GPU client only running test proteins ?

AMD_Gamer

Fully [H]
Joined
Jan 20, 2002
Messages
18,287
I thought all the WU's we have been doing on the GPU client are folding proteins for research but the the thread about the new WU's people are saying that we have been running test units?

what are these test units?:eek:

what is going on here?:confused:
 
They want to make sure that the stuff we are sending back match the results that have been sent back from other cores (IE SMP)

The new stuff we are getting are real from my understanding. This is why they are running harder and hotter. They where made for the GPU environment. The other WU we have been working "fit" into the GPU, but not perfectly.

They really don't want a repeat of GPU1 :eek:

 
They want to make sure that the stuff we are sending back match the results that have been sent back from other cores (IE SMP)

The new stuff we are getting are real from my understanding. This is why they are running harder and hotter. They where made for the GPU environment. The other WU we have been working "fit" into the GPU, but not perfectly.

They really don't want a repeat of GPU1 :eek:


if the new WU's are working the GPU harder and running hotter how do they know that won't cause errors?
 
if the new WU's are working the GPU harder and running hotter how do they know that won't cause errors?

My guess is that they core will EUE before they get back bad data...... that might have been one of the things we have been testing..... dunno.

I'm just glad we are getting real stuff on all this equipment.

 
My guess is that they core will EUE before they get back bad data...... that might have been one of the things we have been testing..... dunno.

I'm just glad we are getting real stuff on all this equipment.


so the 5700 WU's being talked about in those other threads are real science we are working on?
 
if the new WU's are working the GPU harder and running hotter how do they know that won't cause errors?

They have ways of looking at the data they're getting back and telling if something isn't right if the units don't all EUE to begin with.

And as far as "test" units vs. "real" units, work units such as the 50xx ones would say "designed to test the scientific value of the nVidia GPU2 core" on the project description page. The value of every unit ran should be a bit more obvious when remembering the whole GPU1 deal and how that went down.


2.png
 
50xx series are test proteins but they are now very rare. The current crop is indeed real work units but not all is about protein folding per se. A part of them is also for testing new computation algorithms to make protein folding more efficient and they are submitting papers on it as well so they are useful. The 57xx series are indeed the real deal on protein folding.

We are crunching real units since july IIRC.

 
We are crunching real units since july IIRC.


Hey thanks brother Xilikon for clearing that up :D I almost "pooped" in my pants (not really, just a figure of speech :rolleyes:) when I read this post and other posts concerning whether we were just "spinning our wheels" or actually doing something useful. I don't know about anyone else, but it would make me very disappointed in the Stanford F@H project to think the wear and tear on my boxen, the higher light bills, the more attention paid to my boxen maintanance, the purchases of new and better componets to fold with, was all for freaking nothing. :( :mad:

I was under the mistaken impression that the testing of new WU's, clients, etc was done by testers before they were released into the "wild" (mainstream) :confused: It seems to me, with all the beta testers Stanford has (and to listen to some of them you'd think there wasn't anything about computers or folding they didn't know :rolleyes:) , they shouldn't be having all these surprise problems (more heat on GPU cards, CPU usage in GPU2 folding, EUE'ing, crumby WinSMP clients, etc) without at least warning the masses. (like downclocking CPU's, GPU's, etc) ;)

Folding and WCGing for the CURE

 
I was under the mistaken impression that the testing of new WU's, clients, etc was done by testers before they were released into the "wild" (mainstream) :confused: It seems to me, with all the beta testers Stanford has (and to listen to some of them you'd think there wasn't anything about computers or folding they didn't know :rolleyes:) , they shouldn't be having all these surprise problems (more heat on GPU cards, CPU usage in GPU2 folding, EUE'ing, crumby WinSMP clients, etc) without at least warning the masses. (like downclocking CPU's, GPU's, etc) ;)

Folding and WCGing for the CURE


Yes, you are correct. All Wu is getting thru a QA process that I'm also part (with some others like EvilAlchemist, Sunin, etc...) but remember that the sampling is not big enough to catch real issues that the general users is getting. Also, in the past, the Pande Group often release new units without considering beta testers opinions and feeling with disastrous results. Since my email to Vijay outlining this issue, they improved this aspect and now ask us what we think about the next step to do (like releasing public, release as manual download, etc...). This do show a lot by reducing the occurences of bad projects causing hair-pulling issues among the general crowd.

About the 57xx crop, we did the test and we didn't see any significant problem. I noticed that most people who have issues is often running it without proper cooling or with a bad combination of driver and core. Overall, it's generally good and we must accept them since those are the first bunch with a real scientific relation to folding (this should be what matter for those who hate spinning the wheel for nothing).

 
Brother Xilikon, please don't get the wrong impression, I wasn't in any way demeaning the great beta testers that are [H]orde. members, I was inadvertantly refering to some of the posters on the F@H forum. If it wasn't for you people ([H]orde beta testers) a lot of us folders would be completely in the dark. :eek:

I don't want you guys (beta testers) to get the wrong impression I'll probably catch hell for this statement, but if the "higher upper muckity mucks" from Stanford would listen more closely to you beta testers they'd have less problems, more knowledge of computers and their base of folders. Please don't get me wrong, I don't mean you should try to tell them about the science of protein folding (without a doubt Stanford is the best at that), just about the science of computers and the science of computer fanatic "human behavior". ;)

Oh yeah, it'd sure be nice if they'd let you guys look at the beta software first :D

Folding and WCGing for the CURE

 
50xx series are test proteins but they are now very rare. The current crop is indeed real work units but not all is about protein folding per se. A part of them is also for testing new computation algorithms to make protein folding more efficient and they are submitting papers on it as well so they are useful. The 57xx series are indeed the real deal on protein folding.

We are crunching real units since july IIRC.


I'm still getting a lot of 50XX series, i would say its about 50% 50XX and the new ones, seems to switch every other WU:confused:
 
I'm still getting a lot of 50XX series, i would say its about 50% 50XX and the new ones, seems to switch every other WU:confused:

Luck of the draw... I crunch 57xx 70% of the time and 55xx 25% of the time with the remaining doing 50xx. If they are totally useless, they would have pulled them out of circulation by now but it seems they still need the infos. They know better than to waste time of everyone with useless units.

 
This do show a lot by reducing the occurences of bad projects causing hair-pulling issues among the general crowd.

If they are totally useless, they would have pulled them out of circulation by now but it seems they still need the infos. They know better than to waste time of everyone with useless units.

Thanks. I trust you SillyCon and your opinion on the matter. I don't trust the Pandas. I appreciate all that you have done to get their attention and for them to take into consideration, if just for a minute, the thousands of folders who have invested thousands for the cause.





 
Thanks Poc :p

I also forgot that the PG is struggling a bit to have a core working stable with the large WU, hence why they are delayed until now. We are now just starting to crunch them without too much issues and future cores will improve stability a lot.

 
I also forgot that the PG is struggling a bit to have a core working stable with the large WU, hence why they are delayed until now. We are now just starting to crunch them without too much issues and future cores will improve stability a lot.
What about the Ati client? Evil mentioned some interesting new developments in the near future.

 
What about the Ati client? Evil mentioned some interesting new developments in the near future.


Dunno about ATI, I'll let EA answer this. I'm the nvidia specialist and he's the ati specialist and we don't step on each other toes :D

 
The new stuff we are getting are real from my understanding. This is why they are running harder and hotter

I've noticed that some WU's make my GPU run hotter and cause some graphics lag within normal Windows operation. Seems these WU's are coming rather infrequently, though, or maybe only while I'm sleeping. I couldn't comment on which are real or test units though.
 
Dunno about ATI, I'll let EA answer this. I'm the nvidia specialist and he's the ati specialist and we don't step on each other toes :D

That darn Green Team ... Red Team for Life .....
Mr Xilikon is very right, generally we don't step on each other or post in their area.

It is nice thought to have so many [H] members working on beta.

What about the Ati client? Evil mentioned some interesting new developments in the near future.

Things are going well, but slowley.
Having 9x 4850's (Well, will be 10 x 4850 by Friday) helps a lot in testing.
That also make me one of the very few people to have quad ATI gpu systems.

The ATI beta Group has some great members and they are really working hard to test
We can usually get a test ran and results posted within 24 hours about the status.
each core and new work units to ensure they are stable and getting the correct points
At the moment, there are three main issue that need to be fixed.


I know that Mike Houston with ATI has been working overtime latley and putting in many hours to address these issues. I have even seen him online at 1 am still making posts at the FF so he is giving all he has.

Mike also has a baby on the way later this month and I think we all understand the imporatnce of family. That will slow things down a little, but IMO that is 100% acceptable and I wish him the best with his new family addition.
 
Back
Top