Attention GPU Folders: Make sure you are set up for QRB units

Status
Not open for further replies.
My 560Ti handles these great. I just can't keep them coming steady.
Plus, no way these are weighted fairly to other WUs on GPUs.
My 560Ti produces 8 times the PPD with 8057 as an 8054, with virtually identical TPF: 1:40 vs. 1:45.
And, with 8057, my card runs 10C cooler. Rediculous I would say.
 
Hmm, I wonder if the GPU qrb scheme is their new "viral marketing scheme". :)
 
Core32 said:
My 560Ti handles these great. I just can't keep them coming steady.
Plus, no way these are weighted fairly to other WUs on GPUs.
My 560Ti produces 8 times the PPD with 8057 as an 8054, with virtually identical TPF: 1:40 vs. 1:45.
And, with 8057, my card runs 10C cooler. Rediculous I would say.
Click to expand...

Its weird for sure.
 
Jeanjean said:
Yes , I have.

5 hours later , i started again my benchmark with the same WU and the PPD has been increased by two with the same TPF :rolleyes: :confused: :
Click to expand...

Dont forget that the ppd and points are calculated based on total time to complete the WU, so if it was paused for a large amount of time at some point, then the points value and PPD will drop quite a bit.
 
extide said:
Dont forget that the ppd and points are calculated based on total time to complete the WU, so if it was paused for a large amount of time at some point, then the points value and PPD will drop quite a bit.
Click to expand...

This I think.

If im reading correctly, Jeanjean's first screenshot was taken approx 24hrs after you had obtained the WU. In the second pic, it was about 6hrs.
 
coffaholic said:
This I think.

If im reading correctly, Jeanjean's first screenshot was taken approx 24hrs after you had obtained the WU. In the second pic, it was about 6hrs.
Click to expand...

No.

It is the same WU restarted 6 hours later but with twice more points despite the same TPF. :eek:
 
OK, I updated my core just like in the OCN post and just redownloaded and pasted over it again, when I look at the log it still says that its running this - 21:06:31:WU01:FS00:0x15:Version 2.25 (Wed May 9 17:03:01 EDT 2012)

So is there a better place to get the 2.22 core from? Or is it possible I'm not changing the correct core? Because I've downloaded it twice now, and I can't check it before I resume folding but it doesn't look like anything additional is getting downloaded. My TPF is still 4:46 which works but as many have said its what you expect for a 580 using the 2.25 core. Thanks in advance.

EDIT
So I found my problem, when you change the install location it copies the install files to your c:/users/app... roaming... etc location (which is what I was changing) but the cores that it uses are in the *:/programdata/FAHclient/...etc/.. folder... just kicking myself for that one...

I'll let it settle down and see if the TPF is any better.
 
Last edited:
Core32 said:
My 560Ti handles these great. I just can't keep them coming steady.
Plus, no way these are weighted fairly to other WUs on GPUs.
My 560Ti produces 8 times the PPD with 8057 as an 8054, with virtually identical TPF: 1:40 vs. 1:45.
And, with 8057, my card runs 10C cooler. Rediculous I would say.
Click to expand...

Dang, I just noticed my 470 also runs 10C+ cooler when running the 8057 units vs the 8018 units.
 
Skillz said:
Dang, I just noticed my 470 also runs 10C+ cooler when running the 8057 units vs the 8018 units.
Click to expand...
And an 8018 is running even hotter.
The 8057 parameters are massively screwed up, in my opinion, even for a beta run.
I smell something rotten in WU land.
How "convenient" this kicked in right after the #1 ranking changed teams.
 
Jeanjean said:
No.

It is the same WU restarted 6 hours later but with twice more points despite the same TPF. :eek:
Click to expand...
If you're using V7 to get PPD I'm not surprised. I've had an 8101 report 3,276,440 PPD after a restart on a machine that averages 135-140,000 PPD.
For benching it's best to get the average TPF over a dozen frames or so directly from the log and use that to calculate PPD.
 
Amaruk said:
If you're using V7 to get PPD I'm not surprised. I've had an 8101 report 3,276,440 PPD after a restart on a machine that averages 135-140,000 PPD.
For benching it's best to get the average TPF over a dozen frames or so directly from the log and use that to calculate PPD.
Click to expand...

same here. i had v7 show an smp unit that gave me a tpf which equated to 325k ppd on a 3930k. now i hand calculate my tpf after the unit is done and see the real numbers.
 
Core32 said:
And an 8018 is running even hotter.
The 8057 parameters are massively screwed up, in my opinion, even for a beta run.
I smell something rotten in WU land.
How "convenient" this kicked in right after the #1 ranking changed teams.
Click to expand...


I do agree that the points on this WU are pretty crazy, and think they will probably be adjusted at some point. However I truly find it hard to believe that the Pande Groupe is "in cahoots" with EVGA and thus throwing them a bone to get back into first place. I dunno, that's just my opinion.
 
Core32 said:
And an 8018 is running even hotter.
Click to expand...
Skillz said:
Dang, I just noticed my 470 also runs 10C+ cooler when running the 8057 units vs the 8018 units.
Click to expand...


8018 is 2262 atoms, 25,000,000 steps.

8057 is 900 atoms, 50,000,000 steps.

If steps are equal, this makes 8018 25.67% larger than 8057.

My GTX 680 runs 8018 at 00:04:02 TPF, 8057 at 00:01:52 TPF.

Thus 8018 takes 2.16 times longer to complete.


Taking over twice as long to fold a project which is just 26% larger
is a good example of how Nvidia GPU's performance declines as the size
of the WU increases. It also is why Skillz' 470 runs 10C cooler (and my
680 draws 20 watts less) while folding 8057.


Another interesting bit - 8018 and 8020 are the largest successful GPU projects
to date. While it's true that 7640-7644 have more (21,443) atoms, they only
have 2,500,000 steps, making them slightly smaller overall than 8018/8020.

They also fail on a regular basis, and have been on hold since May.

extide said:
However I truly find it hard to believe that the Pande Groupe is "in cahoots" with EVGA and thus throwing them a bone to get back into first place. I dunno, that's just my opinion.
Click to expand...


Agreed.

Two things.

First, it's no secret that Vijay has been big fan of using GPUs from the very beginning of GPGPU.
The fact that EVGA was essentially handed first place on a silver platter was a consequence of the
pursuit of GPU power in general, not a specific attempt to aid a specific team.

Second, it appears Sony and PG have been in 'discussions' for at least several months regarding the
future of the PS3 client. It seems fairly obvious that Sony, who's stock is nearing JUNK STATIS,
was demanding cash in return for allowing folding to continue on playstations. When PG didn't
come up with the ransom, Sony executed all the hostages.

Those PS3s accounted for about 20% of FAH's raw power.

The logical replacement for all those lost PS3 folders would be GPUs.
 
Amaruk said:
it's no secret that Vijay has been big fan of using GPUs from the very beginning of GPGPU.
The fact that EVGA was essentially handed first place on a silver platter was a consequence of the
pursuit of GPU power in general, not a specific attempt to aid a specific team.
Click to expand...

^^ Have to agree with Amaruk, [H] took full advantage of the GPU's last time they were king to extend their lead over everyone else, our problem was not waking up to the EVGA challenge in time to stop it.
 
PG is not sleeping with EVGA the 8057's are just that much faster, did PG implement the QRB to tame some direst at EVGA maybe but who cares we can us this just as well and easily as they can. GPU folding is the easiest and cheapest form of folding there is. You just need a MB K9A2 if you can find 1 with lots of x16 slots a low power processor and a few relatively cheap GPU's a 1200W power supply and you will be at 1,000,000 PPD. Enjoy the cheap PPD while you can and forget about those other guys they can not keep up with us if we mobilise in time. :)
 
Amaruk said:
If you're using V7 to get PPD I'm not surprised. I've had an 8101 report 3,276,440 PPD after a restart on a machine that averages 135-140,000 PPD.
For benching it's best to get the average TPF over a dozen frames or so directly from the log and use that to calculate PPD.
Click to expand...

Ok , thanks for the explanation .

I thought that Standford had divide the PPD of 8057 by 3 . :rolleyes:
 
This is just another example of PG's points inflation. Whenever it seems people are tiring of Folding we get another new process where the lucky ones who look on forums like this find out how to get units that they shouldn't get and points that are ridiculously out of whack. And it stirs up the lapsed folders to fire up again and some people to go out and buy a whole load more kit.

Please everyone, just remember this bait and switch tactic has been done before. We had this when new units got 18K on an 8800GT, we had this when 460's could get 20K.

And then it gets changed and we all whinge about the ppd collapsing. Enjoy it while you can, but please don't build long-term plans around a single beta unit.
 
Happy to report after a days of work... Fedora 17 x86_64 makes it completely impossible to run a NVidia GPU client under cuda/wine. Why in the hell would you want to run any other driver than nouveau in total graphic mode? Another thank you card sent to Stanford for native linux GPU support. On a good note, they got the Fuckdows 8 GPU client working.
 
Guys, the thread is about setting up the GPU clients, not about EVGA or PG, or money swapping hands. Lets keep it on topic.
 
A 450GTS caught a 8057 unit and is reporting 440000 points per day! Insane but hell, make hay while the sun shines!
 
402blownstroker said:
Happy to report after a days of work... Fedora 17 x86_64 makes it completely impossible to run a NVidia GPU client under cuda/wine. Why in the hell would you want to run any other driver than nouveau in total graphic mode? Another thank you card sent to Stanford for native linux GPU support. On a good note, they got the Fuckdows 8 GPU client working.
Click to expand...

What exactly is impossible about it? I am running without issue in Ubuntu 10.10. Is it because the driver doesn't work in GUI mode? That was going to be the next thing I tried.
 
MikeTimbers said:
Insane but hell, make hay while the sun shines!
Click to expand...

Exactly the point of this thread. Many of us probably have some GPUs lying around or in use in some machhine. We may as well crank them up while these inflated points are available. Our competition certainly is....
 
MikeTimbers said:
A 450GTS caught a 8057 unit and is reporting 440000 points per day! Insane but hell, make hay while the sun shines!
Click to expand...

I've had my share of strange looking PPD, so I would not trust any PPD numbers stated by the v7 client.
 
tjmagneto said:
I've had my share of strange looking PPD, so I would not trust any PPD numbers stated by the v7 client.
Click to expand...

My post should of course have said 44000, not 440000, but still crazy. That card is in a rig with a 465 which is reporting 85K ppd for an 8057.
 
I seem to be only getting a 8057 for each 8018 I process... Is this normal...
 
sounds about normal to me. i got a month of 8018's before my first 8057. feel lucky you're getting them as often as you are.
 
Skillz said:
No, that's pretty lucky really. Some people are going through a few 8018s before getting an 8057.
Click to expand...
Then I must be really lucky. I installed a pair of brand new 670's and they both grabbed 8057's right away. I think among all of the 3 GPUs I have set up (the 670s plus a 560Ti), I have had twice as many 8057's as other units.
 
I seem to be getting 3-4 8018's then 3-4 8057's, I actually got an 8054 also. The suspense is killing me lol, I want to get on and buy a GPU or two for folding, but I am going to wait until the points shake down a bit to see where everything will really stand. For all we know, things could change very significantly, or perhaps not...
 
I now have 4 cards running with the 6.41 console client - approximately 575K ppd. It is extremely easy, and I can send you my exact config if needed.
 
musky said:
What exactly is impossible about it? I am running without issue in Ubuntu 10.10. Is it because the driver doesn't work in GUI mode? That was going to be the next thing I tried.
Click to expand...

There is now way that I have found to disable the crappy nouveau driver that gets load by default to run the NVidia driver install app. Blacklisting it and getting the kernel boot parameters to disable it does nothing.
 
-ZS-Carpenter said:
450 at stock clocks, v7 core v2.22 8057 tpf 3:20 62.3k ppd

Should fall on its face again like it did during kepler beta testing

Thanks for the conspiracy theory lolz btw :rolleyes:
Click to expand...

thanks for the info ..and .. I always like to plant the seed when I can :p

..you know somebody was think'n it
 
Status
Not open for further replies.
Back
Top