Selective Folding (branched from Nvidia issues forum)

[shdbcamping]

I think you're forgetting that Fit, Form and Function also applies to Software design. CUDA (and CAL) are "Designed for" GPUs, and therefore have a responsibility not to damage GPUs.

Fit, Form and Function where cuda are concerned only counts for GPU HW "Designed for" CUDA. AGAIN... Graphics cards are only CUDA capable and the CUDA software needs implemented accordingly, as the 24/7 Cuda was not in the Cooling solution applied for the application of the Graphics card.

[shdbcamping]

Hi again all,
Let's get back to ideas to implement User interface choices with the WU or CFG to choose WU classes and let the user throttle them to the DONORS comfort level regarding GPU core heat. Thast way I can make up the difference by letting the non Heat affected WU's go racecourse mode. I have begun stopping some instances of the HOT core11 WU's and deleting until I get a non burner WU. If Pande will not give me a way to ensure my system is used in a manner that I am comfortable with, I will send them back and draw again until I get another one that accomodates. For me, optimum is 75C. I can do 75c massively shader OC'd and mem and core stock on all my GX2's. It is a software design issue with these Overheated WU clases. All who have no problems could let the defaults run and Max them if it worked. I'm not saying slow every one down... Just looking for an opportunity to be able to run them where I can be comfortable. I hate having having to shut down Half of some GX2's because I want to run for the long haul and don't have the money to replace 8 of them if they get burnt up prematurely.

Can someone tell me what happens with Pande servers when they get a bunch of aborted flamethrowers back after they have been prevoiusly assigned? If this is what I need to do to contribute to science and protect my system in a way I wish to... I'm sorry, I just don't have another $10,000.

Looks like cherry picking to me.
Is Stanford going to let that stand?
What action has been taken in the past? And is this going to be done now?
Cherry picking is not right no matter the reason.

With all due respect, to me it looks like protection of equipment. NO more no less. Expensive equipment that is contributing as much science as I can possibly get out of it. I can and do run the flamethrower units on my 2xGPU rig but they are a Danger to my multi-GPU rigs. Understand that once a card is burnt up it is gone for good and will never contribute any more science. I h ave to protect them any way I can. That's the only reason I even consider this strategy.

[v00d00]

Pandegroup appears to address the issue of dumping by adding tracking based on ip and re-issuing the same workunit. On mine when ive had failed workunits its been a pain to get new workunits as it seems that after 3 tries your ip gets blacklisted for 5 mins or so. If a person was seen to dump large numbers over a large period of time im guessing they would either get the shutdown core that kills all folding or just blacklisted if they had a static ip.

The older ones amongst us have been down this road long ago and the result was an improvement in the handling of work unit assignment and the elimination of some of the old ways of cherry picking. As more and more people do it they will continue to improve there handling methods until it is eliminated in one way or another.

Another small fact is if you continually request workunits and dump them you are causing a Denial of Service attack against Stanford University servers by wasting large amounts of bandwidth by requesting data. Also since you are doing this while actually knowing the consequences and what is liable to happen the attack would be classed as a malicious act. Just bear that in mind next time you dump a load of workunits.

[shatteredsilicon]

Pandegroup appears to address the issue of dumping by adding tracking based on ip and re-issuing the same workunit. On mine when ive had failed workunits its been a pain to get new workunits as it seems that after 3 tries your ip gets blacklisted for 5 mins or so.

Sounds like you didn't clean up properly afterwards by deleting *.dat files.

[Bill1024]

I think you're forgetting that Fit, Form and Function also applies to Software design. CUDA (and CAL) are "Designed for" GPUs, and therefore have a responsibility not to damage GPUs.

Fit, Form and Function where cuda are concerned only counts for GPU HW "Designed for" CUDA. AGAIN... Graphics cards are only CUDA capable and the CUDA software needs implemented accordingly, as the 24/7 Cuda was not in the Cooling solution applied for the application of the Graphics card.

The topic here is selective folding, cherry picking, or tossing back work units we don't feel like folding.
Can we please stay on topic?
Cherry picking is bad for science. No one is forcing any one to fold.
There are rules in place for a reason. Cherry picking sets back the projects, plus it is not fair to the folders that do play by the book. We all have choices in life, sometimes we make the wrong ones, even when we think we are justified.
So please don't dump work units, it really is not helping, it is harmful.

[v00d00]

Pandegroup appears to address the issue of dumping by adding tracking based on ip and re-issuing the same workunit. On mine when ive had failed workunits its been a pain to get new workunits as it seems that after 3 tries your ip gets blacklisted for 5 mins or so.

Sounds like you didn't clean up properly afterwards by deleting *.dat files.

Thats the thing i did. That was the whole issue.

After deleting the queue.dat and work/, i got the same workunit again (exact PRCG), which then failed again. So i tried for another and same result, then i was blocked by the AS server for about 5 mins. So i just shutdown folding on that client for an hour and tried again with a different set of flags and bagged something else later on by changing the name on the account and passkey, then swapping it back after.

[bruce]

Looks like cherry picking to me.
Is Stanford going to let that stand?
What action has been taken in the past? And is this going to be done now?
Cherry picking is not right no matter the reason.

With all due respect, to me it looks like protection of equipment. NO more no less. Expensive equipment that is contributing as much science as I can possibly get out of it. I can and do run the flamethrower units on my 2xGPU rig but they are a Danger to my multi-GPU rigs. Understand that once a card is burnt up it is gone for good and will never contribute any more science. I h ave to protect them any way I can. That's the only reason I even consider this strategy.

There are many options to protect expensive equipment that have been mentioned. You can reduce the number of GPUs in a single case. You can improve the airflow. You can reduce the %CPU setting. You can cut back on your folding when the ambient temperature gets too hot. You can modify the hardware (e.g.- water cooling). You can (if it does get to that point) return damaged cards for warranty replacement, though I strongly support your desire to avoid that option.

The Pande Group may very well come up with an option that is not available now, but until then, you have to figure out what works for you. In any case, WU dumping is really, really bad for FAH's science and I strongly encourage you to use some other option.

[Fernando_Celio]

OK, Bruce.

I know you are right, but everyone knows what is going wrong with this projects. Let's be honest, they demands too much time to finish, too much CPU overhead and they are totally crystal clear undervalued.
I have this problem with my team and, believe me, it is not so easy to convince them to do not do that (exclude these WU's).

Someone could say " this guy is worry just about points" but I am sorry to disappoint who think in this way. I don't choose anything. If Stanford wants that I do this, I just do it. My major concern is that my NVIDIA card is in one machine not too fast and I realize that if I receive too many "511 points WU's", I just can't send the results of an SMP client in time. It's a dual Pentium 3.0 GHz. So, in my case is a loose to loose situation. To Stanford and Science, naturally.

So, please, someone from Stanford can answer why there is so many people hating this WU's ?
Is it time to review your values in function of this ?
Is there any way to improve this WU's ?

Sorry for this post.

Best Regards, Fernando.

[bruce]

So, please, someone from Stanford can answer why there is so many people hating this WU's ?
Is it time to review your values in function of this ?
Is there any way to improve this WU's ?

Sorry for this post.

Best Regards, Fernando.

I'm not from Stanford, but I believe I can answer that question. There seem to be several different reasons. For some, it's an issue of heat. For others, apparently it's an issue of PPD. There may be other reasons, as well, but the people with ATI hardware don't seem to hate them.

In any case, the Pande Group does not condone "cherrypicking".

[shdbcamping]

OK, Bruce.

I know you are right, but everyone knows what is going wrong with this projects. Let's be honest, they demands too much time to finish, too much CPU overhead and they are totally crystal clear undervalued.
I have this problem with my team and, believe me, it is not so easy to convince them to do not do that (exclude these WU's).

Someone could say " this guy is worry just about points" but I am sorry to disappoint who think in this way. I don't choose anything. If Stanford wants that I do this, I just do it. My major concern is that my NVIDIA card is in one machine not to fast and I realize that if I receive too many "511 points WU's" I just can send the results of an SMP client in time. It's a dual Pentium 3.0 GHz. So, in my case is a loose to loose situation. To Stanford and Science, naturally.

So, please, someone from Stanford can answer why there is so many people hating this WU's ?
Is it time to review your values in function of this ?
Is there any way to improve this WU's ?

Sorry for this post.

Best Regards, Fernando.

Thank you for the vote of support. I do NOT throw any WU's Back in the "fishing" or "cherry picking" accusation thrown my way prior in this thread. If I come home and see 2 (TWO) 511, 430 or 768 point wu's pushing a GX@ to 105C. I simply close the GPU that has the least progress. If I get back to the computer and That GPU client has more than the other "compliment" GPU client for that card... I'll switch the clients. Unfortunately, too often when I get back the Running client has a more productive WU running. Simply put, 4 Hours or however many days before "Late" is a function of the heat.

Hardly "cherry picking", I let the Clients run... All 80K+ possibly, usually -70k ish. Because of the other problem. Shutting Clients down because my Dual GPU GX2's are at and over 105C is nothing that I will be made to feel "Ashamed" of. I'm kicking as much Science as I feel I can Afford.

Please stop the point stuff. Personally, cut my heat and points.

[Bill1024]

OK, Bruce.

Then shdcamping wrote.
Thank you for the vote of support. I do NOT throw any WU's Back in the "fishing" or "cherry picking" accusation thrown my way prior in this thread. If I come home and see 2 (TWO) 511, 430 or 768 point wu's pushing a GX@ to 105C. I simply close the GPU that has the least progress. If I get back to the computer and That GPU client has more than the other "compliment" GPU client for that card... I'll switch the clients. Unfortunately, too often when I get back the Running client has a more productive WU running. Simply put, 4 Hours or however many days before "Late" is a function of the heat.

Hardly "cherry picking", I let the Clients run... All 80K+ possibly, usually -70k ish. Because of the other problem. Shutting Clients down because my Dual GPU GX2's are at and over 105C is nothing that I will be made to feel "Ashamed" of. I'm kicking as much Science as I feel I can Afford.

First of all it was not me that moved the posts to here from the "heat" threads.
So some one else did see your confession too.
I posted your own exact words.
Not an accusation by me. It was a confession by you.
You wrote

I have begun stopping some instances of the HOT core11 WU's and deleting until I get a non burner WU. If Pande will not give me a way to ensure my system is used in a manner that I am comfortable with, I will send them back and draw again until I get another one that accomodates. For me, optimum is 75C. I can do 75c massively shader OC'd and mem and core stock

You clearly say, you have begun stopping and deleting until you get a non-burner.
Now you say "you do not throw them back, or cherry pick"..
What is the truth? Either you do, or you don't, cant have it both ways.
If you did not do it, why did you say you did?

[shatteredsilicon]

For me, optimum is 75C. I can do 75c massively shader OC'd and mem and core stock

Wow! All I can say is wow. I cannot even begin to come up with words to respond with to that. So you are expressing vehement concern about temperatures and yet you are admitting to "massively" (your words, not mine) OC-ing. I am awed.

[Amaruk]

I have begun stopping some instances of the HOT core11 WU's and deleting until I get a non burner WU. If Pande will not give me a way to ensure my system is used in a manner that I am comfortable with, I will send them back and draw again until I get another one that accomodates.

I do NOT throw any WU's Back in the "fishing" or "cherry picking" accusation thrown my way prior in this thread. If I come home and see 2 (TWO) 511, 430 or 768 point wu's pushing a GX@ to 105C. I simply close the GPU that has the least progress.

First, one of these statments is false. Second, if you 'simply close the GPU' until the deadline passes the result is the same as deleting it.

Personally, cut my heat and points.

There are options available RIGHT NOW to do just that.

As for the 'Please stop the point stuff' bit:

letting the non Heat affected WU's go racecourse mode.

leave the option for WU's that Fly on my 3X GX2 systems get the lead out.

the Running client has a more productive WU running.

Faster = more points.
More produtive = more points

[Amaruk]

...but everyone knows what is going wrong with this projects. Let's be honest, they demands too much time to finish, too much CPU overhead and they are totally crystal clear undervalued.

So, please, someone from Stanford can answer why there is so many people hating this WU's ?

I think you may have answered your own question.

Also, I don't think these larger proteins are 'totally crystal clear undervalued' - in fact they may be overvalued. The benchmark for GPU is 1500 PPD. Yet my lowly 8800 GT gets 3344 PPD.

That's a 123% bonus above benchmark.

Higher end cards can exceed 300% above benchmark on these larger proteins.

...but the people with ATI hardware don't seem to hate them.

I certainly don't hate them. In fact, I've been looking forward to them. One of the benefits of high performance clients is their potential to fold larger proteins quickly.

There is a post here about Project 2681 that takes 3 days to run on a dedicated 8-core machine.

Perhaps one day Panda will develop a way for an 8 GPU rig to use all of them to fold one massive protein.

[Fernando_Celio]

Easy guys !

There is no reason to start a war here.

"Cherrypicking" is just what we want to avoid.

And, remember, some people does not know English so well, at point to write, read and understand correctly.

..........................
I'm not from Stanford, but I believe I can answer that question. There seem to be several different reasons. For some, it's an issue of heat. For others, apparently it's an issue of PPD. There may be other reasons, as well, but the people with ATI hardware don't seem to hate them.

In any case, the Pande Group does not condone "cherrypicking".

Two points, Bruce.

First, if there is no complains from ATI camps, lets send all these WU's to them !
I know that Pande Group can easily identify the video cards. Start with wu's that using more than 20% of cpu overhead (in Windows Vista, of course), so I will not have any problem with my SMP's .
Just a joke.
Even because if Pande starts to do that, the AMD guys will start to complain very soon. I have one guy from my team that is complying right now and another one that just stopped to fold with his ATI video card.

Second, you know that any individual, with a minimum knowledgment of network can do "cherrypicking" and never be caught. So, there is a problem yet to be solved.

Best Regards, Fernando.