New 2665 WU's

Honestly I think what they need to consider is they set a standard with the 2653's, 3605, 3602, etc and they avg out to about 3.5k PPD for a 3.2Ghz Quad. With that being said they need to work backward into the new base PPD for a quad. I think 1760 if used with this new core results in my quad churning 1450PPD something is out of wack... so maybe teh bench needs to be something in the neighborhood of 1760 per Dual core... I mean they set the standard by allowing points like that and now they want to change it all... I'm about 6k invested in F@H I'd say that pretty much blows... Again consistency.. if you pay a man a 35k salary for his job and come back a week later and say oh sorry that was a glitch and really I'm only going to pay 15k I'm guessing that man would quit on the spot. They really need to think of the impact of things like this. Either a quad does 3.5k worth of science or they should never had released the prior core. I know betas have issues, but um they pull betas that are so far out of wack they break the game or in this case the measure of science done/benefit to F@H!

 
7im said:
We'll have to see what the 8 core 16 thread Nehalems with no Front Side Bus and tri-channel DDR3 memory brings us later this fall.
Click to expand...

I think it will bring me a large hole in my wallet unless they can't be overclocked as rumored. :D




 
I understand the outrage felt here (I've been folding a long time too), and Stanford should have done a better job, either in setting the points on the old scale, or communicating why the points were set differently on this series of 266x WUs. That's a given, and will become apparent why at a later date, as Xilikon can attest.

But on the other side of the street, how to you reconcile your outrage when things like this are clearly stated in the SMP FAQs?

1. This is a beta client, anything goes.
2. Bonus points can change at Stanford's discretion, so anything goes there too.
3. If you or your machine cannot tolerate even the slightest instability or problems, do not run a beta client.

Sorry for appearing like a moronic facist fanboy, Relic, but maybe you can answer this one for me. Thanks.
 
SmokeRngs said:
Actually, right now I would still buy nVidia video cards even if they were outperformed by ATI cards. Mainly because in my uses, ATI cards generally get their asses handed to them in Linux compared to nVidia. Until those driver issues with ATI change, no ATI for me.

The only way I wouldn't buy a new nVidia card is if they bring out another "dustbuster" FX5800. At that point I would just wait for the next generation.

Also, price can make a big difference in regards to building machines just for folding. My last three systems were Intel quads because I had the money and they had the best performance. If the current ATI folding client and work units had worked on the ATI cards back when I bought those systems, I may have changed my purchases. Back in November I purchased an X2 system due to some great deals I got. I wanted a quad to replace my E6400 at the time but I was able to build the X2 system cheaper than the price of a Q6600. I still get around 1350PPD out of the X2 running the SMP client. Basically, price vs folding performance can make a huge difference in some of our purchases.

As it sits, it's almost worth grabbing an ATI HD3870 to drop in my X2 system with the points they are putting out right now. It would require me to install XP on that system but I could run the GPU client as well as a single regular client for a nice increase in PPD over what the system is currently producing.

These are the reasons we need transparency and consistency. Some of us are forking out thousands upon thousands of dollars to build folding rigs. Some others aren't spending as much since they can't afford it but they want to get the most out of what they can afford.

Click to expand...

Just an FYI for those that don't know...I am one cheap ass bastard when it comes to my folding computers. THe lower the cost, the better ;)
 
7im said:
I understand the outrage felt here (I've been folding a long time too), and Stanford should have done a better job, either in setting the points on the old scale, or communicating why the points were set differently on this series of 266x WUs. That's a given, and will become apparent why at a later date, as Xilikon can attest.

But on the other side of the street, how to you reconcile your outrage when things like this are clearly stated in the SMP FAQs?

1. This is a beta client, anything goes.
2. Bonus points can change at Stanford's discretion, so anything goes there too.
3. If you or your machine cannot tolerate even the slightest instability or problems, do not run a beta client.

Sorry for appearing like a moronic facist fanboy, Relic, but maybe you can answer this one for me. Thanks.
Click to expand...

As discussed and pertaining to your comment on instability, Prime95 is no longer the ultimate test, at best is has become only a base line.

That said perhaps you could discuss with the boys over at Stanford putting together a test work unit that would simply loop. Not a lot of work involved and it could be distrubited as an ISO so everyone would have the same benchmark.

Just a thought
 
We're getting off topic, but hey, it's your forum. ;)

I've pushed for a standardized folding benchmark since day one. But with such a diverse lineup of clients and platforms, I'm sure you can see how difficult it would be just to write, let alone maintain. And just like Sunin's comment about not knowing what produces the best bang for the buck, that also illustrates how hard it would be to keep a Stanford Benchmark tool up to date. That's a lot of time for the project staff to be away from doing the science of the project. We'd love the tool, but a benchmark tool doesn't directly help make the project go faster. Making an NV GPU client is probably higher on the priority list. ;)

TechReport has a benchmark ISO image of some standard FAH WUs they collected. But last I checked, it lacked an SMP WU.

Gromacs.org has the StressCPU v2.0 tool that stresses better than Prime.

Gromacs.org has also gotten gromacs code added to the SPEC 2006 benchmark test suite, but it doesn't correlate to folding PPD very well. It can be used to compare raw CPU performance, which does translate to FAH performance. But it's a pain to run, and not something a typical user could do.

Like you now know with the p2665 issue, there aren't always easy answers, even to finding a simple benchmark.


EDIT: P.S. I'm flattered to have a whole thread dedicated to me. :D
 
I think the bottom line to what I'm reading here is that if the investor (read folder) doesn't receive what they perceive as a sufficient ROI on their time and the amount of $$$ spent on maintaining their hardware, then Stanford will see a decline in interest / investment. That being said, I also agree that Stanford needs to standardize their points system to reflect consistancy across the spectrum. It has been pointed out several times in this thread alone how much of a difference exists in points depending on the WU. Why can't this be adjusted to reflect more consistancy? Who defines the points per WU scheme that is currently in place?

Ax

 
7im said:
Stanford defines a standard benchmark process in the FAQ for each client.

CPU Client Benchmark process
GPU2 Client Benchmark
PS3 Client Benchmark
SMP Client Benchmark
Click to expand...

Let's use your links to bring what we are trying to say :

Our goal is consistency within a given definition of a reference machine setup (described above), but beyond that, the natural variation from machine to machine and WU to WU will never allow any point system to perfectly predict what you get on your machine.
Click to expand...

 
7im said:
Stanford defines a standard benchmark process in the FAQ for each client.

CPU Client Benchmark process
GPU2 Client Benchmark
PS3 Client Benchmark
SMP Client Benchmark
Click to expand...

I'm not sure I buy into the explaination that Stanford gives on how the SMP WU points are assigned. Read through this post by Nova quoted below, and tell me how this senario is possible if Stanford follows their own guidelines for determining points:

novadude said:
to add some fuel to the fire, here are some numbers from a q6600 at 3.4GHz, Vista64, 1 instance of the 5.91 beta client:

project 2653's were running at ~8.5 minutes per 1%, FahMon PpD was ~2980

project 2665's are running at ~12.5 minutes per 1%, FahMon PpD is ~1470

the 2665's are using 94-140 MiB's of RAM, while the 53's were sitting lower at ~65MiB's IIRC


so for approximately 1.5x the time you get half the points

and for what it's worth, Stanford sure as hell needs to respect the people devoting time/energy/computers to do their work. without the people in the f@h community selling them their computing time for only internet points in return, they'd be stuck on their asses trying to get funding for a supercomputer. So that said, 1 billion internet points to everyone that sends me a new/slightly used Power6 rack :p
Click to expand...

If this senario is true, then it seems to me that somebody at Stanford - whoever might be responsible for setting up and following those guidelines - isn't doing their job properly. I'd have to call BS...

 
Feel free to use those links. Those beta client FAQs also say something like this:

1. This is a beta client, anything goes.
2. Bonus points can change at Stanford's discretion, so anything goes there too.
3. If you or your machine cannot tolerate even the slightest instability or problems, do not run a beta client.

No easy answers, are there?


Look, there is more going on behind the scenes than I can say. Xilikon can attest to that, but won't step up in front of his [H]omies and look like a traitor. I don't blame him. BillR knows more than he is saying also, and still rails me in the Demod 7im thread. Whatever.

Anyway, demod me or ban me, do whatever you feel you must do. Hopefully you will be just as enthusiastic to reinstate me when this gets cleared up.
 
7im said:
Feel free to use those links. Those beta client FAQs also say something like this:

1. This is a beta client, anything goes.
2. Bonus points can change at Stanford's discretion, so anything goes there too.
3. If you or your machine cannot tolerate even the slightest instability or problems, do not run a beta client.

No easy answers, are there?
Click to expand...

Hiding behind a disclaimer to avoid presenting qualified information doesn't help anyone out there who is trying to help Stanford, or this whole project. :mad: How about some *real*, qualified information???

 
Axdrenalin said:
Hiding behind a disclaimer to avoid presenting qualified information doesn't help anyone out there who is trying to help Stanford, or this whole project. :mad: How about some *real*, qualified information???

Click to expand...

Send a PM to Xilikon or BillR, so they can clue you in without looking bad in public. As a Mod, I'm not supposed to share more than I have. Then again, that may not last long either.
 
7im said:
I understand the outrage felt here (I've been folding a long time too), and Stanford should have done a better job, either in setting the points on the old scale, or communicating why the points were set differently on this series of 266x WUs. That's a given, and will become apparent why at a later date, as Xilikon can attest.

But on the other side of the street, how to you reconcile your outrage when things like this are clearly stated in the SMP FAQs?

1. This is a beta client, anything goes.
2. Bonus points can change at Stanford's discretion, so anything goes there too.
3. If you or your machine cannot tolerate even the slightest instability or problems, do not run a beta client.

Sorry for appearing like a moronic facist fanboy, Relic, but maybe you can answer this one for me. Thanks.
Click to expand...

LOL wow... um that's just like having the sign that you wave all rights to sue a store by entering... You can't create such BS blanket statements and for them to hold any value in the mind of the contributor!


 
7im said:
Send a PM to Xilikon or BillR, so they can clue you in without looking bad in public. As a Mod, I'm not supposed to share more than I have. Then again, that may not last long either.
Click to expand...

My own somewhat uninformed opinion of this is what are we working on that is so effing secretive?

"I'm not supposed to share more than I have."
Click to expand...

WTF?!? Are the thought police going to come after you if you do? Points are bullshit anyway, supposedly they should have something to do with the value of the science being contributed, but really they are there just to gauge progress for some and grow ePenor for others.

If point values are supposed to be set by a certain benchmark, and they are not equaling out, there is something wrong. 2+2=|=5

The point values of late pretty much mean one of several things:

1) The benchmark has changed, not big deal, the science is still being done.
2) The work units are crap, and need to not be wasting our computer resources.
3) XXXXXXXXXXXX<---Censored by the CIA


Then again, that may not last long either.
Click to expand...

Don't come around here looking for sympathy. Your woe is me bullcrap wont fly around here. You have been running people off of your forum with your pompous talk long enough to make quite a few enemies. Don't come to their back yard talking crap and not expect a reaction.

 
Wheresatom said:
My own somewhat uninformed opinion of this is what are we working on that is so effing secretive?
Click to expand...

Later in life we find out we weren't folding proteins but cracking China's encrypted messages! :)

LOL



 
SmokeRngs said:
These are the reasons we need transparency and consistency. Some of us are forking out thousands upon thousands of dollars to build folding rigs. Some others aren't spending as much since they can't afford it but they want to get the most out of what they can afford.

Click to expand...

I really think this sums it up pretty good. I did a lot of research of F@H when I first started folding. It seems that in the beginning, Stanford simply asked for volunteers to donate their spare CPU cycles. As things progressed, thousands upon thousands have spent thousands upon thousands of $$ on folding only boxen. I would think with the immense increase in contribution form the community, Stanford needs to reciprocate this. This isn't spare CPU cycles anymore.....



 
delvryboy said:
I really think this sums it up pretty good. I did a lot of research of F@H when I first started folding. It seems that in the beginning, Stanford simply asked for volunteers to donate their spare CPU cycles. As things progressed, thousands upon thousands have spent thousands upon thousands of $$ on folding only boxen. I would think with the immense increase in contribution form the community, Stanford needs to reciprocate this. This isn't spare CPU cycles anymore.....



Click to expand...

If I had to guess 90% of the support for F@H comes from the "Spare CPU Cycle" the guy that owns a gaming machines and folds on it, etc...

The other [H]arder 10% have invested as Delvryboy has stated. Want my next contribution... F@H is going to have to demonstrate they know how to communicate and set expectations otherwise I'm done ramping up... shame since I have the structure for 23 folding boxes in total and I'm only 9 into the ramp up!

 
Sunin said:
Wheresatom said:
My own somewhat uninformed opinion of this is what are we working on that is so effing secretive?
Click to expand...
Later in life we find out we weren't folding proteins but cracking China's encrypted messages!
Click to expand...
That or the research can also be applied to the development and manufacture of biological warfare agents. Nice thought, eh? ;) (j/k)
 
APOLLO said:
That or the research can also be applied to the development and manufacture of biological warfare agents. Nice thought, eh? ;) (j/k)
Click to expand...

EXCELLENT!

Stewie imitation off! :)

 
APOLLO said:
That or the research can also be applied to the development and manufacture of biological warfare agents. Nice thought, eh? ;) (j/k)
Click to expand...

I know that you mean that as a joke, but with all this back room bullshit and 7im running around like he's Deep Throat talking about I may have already said too much, what the hell IS going on? This is clearly about a hell of a lot more than the point value of the 2665 work unit.....
 
alan2308 said:
I know that you mean that as a joke, but with all this back room bullshit and 7im running around like he's Deep Throat talking about I may have already said too much, what the hell IS going on? This is clearly about a hell of a lot more than the point value of the 2665 work unit.....
Click to expand...
I could add so much more regarding the secrecy issue from my own experience but I have chosen to stay out of the brewing political cauldron that's heating up regarding this and related matters. That's why I declined replying in the other threads. I don't want any involvement. Suffice to say that what everyone is only finding out now to be highly unusual, I was asking myself well over a month ago....
 
I haven't received any P2665 WUs yet. Are these Linux only?

 
This puts my boxes at about 3000 PPD which is acceptable up from 1450. I like that they are working on some methods to even out the points.

encouraging

 
Sunin said:
This puts my boxes at about 3000 PPD which is acceptable up from 1450. I like that they are working on some methods to even out the points.

encouraging

Click to expand...

Me too. I hope they adjust so they can put the money in the right mouth (best science value) because people will follow the points trail no matter what.

 
I've gotten a few and with the sysytem completely idle it takes about 15 1/2 minutes per frame. When actively using the pc it goes up to ~ 16 1/2 min.

Before everyone gets all upset over the point structure I know I'm not the only one who can remember (not too long ago) when 1,000ppd was a huge milestone and required AT LEAST 4 high-end dedicated folders. If points go down the guys with more boxen will still maintain their 4,5,6x production over the guys with 1 or 2 boxen.

I'm still in this for the science no matter how the points fall. Just a thought :D


 
Lord von Schtupp said:
SpoogeMonkey said:
I'm still in this for the science no matter how the points fall. Just a thought :D
Click to expand...
Totally agree there, I only posted what I was getting as a report of sorts, thats all.
Click to expand...
Seconded.

Moreover, if the point structure is revised by Stanford everyone and every team will be affected, perhaps not equally depending which clients/WUs will be adjusted the most, but most participants will be affected to a lesser or greater degree. Thus, there should be parity with the competition. Our points of reference will only need to change.
 
Sunin said:
Maybe true with the lax deadline but then should we consider running Quad SMPs on a Quad?
Click to expand...

Sunin said:
After the Chimp Challenge I will convert a One Quad to a Single SMP and see what that does.
Click to expand...

Did you every complete your test?
If so what were the findings?
I was thinking that going back to one SMP client per boxen might be the way to start going ?
As of now, it is still better, PPD wise, to run two clients, but with more and more "Chewy" units ... the time will come!!????

Here is what I have seen on mt farm: Based on Q6600 @ 3.4Ghz = 2 x SMP

2605 / 2653 = 2200 PPD x 2 = 4400 PPD
3062 / 3064 = 1980 PPD x 2 = 3960 PPD
3065 = 1600 PPD x 2 = 3200 PPD

On my diskless boxen = if one instance get a WU - the second gets the same WU Series
On Windows Clients , they will mix and match.

IIf the WU are different, the PPD goes way up .
Example:
3065 & 2653 Combination = 1800 PPD + 2150 PPD = 3950 PPD.
3064 & 3062 Combination = 1700 PPD + 1900 PPD = 3600 PPD

3065 or 2665 brings it way down. I think they must be very cache / memory intensive.
 
FYI you've just been lucky with the windows boxes... I've had multiple (read 3 quads) get dual 2665 and at the time that was 725PPD each core for 1450 PPD... now it is up to 1500PPD per core x2.. so that makes those at least digestible.



 
I just received my first 2665. I'll see what this puppy does with the frame times. It's being worked on by my dual-quad. Hopefully the new points allotment won't affect the overall production on that machine much. That system is slated for a single VM trial installation in the near future.
 
Just as a teaser, I got 3 2662 WU with the A2 core and the output is about 4600 ppd with a single client on a Q6600 at 3.0 GHz...

 
So, when do they release the A2 core in the wild for the rest of us folders? ;)
 
Axdrenalin said:
So, when do they release the A2 core in the wild for the rest of us folders? ;)
Click to expand...

Still some bugs to kill. The other side of the medal is that anytime you try to stop the client or restart the computer, it destroy the WU :(

 
You must log in or register to reply here.
Back
Top