Page 5 of 15

Re: Highwater Mark

Posted: Sun Jun 29, 2008 2:12 pm
by smASHer88
I'm seeing a drop in PS3 numbers against the increase in GPU numbers. Possible that folders that have GPU client setups are now ditching a PS3 they might have had folding?

Re: Highwater Mark

Posted: Mon Jun 30, 2008 4:41 am
by paydirt
Yeah I'm wondering the same thing... PS3s have declined by about 3000 machines. Maybe some folks are looking at PPD/watt and turning off their PS3s. Otherwise it is folks playing Metal Gear Solid 4, or kids moving back home from college (for the summer) and not have re-hooked up their PS3 (and/or not having broadband Internet at home).

Re: Highwater Mark

Posted: Mon Jun 30, 2008 7:07 am
by PlayLoud
I just bought myself a PS3 for my birthday (even if it isn't for a few days still =), and it is folding about 20 hours a day. I think the drop in numbers is just a dip from the high that occurred when GTA IV was released. Hopefully people will turn their systems back on soon. If not, perhaps when the Summer is over, and it is colder.

Re: Highwater Mark

Posted: Wed Jul 02, 2008 4:02 pm
by paydirt
pre-multi GPU for nVidia

OS Type Current TFLOPS* Active CPUs Total CPUs
Windows 200 210173 2076885
Mac OS X/PowerPC 7 8412 117017
Mac OS X/Intel 27 8656 53096
Linux 74 43333 315227
GPU 886 8058 16498
PLAYSTATION®3 1318 46751 551186
Total 2512 325383 3129909

Re: Highwater Mark

Posted: Wed Jul 02, 2008 4:47 pm
by cbuchner1
Interestingly the numbers have dropped a bit post multi-GPU. Apparently people are now changing the configurations of the rigs, taking them offline temporarily.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 1:12 am
by gryoung
Looking at the latest client stats I find GPU client numbers surprisingly low :? , I know that the Nvidia Beta was much anticipated and of course it is a Beta test so many will not adopt the client until it is finalized. But only 16.5K GPU's with only 8K active seems to indicate a lack of interest. Considering that the GPU number also includes the ATI values I really think that something maybe amiss in terms of publicizing the GPU clients. With literally millions of GPUs out there getting only 16.5K even for a Beta test speaks to me of a level of disinterest or simple ignorance of the program.
Of course I maybe completely off base here and this maybe normal as I've never participated in a F@H beta before and these could be perfectly normal values.

Code: Select all

OS Type 	Current TFLOPS* 	Active CPUs 	Total CPUs
Windows 	200	210173	2076885
Mac OS X/PowerPC 	7	8412	117017
Mac OS X/Intel 	27	8656	53096
Linux 	74	43333	315227
GPU 	886	8058	16498
PLAYSTATION®3 	1318	46751	551186
Total 	2512	325383 	3129909

Re: Highwater Mark

Posted: Thu Jul 03, 2008 11:32 am
by ^w^ing
yes, the number of folding gpus is disappointing.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 1:41 pm
by VijayPande
^w^ing wrote:yes, the number of folding gpus is disappointing.
It's still pretty early days, so actually, I'm pretty happy so far. Recall that GPU1 had about 500 donors a steady state. However, GPU1 was limited scientifically and in speed (so less PPD) and was really, really hard to install (lots of driver issues, etc).

The number of donors generally has to do with two factors: PP$ and ease of running. In time, both will likely get better. We're working on ease of running (better desktop responsiveness, multi-gpu, better installation, better core upgrades, improved viewer) and I think we're pretty close on all fronts. Once these sorts of issues are resolved, we'll take the GPU2 client off of the "beta" label and make a big splash/push for it.

Finally, while CPU's aren't getting that much faster, GPU's are continuing to grow in speed aggressively, so every new generation of GPU will bring a jump in points most likely. That will also start to attract more people I suspect.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 4:51 pm
by paydirt
Yeah GPU client is still in beta, so it's probably best to iron out the kinks before marketing/recruiting masses of more casual / "hey this is fun" users who would outright quit if there were too many bugs.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 8:57 pm
by ^w^ing
I know that its early beta, and I can imagine that many people couldnt put up with the 2D lag, but still, these numbers are pretty low imo. I have no doubts that it will get much better tho. I just expected more.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 10:06 pm
by MtM
VijayPande wrote:
^w^ing wrote:yes, the number of folding gpus is disappointing.
It's still pretty early days, so actually, I'm pretty happy so far. Recall that GPU1 had about 500 donors a steady state. However, GPU1 was limited scientifically and in speed (so less PPD) and was really, really hard to install (lots of driver issues, etc).

The number of donors generally has to do with two factors: PP$ and ease of running. In time, both will likely get better. We're working on ease of running (better desktop responsiveness, multi-gpu, better installation, better core upgrades, improved viewer) and I think we're pretty close on all fronts. Once these sorts of issues are resolved, we'll take the GPU2 client off of the "beta" label and make a big splash/push for it.

Finally, while CPU's aren't getting that much faster, GPU's are continuing to grow in speed aggressively, so every new generation of GPU will bring a jump in points most likely. That will also start to attract more people I suspect.
I am waiting till it comes out off beta, going to try and rub it under people's nose as much as I can without getting on their bad side.

And you're right, I'm seeing allot of fresh blood and though I don't think it's mainly the points, there is an certain aspect to it. Allot of teams have grown big and have huge member lists, active or inactive. With the increased points, folding has goten more attention because there is more to talk about. I am doing the same amount off work in a month now which took a year before, and I'm passing and getting passed more and more to. Maybe it's harsh on some off the people who in the past spend their entire devotion on this project to aquire a million points, something Joe sixpack with a quad core and 9600gt can obtain without trying to hard, but it's made it more appealing to new people.

Even people who fold for a cure should be excited, as scientific value and points are one and the same thing.

Re: Highwater Mark

Posted: Thu Jul 03, 2008 11:43 pm
by codysluder
^w^ing wrote:I know that its early beta, and I can imagine that many people couldnt put up with the 2D lag, but still, these numbers are pretty low imo. I have no doubts that it will get much better tho. I just expected more.
You sound like you believe that the 2D lag is an unfixable bug. I'd rather wait to see what changes come during the rest of beta. They've already made some pretty big changes in that area but that doesn't mean they're done. (and it does depend on whether you're talking about NVidia or ATI, though to some degree it's probably true for both.)

Re: Highwater Mark

Posted: Fri Jul 04, 2008 1:34 am
by ^w^ing
codysluder wrote:You sound like you believe that the 2D lag is an unfixable bug. I'd rather wait to see what changes come during the rest of beta. They've already made some pretty big changes in that area but that doesn't mean they're done. (and it does depend on whether you're talking about NVidia or ATI, though to some degree it's probably true for both.)
No, I never thought its an unfixable bug :) I saw the improvements over core versions and with 1.07 I have little to eventually complain about. By 'I just expected more' I meant that even with these issues, I expected more gpus active folding in this early beta.

Re: Highwater Mark

Posted: Fri Jul 04, 2008 2:36 am
by paydirt
It's really up to contributors to get more people interested in F@H, but I would wait until it is more user friendly. Instead of looking down on the number, that's 840 teraflops that we didn't have before, and it's just a number, AND you can do something about it.

You can get other people fired up about it and it can multiply. PS3 numbers used to be half of what they are now... Here are some strategies:

***FAMILY

Most folks have had family members die of cancer or be affected by Alzheimer's. It would have been awesome if my grandfather didn't get Alzheimer's or dementia because I would have liked to have spent more time with him when he was mentally with it. I had just become mature enough to really want to know about life, yet he was incapable of holding a conversation. If we could have a big breakthrough with Alzheimer's, we could greatly improve quality of life for older people (one day that will be you and I).


SOMEONE WHO IS INTO SCIENCE

Simply put, biology is not yet to the point where chemistry is now. With chemistry, you can precisely predict what will happen when you mix two chemicals and cause them to react. With biology, you may be able to describe what is happening if you can observe it, but there are too many things interacting to make it a predictive science. It is believed that if you can utilize enough computing power, you can eventually predict how proteins fold (folding@home) and what 3D shape proteins take (rosetta@home). It is also believed that there is no such thing as disease--what is happening is proteins either mis-folding or misbehaving. Through studying folding (F@H), it is believed we can come to a greater understanding of alzheimer's and huntington's "disease". Through studying 3D shapes (R@H), it is believed we can know the 3D forms that cancers take and be able to create proteins to cancel out cancers.


PRECEDENT (COMPUTER NERDS, that's me too)

If Folding@Home could have enough computing power to make a breakthrough or help other scientists have breakthroughs, then that would be huge for distributed computing (DC) and maybe we could have breakthroughs with other DC projects with cancer, climate understanding, universe understanding, etc.


All it takes to talk to friend or family or a stranger or co-worker, etc... is to be vulnerable and open yourself up and be inspiring as to the WHY you fold and you want others to fold.

Approaching 1PFLOP

Posted: Sun Jul 13, 2008 1:47 am
by bowman
http://fah-web.stanford.edu/cgi-bin/mai ... pe=osstats

The GPUs are approaching 1PFLOP.. Will there be any fanfare? :wink: