Go Back   Hardware Canucks > HARDWARE CANUCKS COMMUNITY > HardwareCanucks F@H Team

    
Reply
 
LinkBack Thread Tools Display Modes
  #41 (permalink)  
Old November 17, 2011, 07:52 PM
3.0charlie's Avatar
3.0 "I kill SR2's" Charlie
F@H
 
Join Date: May 2007
Location: Laval, QC
Posts: 9,667

My System Specs

Default

From Folding Forum • View topic - PG announcement about Big Advanced Projects:
Quote:
Current deadlines are 4 days, 5 days and 5.6 days for 6900/1, 6903 and 6904 respectively.
Halving them gives 2 days, 2.5 days and 2.8 days respectively.
TPF need to be lower than 28 minutes, 36 minutes and 40m15s respectively for QRB to apply
So. Deux, a 2p 16-core rig with 6128s has the following average tpf for BAs:

6901s: 20:02
6903s: 45:20
6904s: 1:05:41

This means that 6903s and 6904s will be out of reach for this 2p rig, with the base 8-core G34 chips if PG applies both the 16-core limit and twice-as-fast preferred deadline requirements. Now, if PG applies the 75% rule to the 6903s and 6904s (since they are BA12 WUs, unlike the 6901s which are BA8s), then Deux could complete the 6903s and 6904s with ease (min. 54min and 97min respectively). There would be a points (credits) drop though...

It would be interesting for other folders to post their BA tpfs (see Benchmarks in HFM).

This same user has the following benchmarks and hardware, for reference:

Quote:
Now my fast 24 thread machine has the following TPF

13m50s for 6900/6901 Fine no problems
28m29s for 6903 Cutting it close
41m20s for 6904 Misses the deadline

Now this machine has 2 x5670 xeon processors running @2.93Ghz + turbo under linux for max performance
--------------

Here's where I have an issue with those changes. 1st, the last k-factor drop which impacted the big brass folders. Imagine this, a rig capable of a 1M ppd was at our door steps. Still, this change impacted a very small number of folders. Now this proposed change - again, BA is impacted, ppd drops are expected. But this time, the folding population impacted by this change is broader. 1p heavily oc'ed rigs (970/980/980x) were managing some GREAT ppd/w/$ numbers, and were selected by many folders as their hardware of choice.

PG will loose folders with this change. Expecting half the ppd (going from BA to SMP WUs) for those 1p rigs is not acceptable, and the funds invested are too high for the ppd returned. Yeah, I hear those shouting "It's for the Science!". Yup. I'm the first one to say it. I fold for the Science. But let me ask all of you, if it would not be for the competitive side of Folding, and the points system, would you be Folding AS HARD? I'm not sure I would be, and in return this would mean less WUs folded, less Science completed...

Get my drift? We are volunteers. PG needs us. A point system is great for competition, and for returning WUs much faster than ever anticipated. Keep us happy, and the Science will win. Unfortunately, with that FF thread, I'm not convinced they are going in the right direction.

My 0.02$.
__________________
Hydro-Quebec is salivating...
Reply With Quote
  #42 (permalink)  
Old November 17, 2011, 08:33 PM
LCB001's Avatar
Folding Captain
 
Join Date: Feb 2008
Location: Aylmer QC.
Posts: 1,774

My System Specs

Default

Quote:
Originally Posted by 3.0charlie View Post
From Folding Forum • View topic - PG announcement about Big Advanced Projects:
So. Deux, a 2p 16-core rig with 6128s has the following average tpf for BAs:

6901s: 20:02
6903s: 45:20
6904s: 1:05:41


It would be interesting for other folders to post their BA tpfs (see Benchmarks in HFM).
SR-2 2x5520 @3.7

6901: 15.53
6903: 35.27
6904: 52.10

SR-2 2x5639 @3.6

6901: 11.42
6903: 27.21
6904: 37.46

Quote:
Get my drift? We are volunteers. PG needs us. A point system is great for competition, and for returning WUs much faster than ever anticipated. Keep us happy, and the Science will win. Unfortunately, with that FF thread, I'm not convinced they are going in the right direction.

My 0.02$.
Well said, even though I don't like them I can deal with the changes, the attitude Stanford has I could do without.

Do I do it for the Science and not the points, absolutely, but Stanford is not the only science project out there and some of the others actually treat their volunteers with respect and appreciate their effort.

This is one reason the rest of my systems run Boinc projects, running F@H is starting to feel a lot like going to work, and I'm starting to look forward to the weekend...
__________________
Folding For Team 54196

Reply With Quote
  #43 (permalink)  
Old November 17, 2011, 08:58 PM
Hall Of Fame
F@H
 
Join Date: Apr 2011
Location: Montreal
Posts: 1,596

My System Specs

Default

People are going ballistic on the official FAH forum, my PPD production already took a hit, I am no longer doing 20k ppd but more like 12k, and one day in 3 I am getting 20, because then is when my CPUs finish the SMP. in this light I think i might return to GPU folding, slam 2 9800GTs or 2 low profile GT430 for better airflow. I wonder how the next Chimp Challenge will look like.
__________________
Reply With Quote
  #44 (permalink)  
Old November 17, 2011, 09:00 PM
Dead Things's Avatar
Hall Of Fame
F@H
 
Join Date: Oct 2008
Location: Centre of the Universe
Posts: 1,572

My System Specs

Default

Overall, I agree with your assessments 3.0charlie, as too does the larger folding population I believe. And you know me - I rarely withhold the urge to shoot my mouth off, especially when it comes to folding. But I don't feel the urge to shoot my mouth off regarding this change, and the reasons why I am at peace with this decision can be explained in part by responding to your post, so if I may...

Quote:
Originally Posted by 3.0charlie View Post
Here's where I have an issue with those changes. 1st, the last k-factor drop which impacted the big brass folders. Imagine this, a rig capable of a 1M ppd was at our door steps. Still, this change impacted a very small number of folders. Now this proposed change - again, BA is impacted, ppd drops are expected. But this time, the folding population impacted by this change is broader. 1p heavily oc'ed rigs (970/980/980x) were managing some GREAT ppd/w/$ numbers, and were selected by many folders as their hardware of choice.
I am a fairly recent folder, having started in January 2008. Yet I remember well that things were not always as they are now. Take MPI-gen SMP for example. It was a waste to fold with anything more than a dual-core because the client didn't scale well and was buggy besides. GPU folding with ATI cards was hitting the mainstream anyway - so the best rig to build was one with multiple PCIe slots (hella expensive back then) and a bunch of $400 ATI X1950's. You could skimp on the CPU 'cuz it wouldn't be folding anyway. Such a rig would have cost at least as much (if not more) then as a typical 2600k rig costs now. Yet when folders who invested heavily in those rigs were literally left behind with the 100% abandonment of GPU1 it raised little furor in comparison to the current situation. The hardware and the science had both evolved (as they have now). The changes made sense (as they do now).

Quote:
Originally Posted by 3.0charlie View Post
PG will lose folders with this change. Expecting half the ppd (going from BA to SMP WUs) for those 1p rigs is not acceptable, and the funds invested are too high for the ppd returned.
There's really no disputing this fact. A large number of folders are indeed likely to leave the project because of this change. It will be a short-term loss for the project. So it begs the question, why is PG willing to accept such a short-term loss when to them, only the science matters. I mean really - how could PG really care about the points? But they do and I'll tell you why...

Quote:
Originally Posted by 3.0charlie View Post
Yeah, I hear those shouting "It's for the Science!". Yup. I'm the first one to say it. I fold for the Science. But let me ask all of you, if it would not be for the competitive side of Folding, and the points system, would you be Folding AS HARD? I'm not sure I would be, and in return this would mean less WUs folded, less Science completed...
Never has a truer thing been said than this. The whole project exists today because of the points system. Absent of a points-based system of competition, FAH is a dead DC experiment long ago. This is why PG cares deeply about the integrity of the points system. If the points system were lacking in integrity, it would undermine the competitive aspect of the project resulting in its eventual demise. The points must necessarily be at least an approximate proxy for the scientific contribution. But, as is the case with nearly all incentive structures that have ever existed in human history, FAH has had its bouts with both unintended consequences and perverse incentives. PG has always done the right thing and corrected those situations whenever they arose, much to the chagrin of those seeking points of course. And this situation truly is no different, not even in scale or cost.

Quote:
Originally Posted by 3.0charlie View Post
Get my drift? We are volunteers. PG needs us. A point system is great for competition, and for returning WUs much faster than ever anticipated. Keep us happy, and the Science will win. Unfortunately, with that FF thread, I'm not convinced they are going in the right direction.

My 0.02$.
The project obviously relies on donors and we come out en masse because of the competition and camaraderie and team spirit all that good stuff (and maybe even a little bit for the science too). But I don't agree that a perversion of the point system is necessary to keep us happy. As I said earlier, there must be some correlation between science and points, and I think that if we're honest with ourselves we can all acknowledge that bigadv got away from that ideal somewhat. There is simply no truth to the perception that a chip is more than 2x productive when folding bigadv than it is when folding SMP. No truth at all. And yet that's what the points say. Clearly something had to give.

What PGH has done differently this time around is communication. They told us change was coming. And they did it early enough to prevent a large number of people from buying brand-new hardware that would have been affected by the change - namely SB-E and, to a lesser extent, Bulldozer. Is it because of this communication that this decision has caused so much more of a commotion within the folding community than the abandonment of GPU1 did? Perhaps. But would you rather it the other way?
__________________
Follow my folding, mining & benching shenanigans @dt_oc!

Think you can overclock? Then show us what you got!
Join the Hardware Canucks Overclocking team today!

Last edited by Dead Things; November 17, 2011 at 09:53 PM.
Reply With Quote
  #45 (permalink)  
Old November 18, 2011, 05:14 AM
3.0charlie's Avatar
3.0 "I kill SR2's" Charlie
F@H
 
Join Date: May 2007
Location: Laval, QC
Posts: 9,667

My System Specs

Default

If indeed the point system is that important for PG - and we all know it is flawed, then why is there no revision, with this upcoming change... why impact a group of folders dedicated with hardware well above the typical home user? Why not make the SMP WUs more interesting, and get that race going again?

I will take a decision shortly, but I just may support other DC projects coming January 16th.
__________________
Hydro-Quebec is salivating...
Reply With Quote
  #46 (permalink)  
Old November 18, 2011, 05:16 AM
chrisk's Avatar
Folding Captain
 
Join Date: Jul 2008
Location: GTA, Ontario
Posts: 7,406

My System Specs

Default

DT I think folks were more accepting of the dramatic swing in PPD from GPU to CPU a short while back because of the efficiency in electricity CPU folding brought.

But yes, from what I can tell this change was brought about due to the lack of work for bigadv, and the mass migration of folks away from other folding methods and the science those WUs did.
__________________
Fold for team #54196
Reply With Quote
  #47 (permalink)  
Old November 18, 2011, 09:31 AM
Dead Things's Avatar
Hall Of Fame
F@H
 
Join Date: Oct 2008
Location: Centre of the Universe
Posts: 1,572

My System Specs

Default

Quote:
Originally Posted by 3.0charlie View Post
why is there no revision, with this upcoming change...
For the same reason why Barry Bonds will always have an asterisk beside his homerun title - his results are not directly comparable to those of his steroid-free peers.

The issue here is that the SMP, GPU, unproc and PS3 remuneration systems are not broken - bigadv's is. There is, after all, a reason why bigadv is beta, while all the rest are public release. So why would they "fix" something that ain't broke just to avoid fixing something that is? They wouldn't. Hence, their decision - while certainly displeasing to the likes of you and me - is clearly the only choice that makes any sense in this scenario.

Quote:
Originally Posted by 3.0charlie View Post
I will take a decision shortly, but I just may support other DC projects coming January 16th.
I hope that's not the case. I'm biased towards folding, so obviously I'd like you to stay with the project. But I understand. Folding most certainly isn't the only DC game in town, so whatever you decide to do, I salute you for all you've done.

__________________
Follow my folding, mining & benching shenanigans @dt_oc!

Think you can overclock? Then show us what you got!
Join the Hardware Canucks Overclocking team today!
Reply With Quote
  #48 (permalink)  
Old November 19, 2011, 04:51 PM
BrainEater's Avatar
Hall Of Fame
 
Join Date: Mar 2007
Location: Calgary
Posts: 2,527
Default

I agree fully DT.

It'll be interesting to see how the 'bonus-point' thing plays out.

....And they are taking a 2-pronged approach to the regular SMP/-bidadv disparity.....They are both dropping the -bigadv bonus , and adding bonuses to all other WU types....They are taking the best course I think.If regular SMP gets a bonus now too , it will fill a gap thats been sorely obvious, but we'll have to wait and see.

------

And don't say that John.

I know you will continue to fold , even if changes happen.

__________________
Intel 3930k /rIVe/32 Gb vengeance LP/Nvidia TITAN/760/760/Intel 520's/WD raptors/etc...
Reply With Quote
Reply


Thread Tools
Display Modes