GPU2 News

[MikeB12]

http://www.tomshardware.com/news/Nvi...Home,5654.html

5:30 AM - June 13, 2008 by Theo Valich
Source: Tom's Hardware US

Santa Clara (CA) - During Nvidia Editor’s Day, we learned that Nvidia and the Folding@Home research group led by Vijay Pande are making final preparation to launch the first version of the Folding@Home client for Nvidia graphics processors.

The unveiling of the client is set for the next week as part of the launch of Nvidia’s GT200 GPU series. Owning such a card will have its benefits in Folding@Home and will outrun Radeon 3870 cards. The new GeForce cards are expected to hit more than 650 nanoseconds of protein simulation in a single day, while the Radeon HD 3870 is stuck at about 170 ns. The Playstation 3 is able to produce "only" 100 ns of simulation, while a quad-core CPU creates an output of just four nanoseconds. For those who are keeping count: The GeForce GPU will be about 163 times faster than a quad-core processor in this specific application.

Nvidia founded Team "Whoopass", which consists only of several computers that are running the Folding@Home GPU client. Even with just 4-5 test machines, the team quickly moved into the top 5% of all contributors by sheer processing power. Dr. Vijay told us that if only 1% of all CUDA-capable users would start using Folding@Home in their spare time, the Folding@Home machine would quickly be considered the fastest performing HPC computer in the whole world - hitting about 60-80 Peta FLOPS of processing power.

Folding@Home for Nvidia CUDA-capable graphics cards (GeForce 8 and above) should become available next week. The codename for this client is GPU2/NVIDIA. The GPU1 client was retired, while GPU2 client will continue to be updated for both Nvidia and ATI cards.

ATI was first with a client for the Folding@Home project, which was released back in September of 2006 for the X1900 series of cards. Back then, the cards topped out at 375 GFlops. The next GPU generation should provide more than double the horsepower.

to re-iterate:

For those who are keeping count: The GeForce GPU will be about 163 times faster than a quad-core processor in this specific application.

wow, say good bye to my fah quad farm.... for me this means fah will be a gpu crunching project... I may move the cpu's over to wcg, and fold with gpu's only for fah.

[RAW-Raptor22]

I cannot wait, I need all the power a can get for my folding team (88881 for http://www.overclockworld.com) but 163 times the power, that is fricken insane. Holy crap Vijay, hurry up.

[MikeB12]

also look at the 3870@170ns compared to the new GT200@650ns...

if true, this means the GT200 will be 3.82 times faster than a 3870 gpu2.
so if a 3870 can 2000ppd, then a GT200 will 7650ppd...


here's the team mentioned in the article... nvidia team whooopass... they have been folding for 2 days and have 400K...
so they are getting 200,000ppd on 4-5 test machines if the OP article info is correct.
http://folding.extremeoverclocking.c...hp?s=&t=131015

[RAW-Raptor22]

also look at the 3870@170ns compared to the new GT200@650ns...

if true, this means the GT200 will be 3.82 times faster than a 3870 gpu2.
so if a 3870 can 2000ppd, then a GT200 will 7650ppd...


here's the team mentioned in the article... nvidia team whooopass... they have been folding for 2 days and have 400K...
so they are getting 200,000ppd on 4-5 test machines if the OP article info is correct.
http://folding.extremeoverclocking.c...hp?s=&t=131015

Im happy about the 3500 or so i will get from my 8800GT...

I saw the pic, holy siht...

[Kingcarcas]

*picks self off the floor after fainting*
:

[SocketMan]

Typical Nvidia's bravado :Whoop ass" give me a break

No numbers on the new ATI's yet - don't write them off just yet

EDIT**
************************************************** ******
http://folding.typepad.com/news/2008...-increase.html

GPU2 PPD increase

With new revisions to the GPU2 core code, the GPU2 code is doing more science per ever in the same amount of time. In the spirit of having the points equate to how much science is being done, we are going to increase the PPD of the GPU2 benchmark machine (ATI 3850 board -- see the FAQ for details) to 1500 PPD. We will revalue the most recent WU's (472X) and all future WU's will use this new PPD for benchmarking.

************************************************** ************************************************** ****
All 4 of my gpu2 clients have in fact lost ~500ppd per client over the last ~week
So now they gonna bring it back up to where it was? Where is the increase? or maybe it's only for NV?

[DAK1640]

I am prepared to abuse my GPU...Let's go !!!

[angra]

Can't wait. I have a lot of GPUs that are not fully utilized.

[coo-coo-clocker]

same here, but I can't help but think that the extraordinary speed is for very targeted projects...similar to the PS3. At least that is what stanford has been saying all along: SMP client provides the highest flexibility for a wide range of projects, while GPU/PS3 provides high performance on a narrow set of projects. Still holds true with GPU2?

[MikeB12]

Pande corrected the numbers posted in the Tom's article above...
http://foldingforum.org/viewtopic.ph...&t=3117#p28792

PS Note that Tom's H numbers are a bit misleading. We're not getting 650ns/day (yet) on the gtx280 (more like 550) and we're now getting 250ns/day on 3870's in the lab (perhaps 300ns/day in time), and the 3870's are the previous gen ATI cards. The gtx280 is going to be really great at folding though.

[angra]

Also, as seen over and over, raw speed != ppd.

We will see when we see! I'm psyched to find out, though!

[RAW-Raptor22]

I cant wait to see how overclocking the GPU affects the folding output.

[SocketMan]

It will probably come out on the same day as the new NV cards:
the 18th?

[RAW-Raptor22]

It will probably come out on the same day as the new NV cards:
the 18th?

I still cant wait.

[MikeB12]

more discussion on fah forum..
http://foldingforum.org/viewtopic.ph...&sd=a&start=15

someone posted this chart:

[Stijndp]

But the Q6600 looks like a stock speed installed on a Windows OS. We know that baby does +3500ppd and then the gpu's don't give more ppd for the cost. Not only looking at the cost of building a system but also running it.

[MikeB12]

yeah, but that's now. ppd restructuring hasn't happen yet...

[bowman]

The Nvidia client beta was leaked a day or two ago;

http://benchmarkreviews.com/index.ph...1&limitstart=4

I've been running it and fahmon says 2065 ppd with a slightly overclocked 8800GTX. It's actually running slower than the 9600GTs though, so whether that's down to the work unit or the AMD CPU holding it back, I have no clue..

[RAW-Raptor22]

OMG, this is folding so fast.

[bowman]

177.35 drivers 'unleash the G80 powah'. 4980 estimated PPD in FahMon. (More than 500ns/day)

[SocketMan]

The Nvidia client beta was leaked a day or two ago;

http://benchmarkreviews.com/index.ph...1&limitstart=4

I've been running it and fahmon says 2065 ppd with a slightly overclocked 8800GTX. It's actually running slower than the 9600GTs though, so whether that's down to the work unit or the AMD CPU holding it back, I have no clue..

Thanks for the link !
Does it work with winXP32/64 ?
Don't know if rivatuner (or something else) can show "the gpu utilization" but
if it's at 100 (90%+) then the cpu is not holding it back
EDIT**
that's on phenom @ 2600 vv

[RAW-Raptor22]

Noob Question: How many consoles can I still run on my PD820 while my 8800GT is folding?

[bowman]

Thanks for the link !
Does it work with winXP32/64 ?
Don't know if rivatuner (or something else) can show "the gpu utilization" but
if it's at 100 (90%+) then the cpu is not holding it back
EDIT**
that's on phenom @ 2600 vv

Apparently it works on XP, I don't know, I'm on Vista64.

CPU utilization was 100% on one core with the old drivers, now it's below 25% while not using the visualizer (I avoid using it at all, it eats away at the crunching).

The GPU just gets to 57C, if I stress it with FurMark it reaches 70C.

Obviously the texturing units and such graphics-specific units are twiddling their thumbs but somehow I feel they still have more juice to squeeze out of the cards. It's a pre-beta after all.

EDIT:

http://foldingforum.org/viewtopic.php?f=43&t=3186

6.12 beta6 is now officially out, as opposed to the leaked beta1.

[kevinbo03]

Question: Does the GPU client only run on the GPU? Or should I tell it to only use %5 of the cpu? I ask because I've got WCG running on my CPU and I don't want it to interfere too much. A little (%5) is fine.

[MikeB12]

let it utilize a whole core if you're gonna use gpu2, it needs an entire core to run efficiently... otherwise it'll choke and won't turn any ppd... wish it wasn't this way. but it is..