theGryphon
[H]ard|Gawd
- Joined
- Nov 21, 2011
- Messages
- 1,295
Is it really an old GPU WU? Do we know what project this is, what actual work is being done by it? I was thinking they must have ported an SMP WU, because only that would make sense...
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
News from PG re GPU benchmarking
- Thread starter Nathan_P
- Start date
Is it really an old GPU WU? Do we know what project this is, what actual work is being done by it? I was thinking they must have ported an SMP WU, because only that would make sense...
Yes it is... the announcement says they are taking an old gpu3 wu....
and it has the same atom count as an implicit wu...
And I agree that only a ported smp unit would make sense...
I posted the psummary section for you to see...
taken from here
http://www.overclockers.com/forums/showthread.php?p=7324327#post7324327
3770k & 2 670's (I have no idea what the frequencies are - not my rig)
I uploaded it to imageshack because you can't see images on that forum unless you are a member.
Depending on how well V7 is calculating that PPD, (and we all know how we it does that)
That is ridiculous!! If that image is accurate, that is NOT two GTX670's (they are GK104) but 2x 650's or 660's, which only have 384 kepler shaders. It appears they are getting ~100K ppd each, and thats a pretty dang low end GPU!
Xavier Zepherious
n00b
- Joined
- Feb 20, 2012
- Messages
- 58
gtx670 is listed as a gk107 in the gpulist
there has been a note on this posted on FF to get corrected
there has been a note on this posted on FF to get corrected
Skripka
[H]F Junkie
- Joined
- Feb 5, 2012
- Messages
- 10,791
402blownstroker
[H]ard|DCer of the Month - Nov. 2012
- Joined
- Jan 5, 2006
- Messages
- 3,242
Between having the 8101 WUs shoved up SMP folders asses and this, motivation to continue to folder will be harder and harder for some people to justify. Mixing politics, people pleasing, and science has never been a good combo.
Can you tell us your exact configuration?
that is NOT two GTX670's (they are GK104) but 2x 650's or 660's, which only have 384 kepler shaders. It appears they are getting ~100K ppd each, and thats a pretty dang low end GPU!
To clarify,
a 640/650 has 384 Cuda cores
a GTX 660 has 960 and a GTX 660ti has 1344.
To clarify,
a 640/650 has 384 Cuda cores
a GTX 660 has 960 and a GTX 660ti has 1344.
Version 7, running on Win 7, client-type = beta.
What other information would you like?
That's good thanks
17,405 points with a TPF just over 2 minutes. Don't have the actual here in front of me.Keep tabs on what that WU actually scores when you turn it in please.
This is on the 560Ti. Nothing from the 460s so far.
So to clarify:
This is the new QRB WU, on a mid range GPU.
If so, much more reasonable.
No kendrak. He got 17k total points for a Wu that took a bit longer than three hours. That's about 125k ppd from that.
Thanks. Keep us posted.
I've posted the gist of what patriot said earlier in the DAB.
GPUs have always been the future. Dump SMP and run.
http://www.hpcwire.com/hpcwire/2012-10-29/cray_debuts_new_xk7_line_in_titan_supercomputer.html
http://www.bbc.com/future/story/20121026-a-supercomputing-titan
The only reason to cut GPU points would be to cover up the mistake of horribly exaggerated bigadv points.
http://www.hpcwire.com/hpcwire/2012-10-29/cray_debuts_new_xk7_line_in_titan_supercomputer.html
http://www.bbc.com/future/story/20121026-a-supercomputing-titan
The only reason to cut GPU points would be to cover up the mistake of horribly exaggerated bigadv points.
I think we all agree that if a GPU can run THE SAME work unit as a CPU then they should both get equivalent points and any QRB applied. This is what the initial blog post stated and I would be 100% behind that.
Unfortunately at this time this isn't the case. A test WU has been created with the points whacked up 10x what it was and a QRB added on top.
Unfortunately at this time this isn't the case. A test WU has been created with the points whacked up 10x what it was and a QRB added on top.
Xavier Zepherious
n00b
- Joined
- Feb 20, 2012
- Messages
- 58
if a GPU can return that same WU in less time than a cpu... then it deserves more bonus points than a SMP.,it's that simple. What you guys are arguing for it to keep the status quo.. that's not gonna happen.
a previous WU was used - points already established for it, They can do work on both GPU &smp with the unit. the GPU just returns more points because it does the WU much quicker (Quick return bonus).
it's about the project doing equal points for equal work. You guys aren't the judge of what is equal work because you don't know what the work is... stanford does. if there is an issue with the points they will take care of it, because they know what is fair and what is equal work.
judging from what Im hearing...the points are correct and expect this to be the norm
if a gpu can do the WU in 1/2.,1/4 or 1/8 the time it deserves big QRB numbers (non-linear growth in points)
this WU has already gone thru scrutiny at Stanford on CPU and GPU (benchmark machines both gpu and CPU)
points were already establish as fair according to them.
what our (EVGA) DAB rep said
http://www.evga.com/forums/fb.ashx?m=1778909
a previous WU was used - points already established for it, They can do work on both GPU &smp with the unit. the GPU just returns more points because it does the WU much quicker (Quick return bonus).
it's about the project doing equal points for equal work. You guys aren't the judge of what is equal work because you don't know what the work is... stanford does. if there is an issue with the points they will take care of it, because they know what is fair and what is equal work.
judging from what Im hearing...the points are correct and expect this to be the norm
if a gpu can do the WU in 1/2.,1/4 or 1/8 the time it deserves big QRB numbers (non-linear growth in points)
this WU has already gone thru scrutiny at Stanford on CPU and GPU (benchmark machines both gpu and CPU)
points were already establish as fair according to them.
what our (EVGA) DAB rep said
http://www.evga.com/forums/fb.ashx?m=1778909
Skripka
[H]F Junkie
- Joined
- Feb 5, 2012
- Messages
- 10,791
Well for p8057 v. p8056, you have 1/24 as many atoms, you have 13x higher base points , longer timeout...and 10X higher PPD for GPUs than before. Sure sounds like "equal pay for equal work" to me. So no, currently you're NOT turning in the same WU faster and being justly rewarded with p8057.
Well for p8057 v. p8056, you have 1/24 as many atoms, you have 13x higher base points , longer timeout...and 10X higher PPD for GPUs than before. Sure sounds like "equal pay for equal work" to me. So no, currently you're NOT turning in the same WU faster and being justly rewarded with p8057.
Exactly. Why do they say equal pay for equal work and then do that? Sigh
Xavier Zepherious
n00b
- Joined
- Feb 20, 2012
- Messages
- 58
they took a Previous work(already done) with the same atom count
do you see any with 900 atoms present? NO
only one close is 8054 (no k factor yet)
different WU's have differing points (different work and project)
some work is more valued than others as well
otherwise all work would have the same atom count, points and same QRB
compare the time taken...does a gpu return it in faster time?
if so it deserves more points
preferred of 6.86 day and you doing it in 2-3 hrs?
what do you expect (faster means a lot more points)
under SMP it would take 2 days
do you see any with 900 atoms present? NO
only one close is 8054 (no k factor yet)
different WU's have differing points (different work and project)
some work is more valued than others as well
otherwise all work would have the same atom count, points and same QRB
compare the time taken...does a gpu return it in faster time?
if so it deserves more points
preferred of 6.86 day and you doing it in 2-3 hrs?
what do you expect (faster means a lot more points)
under SMP it would take 2 days
"Equal pay for equal work" my axx. Why my 4p 6174 gets a p8101 with 370kppd and if it gets a p8102 that jumps up to 630kppd? It does the same amount of work by running 24/7 non stop for 10 months. How come my gtx670 gets only 15-20kppd and some lucky guy with gtx570 gets 150kppd? Pande Group need to hire more math researchers instead of part-time students with little math skill.
Last edited:
You DAB rep is sitting on conjecture and has no numbers to back it up.
The case has been made that a MUCH MUCH smaller WU is getting an insane amount of points and we have the data to back it up.
Equal PPD != for Equal work in this case.
Listen to the voices of the past;
Progress, far from consisting in change, depends on retentiveness. When change is absolute there remains no being to improve and no direction is set for possible improvement: and when experience is not retained, as among savages, infancy is perpetual. Those who cannot remember the past are condemned to repeat it - George Santayana The Life of Reason
Progress, far from consisting in change, depends on retentiveness. When change is absolute there remains no being to improve and no direction is set for possible improvement: and when experience is not retained, as among savages, infancy is perpetual. Those who cannot remember the past are condemned to repeat it - George Santayana The Life of Reason
Xavier Zepherious
n00b
- Joined
- Feb 20, 2012
- Messages
- 58
As I stated before this is a benchmark unit (work has already been done)
they compare it against the SMP units(that were done) and the time taken and points awarded
it's the speed it's done in which makes for the big bonuses
(not work across different projects)
the beta team are suppose to give feedback on points (or problems)
not general public getting them and not able to provide it to the beta forum
I still haven't got one....be nice...but the beta team should get them first
if stanford thinks they are too high they will adjust the scheme and do another run
(usually how it goes)
this is why they are beta Units
do I think they are high? maybe - I was going to do straight SMP folding(was planning more units- and bigadv) and this comes out...now I have to wait and see how this pans out
I got burned by the last changes going from 140k PPD(SMP only) for bigadv to SMP(35k) ...per rig
if stanford goes this route then we will have to adapt again. Like some of you guys told me bite the bullet and take it. Hardware and software changes happen...so does the FAH project needs
it's also quite possible that the whole QRB system be rebalanced after this as well...so don't expect things to go to norm right away..stanford has a habit of changing things when we least expect them
you guys are in the better boat with the big rigs you have and being able to now adopt GPU again
it's gonna be a balance of both I think
they compare it against the SMP units(that were done) and the time taken and points awarded
it's the speed it's done in which makes for the big bonuses
(not work across different projects)
the beta team are suppose to give feedback on points (or problems)
not general public getting them and not able to provide it to the beta forum
I still haven't got one....be nice...but the beta team should get them first
if stanford thinks they are too high they will adjust the scheme and do another run
(usually how it goes)
this is why they are beta Units
do I think they are high? maybe - I was going to do straight SMP folding(was planning more units- and bigadv) and this comes out...now I have to wait and see how this pans out
I got burned by the last changes going from 140k PPD(SMP only) for bigadv to SMP(35k) ...per rig
if stanford goes this route then we will have to adapt again. Like some of you guys told me bite the bullet and take it. Hardware and software changes happen...so does the FAH project needs
it's also quite possible that the whole QRB system be rebalanced after this as well...so don't expect things to go to norm right away..stanford has a habit of changing things when we least expect them
you guys are in the better boat with the big rigs you have and being able to now adopt GPU again
it's gonna be a balance of both I think
Skripka
[H]F Junkie
- Joined
- Feb 5, 2012
- Messages
- 10,791
p8054 is another new GPU unit...NOT an a3/4 SMP unit....so no they didn't compare it to an SMP unit.
Given the benchmarking, PG is saying they want SMP folders to just give up, and buy new Nvidia cards. As the ATI client is abandonware yet again. Hurray for subsidizing hardware manufacturers with platform-specific software.
Yes.
I stopped GPU folding about 6 months ago.
I have 3 460s and one 570 that I could fold on. Or maybe I could buy a new 5** or 6** something. I don't pay much attention to GPU advances.
freeloader1969
2[H]4U
- Joined
- Dec 13, 2003
- Messages
- 2,238
With all the shenanigans going on at Stanford, I think I'll just stop Folding altogether.
If PG wants equal work, equal ppd, feed the GPU an SMP WU and compare the TPF.
So far nothing of the sort seems to of happened, Skripka you are spot on.
Skripka
[H]F Junkie
- Joined
- Feb 5, 2012
- Messages
- 10,791
Hell, already before the QRB for GPUs most current gen GPUs at stock get better PPD than current gen 1155 CPUs running SMP at stock (I don't know PPD figs for LGA2011 off hand). Granted people always gripe about the ills that come to their hardware configuration. But srsly. If "equal pay for equal work" between the smp and gpu clients were applied, GPUs would have a PPD cut not a massive boost.
SMP seems to have become almost as much of a red-headed step-child as the ATI client in PPD....Seems in the last 2 weeks the 10% bonus for A4 units was turned off as well.
Xavier Zepherious
n00b
- Joined
- Feb 20, 2012
- Messages
- 58
they are probably doing that
they could also have a standard SMP WU they ported to GPU (most likely)
could have ran it internally(on both) and balance the scheme according to their benchmark machines
once they get comparisons back on tpf and points they can go back to the SMP points & tpf for comparison
you have no Idea what they have done and you are conjecturing
like everyone has said let this play out ...
stanford will look at the results and determine what's right... not you, your team, me or my team
let beta testers give feedback and enjoy the points that come from the odd GPU beta WU you get
Last edited:
Amazed by how many people are going to just dump out of folding (or say they are going to).
Some pretty weak sportsmanship right there.
Financial investments are one thing, but it's just something you have to bite the bullet on and power through with. No one is making you buy or fold anything, it's simply the feeling that you're adding to a cause and in return you get some fun competition between team mates and other teams.
People make careers out of financial investing and their choices don't always pan out either... but they typically make it through. Look at all the people who contribute for team 33. We've regained our #1 spot with a HUGE thanks to bigadv and 4P, it wouldn't've been possible without them. With evga handing out funny money to entice folders... it's just kind of sick.
Atlas folding with his huge GPU farm he once had for example. Thousands upon thousands of dollars that produced once AWESOME ppd, but was completely drowned out with bigadv and eventually 4p setups. You sell your gear for a loss (which is how technology always works) and you move onto the next big paradigm (if you want).
When QRB for GPUs is finalized, only then will people be able to tell if 4P is worth hanging onto. If you quit folding all together... well the rest of us will have a good time continuing to advance a scientific cause.
Don't know what else to add. I could probably blabber some more on each point but the people crossing their arms and saying "I quit" well... that's your choice.
Some pretty weak sportsmanship right there.
Financial investments are one thing, but it's just something you have to bite the bullet on and power through with. No one is making you buy or fold anything, it's simply the feeling that you're adding to a cause and in return you get some fun competition between team mates and other teams.
People make careers out of financial investing and their choices don't always pan out either... but they typically make it through. Look at all the people who contribute for team 33. We've regained our #1 spot with a HUGE thanks to bigadv and 4P, it wouldn't've been possible without them. With evga handing out funny money to entice folders... it's just kind of sick.
Atlas folding with his huge GPU farm he once had for example. Thousands upon thousands of dollars that produced once AWESOME ppd, but was completely drowned out with bigadv and eventually 4p setups. You sell your gear for a loss (which is how technology always works) and you move onto the next big paradigm (if you want).
When QRB for GPUs is finalized, only then will people be able to tell if 4P is worth hanging onto. If you quit folding all together... well the rest of us will have a good time continuing to advance a scientific cause.
Don't know what else to add. I could probably blabber some more on each point but the people crossing their arms and saying "I quit" well... that's your choice.
Skripka
[H]F Junkie
- Joined
- Feb 5, 2012
- Messages
- 10,791
I think most of it so far is pissing an moaning that everyone does when the rules of any game are arbitrarily changed drastically in someone else's favor. Odds are when it gets finalized the value judgement will occur of PPD/Watt will either make or break it.
I'm currently chewing on a 6947 WU with 77,113 atoms on a pair of L5640's with a tpf of 1:52 for 42k PPD. Last time i checked a GPU WU topped out at about 2000 atoms and everyone was bitchin about the heat, power draw and tpf when they got one.
Feed a 570 that 6947 WU and see how it does, if its faster then fine it earns more points, if its slower it earns less. I don't have a problem as the work will be the same. Personally i think the problem is moot as the GPU will melt trying to to process the info
Feed a 570 that 6947 WU and see how it does, if its faster then fine it earns more points, if its slower it earns less. I don't have a problem as the work will be the same. Personally i think the problem is moot as the GPU will melt trying to to process the info