Folding Forum

mdk777 wrote:

I strongly recommend anyone who wants to complain about ATI performance vs. NVidia performance and say hardware doesn't cause anything to read the post below:
viewtopic.php?p=117995#p117995

You can't call conspiracy if you haven't first read that post.

No conspiracy, just a lack of effort in support.
The conclusion to the thread you reference is:

As has been explained ad infinitum in this thread and others, our focus is now on the next generation GPU3 client and OpenMM core. The programming language and model used by the old ATI code have been made largely obsolete by the introduction of OpenCL, so we're focusing our efforts there, rather than trying to revamp an older code.

Here is clearly admits that the client was for all intents and purpose depreciated a long time ago. It just was never made official.

Yes, this is all several year old news, but continuing to claim it didn't happen falls into the category of history revision.

Lack of effort? How is Stanford, according to your own words, supposed to make an obsolete code base work better for ATI clients? How is Stanford slacking off when Stanford has to wait for ATI to finish the OpenCL code?

Lack of progress, or halting development, or changing developmental focus is not the same as deprecating the client, and therefore not revisionist. Revisionism takes place when one tries to change the understood facts about a situation, and no one is trying to change the facts here but you. Again, as you said, that's an old thread, with long understood information, but here you are trying to draw new conclusions. No one here has ever claimed that post didn't happen. No one deleted the post. No one edited the post. And for an old post, it still gets referenced quite often to quiet the fanboys.

Let's look at an actual example. The v6.29 SMP was deprecated when v6.30 SMP was released. ATI GPU is much different. The ATI client is still active with no replacement yet, so how can it be considered deprecated? So do you want everyone with an ATI card to just turn it off because the client code is considered obsolete? Ya, sure, let's just turn off 600 TFlops of production because of the lack of progress in developing a replacement. Ya, sure, the ATI client just sucks too badly. Let's all shut them off in protest.

So do you want everyone with an ATI card to just turn it off because the client code is considered obsolete? Ya, sure, let's just turn off 600 TFlops of production because of the lack of progress in developing a replacement. Ya, sure, the ATI client just sucks too badly. Let's all shut them off in protest.

Yes, I believe in accountability
If your product is defective, you do not continue to sell it just because you don't yet have a non-defective product yet.

If you care about your customers, and your reputation: It is the only logical choice.

You obviously do not think that a sub 25% efficiency qualifies as defective. Yes it produces some result. But truth in advertising would have you warn donors that the efficiency is very poor and they could make an informed decision.

Hence when the OP complains, it does not make sense to blame the differences in Hardware (which is the post I responded)

I guess the correct response to the OP is :

You are right, the situation is not optimal, but we are working as fast as possible to step up to the next level.
http://folding.typepad.com/news/2010/10 ... rsary.html


All the rest is just finger pointing and CYA.

The ATI folding cores aren't defective, they're just not as effective at utilising the theoretical peak power as the NVidia cores are, for the difficulties outlined in ihaque's post.

The simple matter is with ATI vs. NVidia, you have to take into account the hardware differences as outlined.

Stanford pulling the ATI GPU client would be like GM recalling all their petrol cars and trucks because Hybrids have a better fuel consumption.

I think the proper comparison would be to a Trabant, rather than any current GM product.
But give Government Motors some time and they will get back there.

http://online.wsj.com/article/SB124381203054570397.html

Returning to the bottom of the topic, I forgot about the most important thing is that the current publication of research results. For several years I participated in the project grid.org, where there was a rule. (Project was closed April 27, 2007 year, that's why I participate in the project folding @ home). I would like the same case here.

filu wrote:I forgot about the most important thing is that the current publication of research results.

You mean like those listed here: http://folding.stanford.edu/English/Papers ?

faithruler wrote:Nvidia has hinted that it has a venture within the works that will enable Nvidia's CUDA technologies on AMD GPUs.

Sure, they'd be happy to make more money off of CUDA, especially if that money comes from a competitor. At this stage of development, NVidia has won the software battle. Brook+ and CAL are history, and everybody knows that, but nothing you said implies that NVidia will give that technology to AMD for free. AMD has a choice whether to pay NV a huge sum of money to purchase the rights to CUDA or not. Whether they decide to do that or if they have other plans to stay in business, is known only to the strategists inside AMD.

Does AMD have something in their back room that is better than CUDA? Will OpenCL be better than CUDA? Can they afford whatever NVidia will charge? I certainly don't know, but technology does move on, both hardware and software, and FAH can't run their CUDA-oriented core on ATI, even if it would run much better than their Brook+ oriented core.

The same sort of thing happened with 3DNow developed by AMD, upgraded to 3DNow+ and then updated again to SSE by Intel. (The legal strategists at Intel did figure out how to market SSE as a totally new product without infringing on AMDs patents.) Eventually AMD had to license SSE from Intel to stay in business.

There are other minor questions, though. What about Intel and other minor players who do happen to make GPUs? They'll certainly support OpenCL for graphics but maybe not CUDA or Brook+ or Stream. Will OpenCL provide enough compute capability to let FAH run on their GPUs?

Wow, that was a big read.

mdk777 wrote:

Yes, I believe in accountability
If your product is defective, you do not continue to sell it just because you don't yet have a non-defective product yet.

You might want to tell this to Nvidia, some would say they dont follow ur idea of accountability:

http://www.hplies.com/ or even better: http://support.apple.com/kb/ts2377

Better hardware? Muahahaha!

filu wrote:...

In my opinion...
Client V7 = Duke Nukem Forever

Well, Duke is now a reality. http://arstechnica.com/gaming/news/2010 ... ootage.ars [NSFW]

And now it's a race to see which one comes out first. Hopefully it won't take another 12 years.

7im wrote:And now it's a race to see which one comes out first. Hopefully it won't take another 12 years.

Duke Nukem?

It may be hard for you to understand me. But people of my Team are no longer folded. My intention is not unreasonable to attack the project, only to draw attention to its shortcomings. Currently, we discuss on the forum of Team Poland our longer term future in the project. Summary of discussion will be sent to Prof. Pande and published on foldingforum.org.

maybe we should ask Gearbox to help develop the v7 client then?

if people from your team stopped folding because of the reasons you outlined in your first post, they would probably stop folding sooner or later no matter what anybody did.

1) client beta-ness ( ) has nothing to do with project results as clients only manage the MD cores, those indeed do not leave stanford untill they are out of beta.
2) those effectiveness figures you mentioned are speculative at best. even if that was true, there is the possibility (or rather probability?) that the current MD core or WUs don't fully utilize these cards.
3) WUs for uniprocessor clients do not have ppd cut. people who want to use spare cycles fire up a uniproc. client and forget about it, and for that they earn "some points". people who want to get involved more fire up some of the high performance clients and earn "more than some points". where is the problem with that? anybody can do that.
4) just another fluke of the points system, and not really the worst imo, as this affects everybody the same.

im not saying PG does things perfectly, as sometimes their attitude (the lack of it mostly) winds me up, but most of what you wrote doesnt really have to do with what PG does. its more how some people misinterpret things about the project.

^w^ing wrote:if people from your team stopped folding because of the reasons you outlined in your first post, they would probably stop folding sooner or later no matter what anybody did.

I don't thing so. The three (Borgis, chillerworks.com, P.Holcman) wrote that they would return when bigWU's will be available at all times. They are at the moment went on to other DC's projects.
I'm sorry, because together with other members of the Team we have put a lot of effort, time and money in attracting new members (setonek).

its more how some people misinterpret things about the project.

Lowering expectations;
the road to excellence...er...no...um.....
how to rationalize mediocrity....er...wait......that's not it.....
to avoid disappointment. Yeah, that's it.

they would probably stop folding sooner or later no matter what anybody did.

Yeah, that is the problem. The lack of folder retention.
You make a symptom sound like a benefit.

In Sales?