (122)
[13:18:20] Sending work to server
[13:18:20] Project: 5015 (Run 6, Clone 456, Gen 52)
[13:18:20] - Read packet limit of 540015616... Set to 524286976.

Have you unlocked the shader clock from the core clock? GPU2 is shader intensive. Find your shader threshold first, then start bringing the core and mem up. For example, this is my 8800GS set-up
Core - 750mhz
Shaders - 1758mhz
Memory - 898mhz
Fans @ 80%
Yeilds 4937PPD on a 480pt WU

Also, are you running a single or dual card set-up? Single card set-ups have the clocks set higher on the shaders only (1836mhz) for my applications. Like any overclocking, you have to find the sweet spot. Just remember that shaders are the rule with GPU2.