Protein Folding: Your Chance To Help (and Whup AnandTech) Starting Dec 1

[lilcinw]

At this point we need 20+ 1080s to catch them.

Why don't you head over to the GPU forum and start recruiting?

 

[bit_user]

I was poking around to see if anyone is using FPGAs, for this. Doesn't seem like it, but if you had one that appeared as an OpenCL device, it seems like it could. Once FPGA-enabled Xeons become more common, I think we might start to see it.

Speaking of Xeons, I wonder if Xeon Phi is supported and how fast they are.

Otherwise, the fastest thing would be a Pascal Titan X (but the best folding per $ would probably be a GTX 1070). My GTX 980 Ti ends up aborting about 33% of its work units, due to reaching the MAX_ERRORS limit. If that limit were set just a bit higher (like 5 instead of 3), most would clear.

 

[bit_user]

http://fah-web.stanford.edu/ is offline. See https://folding.stanford.edu/home/stats-system-is-live-and-transfered/ for new stats server.

 

[bit_user]

So, the results are in. I think we made a good showing, but it wasn't enough.

https://forums.anandtech.com/threads/11th-annual-folding-home-holiday-season-3-way-race.2489602/?_ga=1.75678165.1367970780.1399926135


I'm going to stop, now. I'm disappointed in FAH for failing to address the Nvidia Maxwell (900-series) bad Work Unit problem and I'm not going to waste energy computing so many work units that just end up aborting. If they want me to take them seriously, they should also release proper Ubuntu .deb packages (unless they can fix their generic .deb packages).

This, combined with no 64-bit Windows release, or any official release since mid-2014, makes me feel like FAH is a zombie project. Maybe deep learning is absorbing all the time & energy of people who used to maintain it?

I think FPGAs (and quantum computers) will be a game-changer for folding, in the coming years.

 

[lilcinw]

FAH did address the problem with the new NVidia drivers; it is not in their power to fix.

https://foldingathome.stanford.edu/2016/10/27/core-21-openmm-gpu-issues-new-windows-nvidia-drivers-375-57-375-63/

Judging the "life" of the project by Client software releases is inappropriate. The client software is just the management side of the actual folding work and has no bearing on the results. The majority of the development has always been on the cores themselves as optimizations and improvements there lead to better, faster results.

There is a beta version of the client software but they do not advertise it as they don't want casual folders to develop a poor perception if thing don't work flawlessly. I personally participated in the v7 beta and have not upgraded my client since. I almost never have problems because I am extremely cautious about upgrading my drivers.

If you go to their news page you can see that the project is "lively".

https://foldingathome.stanford.edu/news/

There is a reason that the Pande group has not released (to my knowledge) a Xeon Phi core but have released an Android app. They are focused on widespread adoption. If their are 200 potential Xeon Phis to tap into but 2,000,000 potential android phones, their preference is for the phones.

 

[grimfox]

I'll have to keep an eye out for the next race and try to get my GPU into the fray. I'm a bit concerned about running it that hard for a long time period. Also I pay my electric bill so it'll be interesting to see a comparison between this year where I folded for a week and last year where I did not. That said my new PC would use only a little more power running both CPU and GPU cores compared to just CPU cores on the older PC.

 

[lilcinw]

I have folded GPUs 24/7 for years. As long as your cooling and PSU are good they shouldn't be affected.

 

[bit_user]

FAH did address the problem with the new NVidia drivers; it is not in their power to fix.

Nope, it seems you haven't been following my posts. I'm talking about this:

https://folding.stanford.edu/home/issues-with-fahcore21/

Speaking as a developer, I find their apparent lack of follow-up quite unsettling.

Judging the "life" of the project by Client software releases is inappropriate.

I find your highly selective reading of my post, in light of your apparent defensiveness, to be inappropriate.

My comment about their lack of releases was only one, fairly minor point. If there software didn't have the other issues I cited, then I wouldn't mind the lack of releases.

I actually found their beta, but was put off when faced with a login prompt to download it. If I had found any reason to believe it would solve my problems, then I'd have tried it.

The client software is just the management side of the actual folding work and has no bearing on the results.

On Windows, it's a single package. All of the .EXE's currently in my install dir are 32-bit. For CPU folding, this definitely matters.

If you go to their news page you can see that the project is "lively".

That was really a minor point. I don't care how lively it is. What I really care about is how well they maintain their software. I listed 3 issues (granted, I'd agree that 64-bit CPU folding is a lower priority), two of which I judge to significantly impact participation. So, what I'm really saying is that I feel they don't take these things seriously. And I'm not about to put up with sub-standard software that creates an ongoing headache for me on Ubuntu and results in wasted computation of about 1/3rd of my WU's on both platforms. Whether it's lively or not, those are deal-breakers, for me.

There is a reason that the Pande group has not released (to my knowledge) a Xeon Phi core

I don't know why you're even telling me this. I don't have a Xeon Phi. I'm not going to get one. I never said it should be a priority. I simply mused about whether it was supported (via OpenCL) and how fast it might be. Purely as a matter of technical curiosity.

If the goal of your post was to keep me from being dissuaded, it was more counter-productive than anything. And if you're simply trying to evangelize the project, then I can't see how trying to debate with me is going to help your cause.

 

[bit_user]

I have folded GPUs 24/7 for years. As long as your cooling and PSU are good they shouldn't be affected.

Let's hope @grimfox doesn't have an EVGA ACX 3.0 card. In that case, you might be wrong.

http://www.tomshardware.com/news/evga-addresses-geforce-1080-temperature-problems,32967.html

I had similar reservations about running my desktop GPU for a whole week. In the end, I decided that since it's still under warranty, I'd rather stress it now & get a replacement if it's at all marginal. Of course, that assumes it's not going to fry my PC like some of the ACX 3.0 cards. But I think such issues are pretty rare, thankfully.

BTW, I'd estimate participating in this challenge has cost me about $9, or so. Once you know how much you're paying per kWH and how much power your GPU uses in practice, it's pretty easy to work out.

The measured power utilization of my GPU averaged around 210 W, even though it's rated for up to 275 W. The most I've seen it use, during folding, was about 225 W. Then, there's overhead for the platform (LGA2011-based workstation). I was able to measure at the GPU and at the wall.

 

[grimfox]

Asus Strix 1070 in the new PC. PowerColor 7870 myst edition in the old PC.

 

[lilcinw]

Nope, it seems you haven't been following my posts. I'm talking about this:

https://folding.stanford.edu/home/issues-with-fahcore21/

Speaking as a developer, I find their apparent lack of follow-up quite unsettling.

I indeed misinterpreted your complaint. I have seen numerous users experiencing problems with the latest drivers and assumed that was what you were discussing. I have folded on a GTX 970 for almost two years and was not aware of this issue. This is probably due to my 'set it and forget it' mentality with regards to folding.

I find your highly selective reading of my post, in light of your apparent defensiveness, to be inappropriate.

My comment about their lack of releases was only one, fairly minor point. If there software didn't have the other issues I cited, then I wouldn't mind the lack of releases.

I actually found their beta, but was put off when faced with a login prompt to download it. If I had found any reason to believe it would solve my problems, then I'd have tried it.

On Windows, it's a single package. All of the .EXE's currently in my install dir are 32-bit. For CPU folding, this definitely matters.

That was really a minor point. I don't care how lively it is. What I really care about is how well they maintain their software. I listed 3 issues (granted, I'd agree that 64-bit CPU folding is a lower priority), two of which I judge to significantly impact participation. So, what I'm really saying is that I feel they don't take these things seriously. And I'm not about to put up with sub-standard software that creates an ongoing headache for me on Ubuntu and results in wasted computation of about 1/3rd of my WU's on both platforms. Whether it's lively or not, those are deal-breakers, for me.

When you first brought up your problems with linux I tried to point you to another thread where you could receive assistance. As I don't have experience with folding on linux and you to my knowledge have never posted there to ask the team members who do I did not feel qualified or interested in discussing it.

If you would like the 64-bit beta client it is available here:
https://folding.stanford.edu/home/beta/

I don't know why you're even telling me this. I don't have a Xeon Phi. I'm not going to get one. I never said it should be a priority. I simply mused about whether it was supported (via OpenCL) and how fast it might be. Purely as a matter of technical curiosity.

The reason I brought up the Phi/Android comparison was because of your mention of FPGAs and quantum computing. I don't see F@H getting involved with them in a large scale fashion any time soon as they are not mainstream enough that the user base will likely adopt them.

If the goal of your post was to keep me from being dissuaded, it was more counter-productive than anything. And if you're simply trying to evangelize the project, then I can't see how trying to debate with me is going to help your cause.

I will admit that the tone of my response was rather defensive. The reason being that most of your posts up until this point have been you reporting problems with little to no engagement with those who tried to assist you.

Thank you for your contribution during the race. If you do not feel the need to continue folding that is your choice. I believe that Stanford is doing important research with our help and will continue to support that effort /evangelism.

 

[bit_user]

When you first brought up your problems with linux I tried to point you to another thread where you could receive assistance. As I don't have experience with folding on linux and you to my knowledge have never posted there to ask the team members who do I did not feel qualified or interested in discussing it.

Yeah, thanks for that. But I had already gotten it working, by then. The main problem with Linux is that they only published generic .deb packages. The client package has a dependency which Ubuntu lacks, which means that the package manager thinks the client package is broken and keeps wanting to remove it.

Yeah, I could download the sources and rebuild it. And if that were my only issue, I might. But it's not, and with Ubuntu being one of the dominant distros, it's pretty sad they didn't fix this. I think a lot of users would (and have) given up, when it didn't just install correctly, like nearly every other package they use.

It's a pretty simple packaging issue, and it's pretty rare to see packages broken, like this. And the .rpm packages seem to have their own set of issues. So, that just contributed to the feeling of it being (almost) abandon-ware.

The reason I brought up the Phi/Android comparison was because of your mention of FPGAs and quantum computing. I don't see F@H getting involved with them in a large scale fashion any time soon as they are not mainstream enough that the user base will likely adopt them.

I thought I saw something about CureCoin using some sort of ASIC, but I guess that's for blockchain and not the folding part. But, if there are big, institutional folders, then support for non-commodity hardware would be worthwhile.

And if a quantum computer is faster than 10^9 android phones, that might be worth supporting (assuming someone donates the time on one). That's just a random number I made up, but I have read that folding is specifically one of the areas where quantum computers are expected to excel.

I don't know. I don't have a dog in this fight. I'm not trying to dissuade people from folding. I'm just musing, so feel free to ignore. I certainly don't think people should base their decision about whether to participate on speculation about future technology. However, I enjoy discussing tech and trying to make predictions, which I think should be appropriate, on such tech sites as this.

The reason being that most of your posts up until this point have been you reporting problems with little to no engagement with those who tried to assist you.

I didn't need to post in other threads, because it turns out so many others have encountered the same things that you can pretty much just google them.

Thank you for your contribution during the race. If you do not feel the need to continue folding that is your choice. I believe that Stanford is doing important research with our help and will continue to support that effort /evangelism.

I already spent what is, for me, a lot of time and energy tripping over, investigating, and working around problems that have affected others, as well. To me, that spells trouble for a project that's dependent on mass-participation. I'm just pointing that out.

Regardless, I'd have to say my GPU setup is a bit too expensive for me to keep running (especially since I pay for air conditioning and electricity, but not heat). So, even if I weren't getting any WU's rejected, I'd still be reluctant to continue folding, long-term.