Page 1 of 1
What'd be the effect of doubling WU deadlines?
Posted: Sun Jul 20, 2008 3:25 pm
by Trivolve
Searched the forum and couldn't find out about the short deadlines.
1. I generally understand that short deadlines are good for continuing the project, but it requires my C2D computer to be 24/7, and i want to donate spare cycles and not burn electricity just to fold. And in that point, why not just multiply the deadlines by 2 or 2.5 to about 5 days, so that many users' computers can be turned on 12 hrs a day, sufficient to complete the tasks but not dedicate 10 hours of sleeping hours to just folding (i think it defeats the purpose of using spare cycles).
2. C2Qs would be able to handle SMP WUs twice as fast, so is there any differentiation there? Like C2Ds can get longer deadline WUs.....
Re: What'd be the effect of doubling WU deadlines?
Posted: Sun Jul 20, 2008 3:39 pm
by Bernie64
Each wu is a fragment of a big picture, and to continue the fold sequence requires getting back each wu complete first.
There already is a 100% leeway on the deadlines anyhow as per below.
http://vspx27.stanford.edu/psummary.html
http://vspx27.stanford.edu/cgi-bin/allprojects shows that a complete project may only be fractions of a second.
Then again the server should check your system specs and give appropriate wu's.
Perhaps run multiple smaller wu's instead of the big ones?
Re: What'd be the effect of doubling WU deadlines?
Posted: Sun Jul 20, 2008 3:44 pm
by Trivolve
SMP does go about 25 times faster than the 5.0x and 6.x clients, about 2 days vs about 50 days. Is that really necessary, so are the 5.0x client projects really going very slow then?
Re: What'd be the effect of doubling WU deadlines?
Posted: Sun Jul 20, 2008 5:30 pm
by HackMan
Trivolve wrote:Searched the forum and couldn't find out about the short deadlines.
1. I generally understand that short deadlines are good for continuing the project, but it requires my C2D computer to be 24/7, and i want to donate spare cycles and not burn electricity just to fold. And in that point, why not just multiply the deadlines by 2 or 2.5 to about 5 days, so that many users' computers can be turned on 12 hrs a day, sufficient to complete the tasks but not dedicate 10 hours of sleeping hours to just folding (i think it defeats the purpose of using spare cycles).
2. C2Qs would be able to handle SMP WUs twice as fast, so is there any differentiation there? Like C2Ds can get longer deadline WUs.....
I totally agree with you.
In my opinion a longer deadline for SMP WU's would increase the whole folding performance, people can use there dual or quad system for full-core folding and thats's only 8-12 hours a day and not for 24 hours a day...
Re: What'd be the effect of doubling WU deadlines?
Posted: Mon Jul 21, 2008 12:51 am
by 7im
It's a slippery slope argument. You may want 2x deadlines, but the next guy then wants 3x. Heck, why not make the deadlines 4x so that single core systems can fold SMP work units?
Long deadlines are not in the goals for having a high performance client. If you want to fold part time, install the CPU client.
Re: What'd be the effect of doubling WU deadlines?
Posted: Mon Jul 21, 2008 8:22 am
by jevans64
Overclock your C2D to 3.8 GHz and your C2Q to 3.4 GHz then you can run them for 14 hours a day.
Re: What'd be the effect of doubling WU deadlines?
Posted: Mon Jul 21, 2008 9:50 am
by Trivolve
nah, not necessarily slippery slope, there would be a sweet point where the project can move on reasonably fast and yet have as many users in as possible, to utilize the "86,000,000x speed up in communication".
Re: What'd be the effect of doubling WU deadlines?
Posted: Mon Jul 21, 2008 11:35 pm
by Bernie64
http://foldingforum.org/viewtopic.php?f=8&t=1783
The beta clients are always short so that they can be very often reviewed.
When they go mainstream they size up quite a lot and take much longer.
I don't think the SMP is a "regular" client yet.
Same applies to the GPU client.
Re: What'd be the effect of doubling WU deadlines?
Posted: Mon Jul 21, 2008 11:36 pm
by bruce
Let's ask a different question.
(Note: This is not something even semi-official, so don't assume it will really happen)
Suppose you got less points if you accepted a WU with twice the deadline? (...or maybe more points if you accepted WUs with a shorter deadline?)
Note that the bonus (or anti-bonus) would not be based on when you actually return the WU, only on whether you accepted a WU with a longer deadline or not, but missing that deadline would still be critical.
Re: What'd be the effect of doubling WU deadlines?
Posted: Tue Jul 22, 2008 4:12 am
by 7im
Trivolve wrote:nah, not necessarily slippery slope, there would be a sweet point where the project can move on reasonably fast and yet have as many users in as possible, to utilize the "86,000,000x speed up in communication".
It is a slippery slope, because no matter where you put the sweet spot, the people just below that sweet spot would want the same thing, just like they are doing now. If the deadline was changed so that a C2D folding 12 hours a day could make the deadlines, then the people only folding 8 hours a day at their workplace would then ask for the same thing. How is this not a slippery slope?
Bruce raises a good question...
Re: What'd be the effect of doubling WU deadlines?
Posted: Tue Jul 22, 2008 12:43 pm
by HackMan
bruce wrote:Let's ask a different question.
(Note: This is not something even semi-official, so don't assume it will really happen)
Suppose you got less points if you accepted a WU with twice the deadline? (...or maybe more points if you accepted WUs with a shorter deadline?)
Note that the bonus (or anti-bonus) would not be based on when you actually return the WU, only on whether you accepted a WU with a longer deadline or not, but missing that deadline would still be critical.
That's a good point and you know what, I would agree with that. My point is, you have a really really great client, special for quad cores (nobody else has that). So I want to use this client but unfortunately I'm not able to run my computer the whole night (and day)... So I can't use this great client because of the short deadlines...
And because of that, I would agree with less points but longer deadline
Re: What'd be the effect of doubling WU deadlines?
Posted: Tue Jul 22, 2008 12:53 pm
by MtM
bruce wrote:Let's ask a different question.
(Note: This is not something even semi-official, so don't assume it will really happen)
Suppose you got less points if you accepted a WU with twice the deadline? (...or maybe more points if you accepted WUs with a shorter deadline?)
Note that the bonus (or anti-bonus) would not be based on when you actually return the WU, only on whether you accepted a WU with a longer deadline or not, but missing that deadline would still be critical.
I never did like the idea off running dual smp's on quads just because I like the lowest possible return times, would be great if the points somehow more reflected that
Re: What'd be the effect of doubling WU deadlines?
Posted: Wed Jul 23, 2008 5:47 am
by bruce
Some time back, the Pande Group said that the points system would be changing but they have given no details about what might change except that it's supposed to make the points reflect the scientific value more accurately. We do know that speed is one important factor that improves the scientific value, so it is a logical assumption that something along these lines MIGHT be developed. (It could also be very different that what I've suggested.)
There's also no information about when the changes might happen, so there's not much point in expecting that they'll be soon enough to satisfy your request.
As a PC owner, you do have the option of running multiple console clients instead of a single SMP client and the deadlines are MUCH longer. The only people who really do have a problem is those who own Intel-based Macs. The only client that will run on their machine is the SMP client, so they have to run all night unless they've got a really fast machine.