nVidia F@H ??

[Insidious]

Originally posted by: 7im

Originally posted by: biodoc...

In the end, it's the science that's important, right?
:beer:

Yes, that's why they do things like QMD work units, or release a Linux SMP client months before a Windows client is possible. It's not to screw AMD users, or Microsoft. It's becasue Stanford DOES follow the science, wherever it leads even if it doesn't please everyone, not because someone pays them to do it.
... snip

Payment can be defined in many ways. An agreement to help is fine, but when it is associated with the understanding that this help is specific to the stuff that makes them money, to the exclusion of the stuff that makes the competition money ...... Aren't there other words commonly used in place of "payment"....?

I don't believe you are in a position to truly know the validity of what you have just written. Yeah, lots of forum posts and probably a side discussion or two, but saying it louder and more times won't make it so....

It is my assertion that the situational realities.. such as follow up that was promised but never happened in more than one situation involving optimizing for more than one hardware manfacturer for a given type of processor. (ie: CPU = intel but not AMD, GPU = ATI but not NV, Console = PS3 but not XBOX (?))

Actions speak much louder than words... even harsh ones. I am not the only participant in this project that would need to see more even handed results to believe in such non-biased motives as you describe.

But yet we crunch.... go figure.

To assert that anyone who is folding must be in total and unquestioning agreement with everything that is F@H or else they are not supporting the project... I say BUNK

-Sid

have a :beer: and relax

 

[7im]

Originally posted by: Insidious...

I don't believe you are in a position to truly know the validity of what you have just written. Yeah, lots of forum posts and probably a side discussion or two, but saying it louder and more times won't make it so....

There is a saying that goes... "A wise man can not reason away what a fool believes."

As mentioned previously, I am pleased to now serve the first of several crow appetizers, this one has a hint of GPU encouragement flavoring... see the Folding@home News link... http://folding.typepad.com/new...1/sc2007-demo-new.html

:evil:

 

[waffleironhead]

folding@home forums still down? And does that "story" say anything about the 8800? didnt think so.

 

[Insidious]

Originally posted by: 7im

Originally posted by: Insidious...

I don't believe you are in a position to truly know the validity of what you have just written. Yeah, lots of forum posts and probably a side discussion or two, but saying it louder and more times won't make it so....

There is a saying that goes... "A wise man can not reason away what a fool believes."

As mentioned previously, I am pleased to now serve the first of several crow appetizers, this one has a hint of GPU encouragement flavoring... see the Folding@home News link... http://folding.typepad.com/new...1/sc2007-demo-new.html

:evil:

I see a lot of ATI hype. No mention of a cross mfgr. client to run on NV GPUS that I could find...... are you indicating you now agree with me? I guess that would be a large crow indeed. :Q

-Sid

 

[7im]

Yes, the Folding@home support forum, sponsored by a private 3rd party so that Stanford can not unduly influence our opinions there, has been down for several days. Feel free to donate and keep it running if you like. Maybe the owner forgot to pay the light bill? Don't know. Sorry, but that's a bit of a tangent, don't you think?

Oh, and I think I did mention something about the first of several appetizers to come... too many at once, and the distaste might turn in to indigestion. Some might actually choke, and we can't have that.

If you want a whiff of the next course to come, try to read between the lines of all that so called ATI Hype, Captain Obvious...

 

[Insidious]

Originally posted by: 7im
... snip

Oh, and I think I did mention something about the first of several appetizers to come... too many at once, and the distaste might turn in to indigestion. Some might actually choke, and we can't have that.

If you want a wif of the next course to come, try to read between the lines of all that so called ATI Hype, Captain Obvious...

And on and on and on it goes.....

"appetizers"..... "read between the lines"..... OH COME ON :|

Don't you get it.... ?

F@H does what it needs to for the project to continue.... That's fine

But it's all the nonsensical, weenie stuff... the false promises and enuendo... the denial of fact, etc...

I cannot find credibility in what I've read here today.

-Sid

 

[Rattledagger]

Originally posted by: 7im
Oh, and I think I did mention something about the first of several appetizers to come... too many at once, and the distaste might turn in to indigestion. Some might actually choke, and we can't have that.

If you want a whiff of the next course to come, try to read between the lines of all that so called ATI Hype, Captain Obvious...

Well, as far as Ati goes, it's very unlikely my X700 will work with whatever new GPU-client, and as far as non-Ati, it's unlikely G33 will work either...

But, being terrible at reading between the lines, I didn't see any mentionings if the GPU-client finally not needing one cpu-core to run effectively any longer.

As for new GUI, nice-looking.

How it will actually perform is another matter...

 

[waffleironhead]

Originally posted by: 7im
Yes, the Folding@home support forum, sponsored by a private 3rd party so that Stanford can not unduly influence our opinions there, has been down for several days. Feel free to donate and keep it running if you like. Maybe the owner forgot to pay the light bill? Don't know. Sorry, but that's a bit of a tangent, don't you think?

Oh, and I think I did mention something about the first of several appetizers to come... too many at once, and the distaste might turn in to indigestion. Some might actually choke, and we can't have that.

If you want a whiff of the next course to come, try to read between the lines of all that so called ATI Hype, Captain Obvious...

holy god man! It was a simple question, seems it was around the same time that the fah forums went down that we suddenly see you here. Also seems to me to equate that when the fah forum is back up you'll be back there.

The second part of my post was along the lines of the op. We still dont see any offical support of the nvidia hardware. That site you linked to was all about how great we(fah-amd/ati) are. With NO mention of nvidia.

oh, and the gui is pretty and all, but im a console type person as I dont really care about eye candy.

and for the record I dont care if there is ever an nvidia capable client.

 

[Assimilator1]

7im
Interesting read that ,I look forward to the new GUI
But I see nothing about nvidia or 8800s ,hint ,between the lines or otherwise

waffleironhead
How do you know you won't have an Nvidia card in the future?


Btw folks what are QMDs?

 

[waffleironhead]

Originally posted by: Assimilator1

waffleironhead
How do you know you won't have an Nvidia card in the future?
q]


good point , ok lemme rephrase that, seeing as my upgrade cycle is/has been every 5 years, and I a bout 2 years into this system, Quote me as, "I dont personally care if there is a nvidia clint for at least the next 3 years".

 

[Insidious]

Originally posted by: Assimilator1
7im
Interesting read that ,I look forward to the new GUI
But I see nothing about nvidia or 8800s ,hint ,between the lines or otherwise

waffleironhead
How do you know you won't have an Nvidia card in the future?


Btw folks what are QMDs?

QMDs were a type of Work Unit that would only be assigned to Intel Processors.

It was kind of wierd that "licensing issues" were insurmountable, but the real flap was about a scoring of these work units that was nearly 4x ( edit: when compared to a non-bonused work unit which were the vast majority of what was being assigned at the time to non-intel platforms) the points in a day that the non-intel folks could have.

People would complain and 'certain' folks at the F@H forum would turn the topic over to licensing or offer up ridiculous rationale for why it was fine that the "benchmarking" methods they used for setting scores gave this weighting...

Worst of all (IMO) was a persistant LIE that the Stanford team was trying to make them run on AMD platforms... history proved what fools we were to believe that! {So now I've also explained my disgust for terms like "appetizers" and "read between the lines". }

many people quit F@H (myself for over a year) and finally the QMD Work units were dropped. Again funny that Stanford would rather abandon them all together than simply set the scoring to a point that they were equitable... but far be it from me to call them 'short sighted' or biased by side agreements preventing a cross platform client, or anyting like that.....

-Sid

 

[Rattledagger]

Originally posted by: Assimilator1
Btw folks what are QMDs?

As you're maybe aware, Folding@Home has multiple applications, the really old-one was Tinker, the most common is Gromacs (SSE-optimized), and there's Amber and so on.

For a couple years ago, they had a few QMD-projects using a "QMD"-core of it's own, don't remember if it had another official name, but in any case, it was a SSE2-optimized, and since Intel with accompanying libraries is the "best", for speed-reasons the QMD-core relied on Intel. But, while the Intel-combination is excellent on Intel-cpu's, atleast some of their compilers has also a tendency to explisitely test for intel-cpu, and if non-intel, the program either crashes, or non-optimized portions is run instead...

Meaning, for speed-reasons, Folding@Home choose to only send QMD-wu's to P4-computers.

Now, if this had been the full story, my guess is there wouldn't have been much controversy with the QMD-wu's only being sent to Intel-cpus. But, while FAH benchmarks all wu's on a common computer before deciding how many "points" a wu is worth, they're screwed-up their own system with:
1; Even for QMD that only was sent to SSE2-computers, all wu's is benchmarked (*) with only SSE enabled. Meaning, QMD immediately gave higher points than benchmarked.
2; FAH gives "big-wu"-bonuses, either due to high memory-usage, or because the client sits idle doing large uploads/downloads. QMD was "big".
3; FAH gives bonuses for "advmethods", wu's that either is just out of beta or for beta-cores, QMD was "advmethods".

So, my recollection is really fuzzy, and it's quite possible it's wrong, but has a hazy idea the bonuses before was 100%, while they're now "only" 50%. Others will be better remembering this point.

In any case, while QMD was active, any P4 with enough memory and big/advanced configured was basically guaranteeded to get a QMD-wu, and get atleast 2x more points than non-Intel would normally get.

(Hmm, sees Insidious has already answered, but anyway, to get his 4x-number to add-up, it means the bonuses before was 100%...)


Has Folding@Home learned anything after all the QMD-controversials?

Well, with current SMP-bonuses, I would say no, FAH's point-system is if possible even more screwed-up than it's ever been before...

There's been hinted-at some changes to the points-system then v6 is out of beta, but there's still not been any real info on impact.



(*) Well, the benchmark-rules has changed resently, to accomodate PS3, GPU and SMP, but for the "normal" cores it's still the old non-SSE2/3/4-benchmark-rule.

 

[Insidious]

yep, the SMP scoring is ridiculous too !!!!

My last trip(s) to the F@H forum were to participate in a "discussion" about possible remedies....

Lot's of unproductive banter with the usual pretext... "we're gonna fix it", "just you wait", "here's a link to a hint", etc, etc, etc,

Also a BIG flap about people deleting WUs which were scored unfairly low. Commonly refered to as cherry picking.... But many months later.... still BIG scoring inequities, still people feeling the need to delete WUs because their points are too low, still an angry F@H forum moderated from a bully pulpit ........ uh, you've met one of them.

And still we fold.....
Someone buy us a :beer:



-Sid

 

[mgpaulus]

Originally posted by: Insidious


And still we fold.....
Someone buy us a :beer:



-Sid

Here's hoisting one to you, Sid.......
:beer: :beer: :beer:

Would you like yours dark, light, IPA, Amber, Brown Ale, stout, porter, etc??

Me, I would like either a New Belgium Fat Tire, or a Guinness......

 

[Insidious]

:beer: :beer: :beer: ( <- the big G )

:thumbsup:

 

[Rattledagger]

Originally posted by: Insidious
Also a BIG flap about people deleting WUs which were scored unfairly low. Commonly refered to as cherry picking.... But many months later.... still BIG scoring inequities, still people feeling the need to delete WUs because their points are too low, still an angry F@H forum moderated from a bully pulpit ........ uh, you've met one of them.

Yes, I've met a couple of them...

And still we fold.....
Someone buy us a :beer:



-Sid

For various reasons my FAH-production has always been low and erratic, but should still "soon" hit 200 returned wu's. :beer:

 

[GLeeM]

The QMDs came out right after all that P4 HT squabble.
It did no good to run two QMDs on HT, so it only made sense to run one.
Seemed to me like they made them to end the HT disagreements.

Now the forum concensus is that current WUs cannot do better by running two on HT. I have not found this to be true, but the advantage is a tiny bit less. But I wasn't about to say anything there.

Originally posted by: Rattledagger
3; FAH gives bonuses for "advmethods", wu's that either is just out of beta or for beta-cores,

I don't think there is a bonus for "-advmethods" WUs.

Occasionally you can get bonus WUs more often by using -advmethods. Currently I think you are better off NOT using it.

 

[biodoc]

I would like to add a couple of points and also clarify the $/support situation that at least on the surface, appears to be the case at F@H. I have been in biological research in both industry and academics for the last 28 years so perhaps I can offer a relevant perspective? In the end, it's your call.

The principal investigator (PI) of the Stanford lab has a very difficult job (as do all PI's in academics). Their principal jobs are to bring funding into the lab, provide scientific guidance to the grad students and postdocs, and to publish frequently in peer reviewed scientific journals.

Unfortunately, most PI's spend the vast majority of their time in search for funding and also are on the road at scientific conferences presenting their work (Public relations, etc).

In the case of the F@H PI, it's fairly clear to me that if a corporation provides funding to his lab, that corporation is rewarded with F@H clients that support the corporation's hardware (Intel, Apple, ATI, Sony). I personally don't have a big problem with that, but please don't tell us that other hardware will also be supported with the same effort (AMD, NV, Xbox, etc). Let us not forget the age old promise of BOINC support too!

If these agreements are exclusive for a period of time, then that doesn't bode well for an NV GPU client.

Interestingly, now that AMD owns ATI, how's Stanford going to deal with that inconvenience? Will they eventually support the new Firestream processor base on ATI technology?

Now comes what I feel is my most important point. Peel back that outer political/sales layer of PI & "loyalist" mantra and you will find highly dedicated and hard working grad students/postdocs who are really driving the science of this project. I really feel that F@H and Rosetta@Home are both valuable projects and are worth crunching for future cures to disease.

Kudos to GLeeM who always take the high road!

Hey Sid, nice job handing the Stanford lab "press Secretary" today! I enjoyed it! :beer:

 

[7im]

It's hilarious how far from the truth things can get 3 years after the fact. It's like that kid's game of Grapevine, where the original message at the first child is nothing like what comes out at the last.

1. QMDs ended when the Principle Investigator graduated, and for NO other reason.
2. QMDs were only assigned to Intel systems because Intel chips performed more than 2 times faster than AMD chips. AND because of licensing issues with the science libraries used in that core. Intel's library didn't enable SSE2 on AMD chips and ran really fast, and the AMD libraries didn't enable Intel chips, and ran slowly. And since more than 75% of the users at the time were Intel users, there was little advantage to catering to the minority. With no performance improvements, there was little need to develop an AMD only QMD capable client. AMD chips were just as productive on regular work units, so that is what they we assigned. As I said, the science comes first, even if the public opinion of a very small minority doesn't agree.
3. The standard benchmark procedures were followed to set the points. Say what you will about it, but if they had diverged from policy, you would be complaining about that as well. I never saw you complaining about AMD chips scoring 20% better than Intel chips on Tinker work units. :roll: Sounds like a very onesided fanboy complaint to me.
4. I ran the numbers back in the day, and less than 10% of the project participants ever opted in to fold a QMD. No big deal as far as the overall project was concerned.

And bringing up insignificant problems from 3 years ago in your attempts to support your complaints today helps your arguments very little. Past history, long forgotten by most. It's like pointing at Roswell as proof of aliens, when there is so much more current evidence available in the present, and yet people chose to ignore it.

And just because the project accepts help from some vendors, is not a valid reason to assume competing vendors are shorted. You don't see Stanford shorting the Microsoft clients because Apple donated hardware last year. Give me a break, that argument is ridiculous. And PS3 vs. Xbox, are you kidding me? The Xbox hardware would fold slower than a G4 Macintosh. And the Xbox overheats just running a game, let along a much more demanding application like F@h. No valid argument there either, and no reason to waste limited resources to develop a client, especially when MS wouldn't lift a finger to help. And if you choose to selectively ignore the hints about NV, that IS your choice, but I'll gladly come back and serve crow again. And that will shoot down the last of your supposed discrimination claims. I can hardly wait!


P.S. And not that you'd seem to care, but the Folding Forum is back up. And how much more of a NV hint do you need? http://forum.folding-community...post203996.html#203996

 

[Insidious]

Thanks for visiting!

Don't let the door hit you in the butt on your way out!

 

[7im]

I've been here a long time, much before my registration date. And I'm not going anywhere now, when things are just about to get really fun.

 

[GLeeM]

Originally posted by: biodoc
I would like to add a couple of points and also clarify the $/support situation that at least on the surface, appears to be the case at F@H. I have been in biological research in both industry and academics for the last 28 years so perhaps I can offer a relevant perspective? In the end, it's your call.

The principal investigator (PI) of the Stanford lab has a very difficult job (as do all PI's in academics). Their principal jobs are to bring funding into the lab, provide scientific guidance to the grad students and postdocs, and to publish frequently in peer reviewed scientific journals.

Unfortunately, most PI's spend the vast majority of their time in search for funding and also are on the road at scientific conferences presenting their work (Public relations, etc).

In the case of the F@H PI, it's fairly clear to me that if a corporation provides funding to his lab, that corporation is rewarded with F@H clients that support the corporation's hardware (Intel, Apple, ATI, Sony). I personally don't have a big problem with that, but please don't tell us that other hardware will also be supported with the same effort (AMD, NV, Xbox, etc). Let us not forget the age old promise of BOINC support too!

If these agreements are exclusive for a period of time, then that doesn't bode well for an NV GPU client.

Interestingly, now that AMD owns ATI, how's Stanford going to deal with that inconvenience? Will they eventually support the new Firestream processor base on ATI technology?

Now comes what I feel is my most important point. Peel back that outer political/sales layer of PI & "loyalist" mantra and you will find highly dedicated and hard working grad students/postdocs who are really driving the science of this project. I really feel that F@H and Rosetta@Home are both valuable projects and are worth crunching for future cures to disease.

Kudos to GLeeM who always take the high road!

Hey Sid, nice job handing the Stanford lab "press Secretary" today! I enjoyed it! :beer:

Thanks for the peek inside how things work :thumbsup:

Even though it is not completely true, thanks for the kudos, maybe it will stop me next time I want to post over at FCF just to get "someones" goat (I can't help it, it is so easy ) Actually the few times I've done that is usually to take the whipping instead of whoever is getting or about to get it :roll:
I just have to say this because of the pun: I don't like to see whipping posts over there either.

 

[Assimilator1]

Thanks for the info Sid,RD & biodoc

Originally posted by: mgpaulus

Originally posted by: Insidious


And still we fold.....
Someone buy us a :beer:

-Sid

Here's hoisting one to you, Sid.......
:beer: :beer: :beer:
Would you like yours dark, light, IPA, Amber, Brown Ale, stout, porter, etc??
Me, I would like either a New Belgium Fat Tire, or a Guinness......

Bleh! ,give me a Kronenberg or a Carling anyday

7im
Interesting points ,1 though is unclear (worded unclearly)

Intel's library didn't enable SSE2 on AMD chips and ran really fast, ... <--- on Intel chips?
... and the AMD libraries didn't enable Intel chips, and ran slowly <--- on AMD chips?

Also you said And if you choose to selectively ignore the hints about NV, that IS your choice
Assuming that you are talking about the link you posted earlier, we are not ignoring any NV hints, their simply aren't any! ,prove us wrong by copying & pasting the lines you have in mind

On point 2 (bear in mind I'm a F@H newbie so I'm only going by what's in this thread) ,you said that the SSE2 made the QMDs x2 faster on Intel chips ,& yet (apparently) they got x4 the score! ,this is illogical as far as I can see & also I can quite understand why it would upset people ,a boost in the score equal to the boost in performance would of been the sensible & fair thing to do.
I guess that stems from testing an SSE2 client with SSE2 turned off? ,I would of thought that if you optimise any program to use SSE2 etc that you'd want to benchmark it with SSE2 on to be able to see the difference ,& not hide it.Or have I missed something?

Btw when you mentioned the Xbox ,did you mean the original one? ,I believe Insidious was talking about the Xbox 360 (whether he actually said 360 or not) ,now I know the 360 does have overheating problems but its a lot more powerful than the original Xbox of course

Insidous
Worst of all (IMO) was a persistant LIE that the Stanford team was trying to make them run on AMD platforms... history proved what fools we were to believe that!

How do you know they didn't try?

 

[Insidious]

How do you know they didn't try?

Are you asking how do I know that is my opinion? (IMO=In My Opinion)

-Sid

edit: With regard to how I formed my opinion.... You had to be there.
It was well documented as I posted a lot during that time. Not a flame war I really want to dredge up here. It occured at the F@H forums and was really only appropriate in the environment their moderators maintained... not at AnandTech.

 

[7im]

Sid,

If Stanford did not try to make QMDs work on AMD chips, then how did Stanford know the AMD science libraries made the QMDs run so much slower on both the AMD and Intel chips? How did they know the AMD libraries disabled SSE2 on the Intel chips just like the Intel libraries disabled SSE2 on the AMD chips? Both Intel and AMD play dirty.

The decision was simple, run the faster libraries, which also happened to match the largest demographic of fah contributor. It was a no-brainer.

QMDs benefitted from a double bonus, 2x for Big WUs, and 2x for SSE2. 4x was the standard benchmarking procedure at the time. IIRC, SSE2 WUs still get 2x bonus, Big WUs slightly less now.


Assimilator1,

I posted a link where Vijay specifically mentions the G80, they are working on it, and he hinted at a time frame. http://forum.folding-community...post203996.html#203996

As for the hints in the News post, answer me this question... "What does Stanford state is the single biggest problem getting the NV cards to fold?" No Working Drivers!

And then what does the News post specifically say about that topic? http://folding.typepad.com/new...1/sc2007-demo-new.html

"...and perhaps most importantly, much better driver support (we've fixed most if not all of the bugs leading to driver incompatibilities)."

I have drawn my conclusions, feel free to do the same. Take it for whatever you think it's worth...