SPCR

Found this just then, thought some people might be interested.

http://www.stanford.edu/group/pandegrou ... -612b6.msi

Works ok for me so far, though it makes my pc noticably slower.

Here's a screen grab of the Display:

1% every 20 seconds on an 8800GT.

Yup, just got it running today too - approximately 25 secs / % on a 9600GT for project 5002. PPD is approximately 3500!!!??

GPU temp is fluctuating between 51-53C with a passive S2 cooler.
CPU is back to running a single standard client, pretty poor showing of 387PPD compared to the 1920 I was getting with the SMP version. Though this total of ~3800PPD is almost 2x more...

This is a big deal. I know the ATI folks have had GPU2 for a while, but I'm really happy to finally be able to use it myself (I only run NVIDIA.)

You'll need CUDA drivers from http://www.nvidia.com/object/cuda_get.html or the leaked beta 177 driver and a modified INF file ( http://foldingforum.org/viewtopic.php?f=43&t=3256 .)

On my Gigabyte 8800GT at 700Mhz, fahmon shows 4700 PPD for project 5002. My points output for the entire month of May can now be surpassed in a couple of days with GPU2.

Of note, one of the busiest threads on the folding forums has been about coil whine. Apparently some folks never noticed it before, or never actually had it til they ran GPU2: http://foldingforum.org/viewtopic.php?f=43&t=3212 . Mine has been there since I got the 8800GT, but is far more noticeable running GPU2 than with any games. With the problem obviously widespread, one has to wonder if there's any way to bring pressure to bear on nVIDIA or a few card OEMs to provide a remedy. But I digress...

My stats: http://folding.extremeoverclocking.com/ ... =&u=335156 Note the tremendous increase in point production since yesterday!

Hah, I skimmed the GPU client FAQ on the Stanford site and mistakenly thought that these new WUs would be worth 1500. That would've been nice.

4980 PPD for me. It's good to be producing again after stability issues with the SMP client.

On the downside, folding on the GPU has definitely uncovered some issues with my card. I get a high-pitched whine under load (changes frequency depending on the level of load, or my clocks -- currently I'm at 725/1080 memory and I get a nice high F#). I think my problem is actually with a voltage regulator (if those little black cubes next to the caps are voltage regulators), which is loose enough to be jiggled around and which affects the tone and volume of the whine depending on which way it's jiggling.

Is anyone else getting this sort of issue while folding? Apparently the problem is pretty widespread across the G80 family.

mightysoy wrote:Hah, I skimmed the GPU client FAQ on the Stanford site and mistakenly thought that these new WUs would be worth 1500. That would've been nice.

4980 PPD for me. It's good to be producing again after stability issues with the SMP client.

On the downside, folding on the GPU has definitely uncovered some issues with my card. I get a high-pitched whine under load (changes frequency depending on the level of load, or my clocks -- currently I'm at 725/1080 memory and I get a nice high F#). I think my problem is actually with a voltage regulator (if those little black cubes next to the caps are voltage regulators), which is loose enough to be jiggled around and which affects the tone and volume of the whine depending on which way it's jiggling.

Is anyone else getting this sort of issue while folding? Apparently the problem is pretty widespread across the G80 family.

Yeah, common problem sadly.

I also have it in a much more toned down version in intense games like Crysis.

So what is a good iter/ sec? I get 1600-1800 on average.

shadestalker wrote:
You'll need CUDA drivers from http://www.nvidia.com/object/cuda_get.html or the leaked beta 177 driver and a modified INF file ( http://foldingforum.org/viewtopic.php?f=43&t=3256 .)

All I needed was the latest nvidia driver for XP (175.16, released 16th May)

mightysoy wrote:Is anyone else getting this sort of issue while folding?

I have that problem, though I didn't notice it with my headphones on last night.

darkb wrote:I have that problem, though I didn't notice it with my headphones on last night.

My headphones were actually picking up and emitting the same tone (I was plugged into a front 1/8" jack). Switched to the output in the back and I haven't had the same problem, although the whine is still loud enough to be audible when I have my headphones on. Granted, they're pretty crappy headphones.

After realising that the noise increases in pitch relative to my core/shader clocks, I've been overclocking incrementally in an effort to force it out of my perceptible range. I'm at 750/1875 MHz right now and it's a lot less frustrating than it was at 700.

On the plus side, that means I'm now doing 16-17 seconds/%, approx. 5300 PPD.

Finally I'm running the GPU2 core as well

I had some trouble finding the correct drivers (for me).
The CUDA_get site now lists 177.35 drivers, which didn't work here (I suspect it is only meant for the new G200 cards - I couldn't even install the driver).

I tried a newer driver (from the original mentioned 174.55), 175.16, but this didn't seem to support CUDA.

On this page I found the driver which works for me: 174.55 (x86 for "normal" 32 bit systems, x86-64 for 64 bit systems).

The visualisation ("display" in the right click menu) is highly unstable here, and will shut down the core every time. There will be a new version where this will work (not important for power folders, as they are used to a DOS box). Hovering the mouse over the system tray icon will show the progress.

To me this is a major improvement!! Preliminary PPD (FAHmon): 4400 (overclocked 8800 GTX, Athlon 64 X2 5200+ @ 2.8 GHz).
...can it...?

Shadestalker mentions a 4700 PPD with a 8800GT. I read somewhere, CPU speed also has an effect on the final PPD results. Intel CPU's will have an advantage here as well.
Or... 700 MHz... I haven't OC'ed to that high a core clock.

Mightysoy even gets a mighty 5300 PPD but I'm happy.
Temp seems to be stable at 65 °C at the moment.

shadestalker wrote:Of note, one of the busiest threads on the folding forums has been about coil whine. Apparently some folks never noticed it before, or never actually had it til they ran GPU2: http://foldingforum.org/viewtopic.php?f=43&t=3212 . Mine has been there since I got the 8800GT, but is far more noticeable running GPU2 than with any games. With the problem obviously widespread, one has to wonder if there's any way to bring pressure to bear on nVIDIA or a few card OEMs to provide a remedy. But I digress...

Oops, sorry, didn't realise you'd already mentioned that. In retrospect I've realised that I was also hearing the sound when playing certain games (I'd always thought it was my hard drive).

spookmineer wrote:I read somewhere, CPU speed also has an effect on the final PPD results. Intel CPU's will have an advantage here as well.

Probably true, the GPU client also uses about 50% of my CPU power. For reference: I'm running an E6550 @ 3.15 GHz, 8800GT @ 750 MHz, PPD = 5292.

shadestalker wrote:My stats: http://folding.extremeoverclocking.com/ ... =&u=335156 Note the tremendous increase in point production since yesterday!

Haha, you and I have left a swath of destruction and red arrows in our wake since this client became available.

I've also heard of some instability if using remote desktop protocol to remotely connect to the folding box and can confirm that happens too. It shuts down the core - probably due to the unconventional rerouting of video data. The core will shut down after completing a project and will not restart until you log on locally and restart the service. Kinda sucks, but should be rectified.

Never heard coil whine before, but now its apparent when I was checking behind the box to route some cables. Its under the noise floor of my place until about 2ft. It definitely goes away when I turn off the core so I guess that's what it is.

shadestalker wrote:On my Gigabyte 8800GT at 700Mhz, fahmon shows 4700 PPD for project 5002.

mightysoy wrote:

shadestalker wrote:My stats: http://folding.extremeoverclocking.com/ ... =&u=335156 Note the tremendous increase in point production since yesterday!

Haha, you and I have left a swath of destruction and red arrows in our wake since this client became available.

It is confusing... My actual PPD is lower than FAHmon indicates, same as with shadestalker's stats: his PPD is "only" 3000.

I managed to work my way up to fourth on the daily producers list, but at the cost of an uncomfortably stuffy room. I'm going to stop for now, and start up again when the weather is cooler (i.e. in some months' time). Still, I think the GPU client has proven its merit -- in the week that I ran it, I averaged about 4500 points a day.

Good/continued luck to the rest of you (I noticed that the others who are running this client have also shot up the rankings).

I might test this on my 8600GT equipped WS2, which is currently the only folding@home box I have...

After a miserable experience last year with my ATI card I gave up on GPU folding, but now that I've seen the numbers I must say this seems a better way to go over building another system.

I have 2 dualies with integrated video and empty PCI Express X16 slots.

2 Questions though:

1. Is this Windows only or can I do this with Ubuntu as well?

2. Has anyone measured how many watts their video card pulls?

Thanks very much.

Aris

aristide1 wrote:1. Is this Windows only or can I do this with Ubuntu as well?

Windows only for now, possibly linux/OS X in the future (aka no work on it yet)

2. Has anyone measured how many watts their video card pulls?

xbitlabs are usually pretty good at that. I'd imagine these kinds of calculations will at least move the power draw close to peak 3d power draw.

Yes I found that it's Windows only, and the extreme PS calculator seems to suggest about 75 watts for the card I am considering.

Biggest bang for the buck seems to be 8800GTS G92 processor over 5000PPD, at 75 watts!

Is anyone here actually doing this now?

My only problem is I will have to buy Win XP or Vista, as my other folders are running Ubuntu and 2000Pro as well as an Artic Cooler silent video cooler. $90 + $30 extra. But still, such a deal!

Also - Can you still run SMP? I know people are running the GPU + a single here and there.

Thank you!!!

Aris

aristide1 wrote:Yes I found that it's Windows only, and the extreme PS calculator seems to suggest about 75 watts for the card I am considering.

Biggest bang for the buck seems to be 8800GTS G92 processor over 5000PPD, at 75 watts!

As you can see here, an 8800 GTS will draw approx. 105 Watt.
This graph shows the same power consumption.
The above measurements are not for an entire system, an entire system needs more power.

aristide1 wrote:Is anyone here actually doing this now?

Also - Can you still run SMP? I know people are running the GPU + a single here and there.

Thank you!!!

Aris

Maybe you missed my post in the other thread:

spookmineer wrote:

kittle wrote:Im still trying to get the GPU client and the "new" smp client to play well together. I have to use the older version of their beta because the new one doesnt support 64-bit windows.

As I read when the new GPU client was just released, it is not recommended to run the GPU client together with an smp client, as the GPU client will use a considerable amount of CPU power (here it's at about 52%).

Hopefully (and hinted at by Stanford) a newer GPU client won't use as much CPU power.

Anyway, if you get both of them working, great

I haven't tried running both GPU2 and SMP as it was not recommended.

spookmineer,

On the extreme pc calculator I went from the lowest video card they had to the 8800GTS and it went up 75 watts, but I guess even the lowest card was 25 watts.

I may instead replace the video card (ATI X1950GT) in my main PC with an 8800GTS, and that will raise my power consumption appx 15-30 watts total.

spookmineer wrote:Maybe you missed my post in the other thread:

You are correct, I ignored the GPU folding for quite a while. My mistake. Where is it?

mightysoy wrote:Probably true, the GPU client also uses about 50% of my CPU power. For reference: I'm running an E6550 @ 3.15 GHz, 8800GT @ 750 MHz, PPD = 5292.

Nice. I've heard some people do the GPU + a regular single core WU for another 300 or so PPD.

As far as I'm concerned less CPU workload also means less power consumption.

So if your CPU is not folding anymore, only supporting the GPU(s) then what would be the point of getting a quad core?

aristide1 wrote:So if your CPU is not folding anymore, only supporting the GPU(s) then what would be the point of getting a quad core?

FAIK there would be no point. But apparently a fast dual core makes some difference. Mightysoy doesn't seem to be folding anymore, but when he was he got more PPD than me with a "slower card" (*), but faster CPU (he did OC* the video card a lot too).
Typically, the GPU client forces one CPU core to be loaded 100% while the other core takes on other tasks.

The OC of his card alone might explain his PPD: 5292* (612/750) = close to my PPD.

So what I would like to see is a quad core, with SMP forced to run only on 2 cores, and with 2 video cards, each with one of the remaining CPUs working for it.

Yummy.

This thread eventually shows which NVidia card gets the highest PPD (good) and also the highest PPW (even better).

The latter 4500 PPD at 60 watts / per hour of execution on the 8800GS.

http://foldingforum.org/viewtopic.php?f ... 644#p47644

well I ran just the GPU2 client over the weekend and it was getting really small WUs -- 68 points each. Then it would run through a stage where it downloads a wu, fails to run and then goes to grab another one. some i got credit for. some I didnt.
so I see 480points for a day and 8 WU. and the next day 960points and 2 WU.
go figure :/

Now I have the SMP and gpu running side by side. I'll report back in a few days on how things are running.

Kittle, what card are your using?

I knew something was interesting when I heard of the video cards taking checkpoints every 75 seconds.


Has anyone started looking at motherboards that support 2, 3, 4 video cards? You can get a quad AMD board for $150, but the Intel boards are still very high. Sole exception:
http://www.newegg.com/Product/ProductRe ... 6813157120

im using the bare minimum for a card - an 8600. It was one that MikeC offered for sale on the spcr lab clearance.

mine does 1% every 10-12min.

ive been eyeing a GTX 280 or a 9800 card but the noise factor has me very wary of getting one.
Great folding and also great short term upgrade for my system.

kittle wrote:im using the bare minimum for a card - an 8600. It was one that MikeC offered for sale on the spcr lab clearance.

mine does 1% every 10-12min.

ive been eyeing a GTX 280 or a 9800 card but the noise factor has me very wary of getting one.
Great folding and also great short term upgrade for my system.

I used an EVGA 8800 GS 384MB ($64, delivered after now expired promo code and MIR from newegg) at stock settings. It was inaudible in my system (NSK-2480, 2x120 Thermaltake exhaust fans at 700RPM, fanless Ninja Mini) and produced nearly 3800PPD. If you can stand a bit of white noise (fan at 70%), you may be able to crank it up to 5120PPD, like I have done. I'm in the process of moving the fan speed down to balance the noise and temperature with the PPD.

Then 8800GS 384MB produces the most PPD per watt. Far more than either a GTZX280 or 9800. Source: http://i28.photobucket.com/albums/c241/ ... ppdper.gif

Mu GPU2 client expired today. Downloaded and installed the new version 6.20.

Due to a power outage at Stanford the servers have been down... My GPU2 hasn't been working ever since I installed it a few hours ago "attempting to get work packet".

The EVGA discount didn't last long, just over the weekend, now it's $80 after rebate. But.....

http://www.newegg.com/Product/Product.a ... 6814150276

is also $80 without a rebate, free shipping, and it's faster. Whoohoo!