PPD restructuring

[coo-coo-clocker]

Well, it's about time!

Points in general (http://folding.typepad.com/news/2008/05/gpu2-points-benchmark-machine-and-points-in-general.html)

I should also remind donors that we've been talking about rebalancing points in general, to try to get SMP, GPU, PS3, and classic points in line with the scientific productivity done by the clients. As long as more points = more science, donors can optimize for points and that will automatically optimize for maximum scientific output.

[DAK1640]

finally...:up:

[SocketMan]

GPU2 increase kicked in today,:up:
EDIT**
Well on a second thought it looks like when a store raises the price by 40$
then says save 20$ seems like it's just back to where it was ~ 2 weeks ago
so the jury is still out on this "increase", will wait for another project to compare..........

[SparkyJJO]

Any idea how long it'll be before everything else kicks in?

[SocketMan]

Negative

Now I am not even sure if the GPU2 point increase is part of this "restructuring".
How knows maybe Nvidia asked for more ppd's:rolleyes:


edit**
http://folding.typepad.com/news/2008/06/gpu2-ppd-increase.html

http://folding.typepad.com/news/2008/06/what-does-the-smp-core-do.html

[MikeB12]

hmmm.. is that 1500ppd 3850 benchmark really an increase as advertised?
how much ppd are your 3850's turning now after the supposed increase, socket....
weren't you already getting like 1700ppd on each 3850 before? http://www.xtremesystems.org/forums/showpost.php?p=3013467&postcount=9

--------------------------------------------
:2cents:
imo, I think smp is in it's "lasts days" for fah enthusiast users... what use does it have when a single gpu2 can dwarf it's scientific output. I mean it's useful today and in the past and everything, but it makes more sense to shift wu focus to the more efficient clients; so over time smp gets phased out by us enthusiast folders... I imagine once the ppd gets restructured and smp gets a ppd reduction to put it on equal output footing with the other cpu clients and proportionate to gpu2, then it won't make as much sense for us to keep all the oc'ed quad farms sucking $$$ down the elec bill... it'll make more sense to put the power consumption into hardware that can crank out the Gflops and ppd (after restructuring). basically the same thing that happened when smp got introduced with high ppd value last year, everybody started building quads, quads, quads... now if gpu2 comes along and does something similar, everybody will be building dual gpu2 farms.. I dunno, just speculation.

[Kingcarcas]

Yup, after all the whining with the SMP PPD decrease, i expect a lot of people will jump ship to other DC projects. GPU is the way to go and i'm sure everyone who stays will make up the output or even set new records :up:

[MikeB12]

yeah, I'm thinking about a 4870x2 for my main rig to replace the 2900pro that's in there now.. then fold on 2 gpu2's on that rig instead of smp...

not sure what I'm gonna do with my other 2 quads, they have 8500gt and 8400gs right now, just budget quad crunchers... I'll probably move them to wcg...

then I'll have a 4870x2 on fah, and 2 budget quad crunchers on wcg...
the rig's in my sig...

it'll be a reduction in my overall ppd for fah, but I'll be crunching on 2 projects instead of just fah, like now..
plus that will alow me to crunch for xs wcg; my fah efforts were already devoted to another team when I became active at xs.. so I've been absent for xs crunching.. been wanting to change that for a while, but didn't want to jump ship on my orig fah team.. now, this may give me a good reason to shake things up..

[sc00p]

You buy a car and drive it gladly for awhile... it was a good performer at it's time. Oh boy, it did cost a bunch of apples back then...
Then comes a rival model. Which goes considerably faster with about the same amount of gas... but of course overall costs could be slightly higher, who knows.
Do you go ballistic/mad that someone had the nerves to develop such a thing and give your current car away (for free) to your friend or even to a stranger... yet you keep buying gas for it?
Next of course in addition you buy this new better car...




Yup, after all the whining with the SMP PPD decrease......I still don't understand it...
I've lost taste of folding.

[SparkyJJO]

...I still don't understand it...
I've lost taste of folding.

Dude don't stop. We lost UnG again, Jimwah seems MIA... don't let a few whiners ruin it for you. Some may complain about stuff but if you understand it all why let them bother you?

[coo-coo-clocker]

yeah, what he said. Adapt, and forge ahead.
I for one am happy that I will finally be able to use my graphics cards. At the risk of ticking off the ATI fans, I still like Nvidia more and chose that when I built my dedicated folding rig. Now I can use them! MUWAHAHAHAHAHAAAAA

[MikeB12]

personally, I think the ppd restructure and accompanying nvidia client will increase all the enthusiast folders' ppd...

if the numbers posted at tom's are correct... how many of you gamers already have 8800gts g92's. even on dual core boxes. 4700ppd. that's a 3000 ppd increase for some of the gamer/folder pc's that were folding on dual cores... or about a 1000ppd increase for gamer/folders on quads... and consider that the quads w/8800gts can now run smp and gpu2, even more... I'm just using a 8800gts as an example, because I know those are mature tech and tons of people already have them, so no money investment needed..

http://media.bestofmicro.com/gtx-280-cuda-folding-at-home,N-H-111005-13.jpg


I wouldn't let the ppd restructure get you down, though it may take some strategy/rethink to remain in the game as a an enthusiast folder. that's how it's panning out for me anyway...

Scoop had his biggest month in May, so he's not out, just on a little sabbatical...
http://img395.imageshack.us/img395/293/59453745bz2.jpg

[angra]

Interesting - I wonder where a 9800gx2 fits on that panel. I have a rig with 2 of those running - with a third waiting in the wings :)

[MikeB12]

http://www.xtremesystems.org/forums/showthread.php?t=191513
just glanced quickly at some of the new info after release today, it looks like it's not ready for dual gpu's or vista 64 yet... but check out the links in the above thread.

faq, fah nvidia gpu forum , and dl link for nvidia gpu2 beta 6.12 (new links just posted at fah this afternoon)
you could probably dl the beta and get a ppd value on the 9800gx, even if it's only half of it for now.

[coo-coo-clocker]

Have we figured out though how much loss in ppd a quadcore/smp will suffer when you add a GPU2 client? Doesn't it take an entire core to drive each GPU?

[direowls]

Have we figured out though how much loss in ppd a quadcore/smp will suffer when you add a GPU2 client? Doesn't it take an entire core to drive each GPU?

Will test later this evening. Didn't slow down one bit with 2 cores on WCG, though. CPU usage is minimal, at least in Vista.

[MikeB12]

if it's anything like the ati gpu2, if you don't feed the gpu2 client a "slightly higher core pririty" setting in the settings of the gpu2 client, then smp chokes the hell outta the gpu2 client... at least that's how the ati one does... I only got like 350ppd on the gpu2 client on a 2900pro without the setting, but if I checked off the higher core priority setting I got like 2000ppd w 3d clocks enabled.

it will be interesting to see. because if you don't have to use a core, then that means the nvidia gpu2 client is not cpu dependent, which seems like a long shot expectation to me. but I really don't know, I don't have any 8800's right now...

[SocketMan]

I don't understand how and why (well I think I know the why) there are comparing 1 project (50xx) that is run on Nvidia to a completely different project run on ati. Yes I know it's ppd's but how do I know that ati wont get
the same ppd if they run the same (as NV) project? Same for NV:
how do I know whether the ppd's wouldn't decrease if 47xx(ATI's projects)
were run on a Geforce?
The "benchmark" machine for NV is the same, says NV faq
http://folding.stanford.edu/English/FAQ-NVIDIA

What I want basically is a project that will run on both NV and ATI.

Because the way it looks right now for many,many ,many people, (who are also potential customers for the gtx280/ati48xx) is that 8800gt
is at least 2 times "better" then 3870 in folding which may or may not be
Seems very unfair to me.

[SocketMan]

hmmm.. is that 1500ppd 3850 benchmark really an increase as advertised?
how much ppd are your 3850's turning now after the supposed increase, socket....
weren't you already getting like 1700ppd on each 3850 before? http://www.xtremesystems.org/forums/showpost.php?p=3013467&postcount=9

--------------------------------------------
:2cents:
imo, I think smp is in it's "lasts days" for fah enthusiast users... what use does it have when a single gpu2 can dwarf it's scientific output. I mean it's useful today and in the past and everything, but it makes more sense to shift wu focus to the more efficient clients; so over time smp gets phased out by us enthusiast folders... I imagine once the ppd gets restructured and smp gets a ppd reduction to put it on equal output footing with the other cpu clients and proportionate to gpu2, then it won't make as much sense for us to keep all the oc'ed quad farms sucking $$$ down the elec bill... it'll make more sense to put the power consumption into hardware that can crank out the Gflops and ppd (after restructuring). basically the same thing that happened when smp got introduced with high ppd value last year, everybody started building quads, quads, quads... now if gpu2 comes along and does something similar, everybody will be building dual gpu2 farms.. I dunno, just speculation.

Exactly :
about 2 weeks ago "new" projects came out and that opty's (link) ppd's went
down from ~3500 to less then 2500 (500ppd per card)

after the "increase" it up to `3800 ppd, but I did put my 3870 in instead of the 3850 (256mb) so it's about the same as it was:(

Also the projects change: they'll run 1 for a week or 2 then switch to another,rarely going back to the "older" ones

[Kingcarcas]

Only people who were whining were the people who had Quad farms, i don't so i will continue to Fold on :D I'll slap my 3870 into the family PC with the X2 and buy a 4870/9900GTX for myself :up: I'll probably have to take my Ps3 offline though....

[angra]

Mike, one thing I haven't yet figured out - when they say the client isn't dual GPU ready, do they mean that you can't run multiple copies, each grabbing a different GPU? Or just that a single client doesn't exploit multiple GPU yet?

[ZOMGVTEK]

Those charts from toms are a little low if you ask me, my 8800GTS 512 is getting 6048 PPD according to FahMon...

But its overclocked a good bit...

[MikeB12]

Mike, one thing I haven't yet figured out - when they say the client isn't dual GPU ready, do they mean that you can't run multiple copies, each grabbing a different GPU? Or just that a single client doesn't exploit multiple GPU yet?


here's the 9800gx2 thread whwere people are discussing it.
http://foldingforum.org/viewtopic.php?f=43&t=3189

I'm assuming when they say multi gpu, they mean multiple clients also.. but who knows, it may be a sli type thing with just one client (that seems a little complicated for a client in this baby stage of developement though). I bet they're talking about running multiple clients on multiple gpu's in the same os installation. dunno, I guess time will tell. I don't see a list of flags that work with the 6.12 beta client, so that's another question mark..


but Bruce did mention something that looks relevant to your situation, someone apparently was getting errors on a gx2 and this is what Bruce posted...

When multiple GPUs are recognized by Windows, the FAH client must be told which one it should be talking to. If it's trying to initialize the board that you've disabled, I would expect a mdrun_gpu returned -1 error or some other equally obscure error.
See what happens if you leave everything enabled.
Then see what happens if you add the flag -gpu 1 or -gpu 2 (etc.) to the command line when starting the client.

[SocketMan]

The ati flag starts with 0 :
-gpu 0 (primary) -gpu 1 (second)
not sure if the flag works for NV

[coo-coo-clocker]

We better step on it!
Team Whoopass is crankin out > 500k PPD!!

[angra]

but Bruce did mention something that looks relevant to your situation, someone apparently was getting errors on a gx2 and this is what Bruce posted...


I was able to get two clients launched, one with "-gpu 1" and one with "-gpu 2", but the WUs each took 2x as long to run. Looks like the flag is not honored on this client.