Folding Forum

cristipurdel wrote:...I simply hate the fah client.
I will start again only if there was a wrapper, so that I can also manage other projects.
I know I'm not representing a large group, but that's the way I feel.

If it is because of the hassle involved with configuring the Clients, then there is good news. A completely new Client V7 is being written by professional programmers (previous Clients were written by PG who were first scientist then programmers) and it is expected to reach open public BETA within two or less months provided that there aren't any major bugs discovered which will delay it. Maybe you can try F@H by the end of this year so hopefully the Client will be out of BETA Stage and finalized.

Bruce, I'm Polish. Pande Group treats folders as the communist authorities used to treat Poles. Pande Group is always right, Pande Group needs time to make it good, Pande Group must have everything under control and everything has to be secret, etc.
Explain to me please what your beta testers do, because from what I can see that we're beta testers. For example, why the Pande Group, opened the new Linux fahcore_a3 and after a few days blocked. Probably it was never tested.

The v7 client will have some new features which will help donors manage multiple GPUs, CPUs, and even multiple machines. Please check it out once it's released to see if it suits your needs.

filu wrote:Bruce, I'm Polish. Pande Group treats folders as the communist authorities used to treat Poles. Pande Group is always right, Pande Group needs time to make it good, Pande Group must have everything under control and everything has to be secret, etc.
Explain to me please what your beta testers do, because from what I can see that we're beta testers. For example, why the Pande Group, opened the new Linux fahcore_a3 and after a few days blocked. Probably it was never tested.

I'm sorry you feel that way. The reason for having a separate beta tester group is to catch bugs before release. If you like, the Mods can make you part of the beta testing group and you can see what goes on there. However, note that this group has special responsibilities to give us feedback to try to improve our software before release.

I also think your statement that "Pande Group is always right" is not an accurate characterization. We are quick to admit our mistakes and move on (there are many examples of this on the forum). However, our priorities can be different than some donors -- we are trying to push the science as far as possible, while some donors are more interested in the competition aspects in FAH. I'm not saying that's a bad thing, just that such differences in goals can lead to differences in opinions on the priorities for future work.

It's nice to see that the fah client can improve so much, and no need for micro-managing will be required. SMP along with GPU clients it's a real nice touch.

But, it would be nice to visualize somehow the progress made by the donors. I'm looking at wcg and what tremendous job they did with their site. You can choose what kind of project to run, and you can see how much time you spent on a project. It's really neat. I don't think it would be that difficult for fah to setup a stats page, where you can choose what type of research you want to contribute, and receive the appropriate project number. And quantify it afterward (eg. Alzheimer, sum of WU (proj nr. x + proj nr.y + ...) ), even put some sort of icon,badge for when you reach 100WU, or one million.
From my page http://fah-web.stanford.edu/cgi-bin/mai ... e=teampage I'm not getting any tangible information.
Maybe I would prefer to crunch only for cancer research, since members of my family died from it, and I wouldn't like to crunch for eg. AIDS or Alzheimer.
The above example is taking from WCG. If theirs is working, why don't you use it. The rest of the boinc doesn't see badges or which crunching projects they are using. They just multiply the wcg points by a coefficient and that's it, just one number which you add to your stats.

Regarding the wrapper , not to be off-topic, I can present two extreme, but possible scenarios.

1. Worst scenario: there is no wrapper, fah ATI servers clients are down, for one week. Thousands of GPUs are sitting idle, and for one week, 1PFLOP is sitting idle. Users become frustrating, maybe some of them start leaving.

2. Best scenario: there is a wrapper, fah ATI servers do go down, everybody starts crunching for their back-up project or their shared projects.Some will complain, but they'll keep on crunching.
Best part, what if gpugrid or wcg goes down, where do you think they'll end up?

I know that these are extreme cases and they don't apply to all the folders, but that's what I would do.

My suggestion, have a chat with wcg in terms of managing the stats, gui, security and hatred towards boinc

Of course, I'll try the v7 when it comes out, and probably will like it, but I will not want to run boinc and folding in the same time, micro-managing them, just in case folding has some server issues.
Right now, on one machine I'm running boinc (cpu clients) in ubuntu under virtualbox, and on windows I'm running boinc cpu & gpu with back-up projects (in case the main projects don't have work), and also a non intensive cpu project.

If my post has offended anyone, it wasn't my intention.

Alternately, many people arrived at this project because they disliked BOINC. And they continue to fold dispite a few client and server rough patches from time to time because they have no interest in running other projects.

If someone writes a wrapper for fah, and it doesn't screw stuff up in the process, then more power to them. If done well enough, Stanford might even make a EULA exception. But if done badly, it will be blocked quickly. Best of luck.

7im wrote:If someone writes a wrapper for fah, and it doesn't screw stuff up in the process, then more power to them. If done well enough, Stanford might even make a EULA exception. But if done badly, it will be blocked quickly. Best of luck.

On this page you will find the authors of wrapper for f@h. http://www.dnetc.net/.
Sesef and OxyOne are Poles, so you will not find anything yet English. I tried to contact them, but so far not received any response.

Wrapper will be done under the GPU client. The server will be packed in packages of 4 wu to 1 wu, wu is one of the four will be F@H. At this point, ATI's GPU will be charged at 100%, thus will be 2-3 times more efficient than NV. So now they have done in the DNETC.

Old discussions rehashed, fun thread though.

I'm waiting for v7 as well, especially since it seems to do what I wanted to do myself and failed to do so ( because there was no way to comply with all EULA's and still offer a boinc like GUI to install/monitor and manage clients ).

There was no more reply to the thread started by the developers after I asked to either open up the api so 3rd party dev could write such an interface, or confirm that it would not be needed anymore. I even stayed of the forums and decided to stop folding partly because of that ( and partly because I had other priorities like a newborn ). But reading this thread ( I know.. just now, I have been away for months before coming back I am sure there are tons of other threads I missed out on ) makes me feel happy that the many many suggestion which were made long long time ago finally are going to show up.

filu wrote:Wrapper will be done under the GPU client. The server will be packed in packages of 4 wu to 1 wu, wu is one of the four will be F@H. At this point, ATI's GPU will be charged at 100%, thus will be 2-3 times more efficient than NV. So now they have done in the DNETC.

What do you mean by "charged at 100%"?

How do you plan to make ATI 2-3 more efficient than NV?

If you're actually talking about FAH that would require rewriting the analysis code known as FAHCORE_11. It has nothing to do with the client or whether that client runs under a BOINC shell or not.

On the other hand you may be talking about something entirely different than FAH. If that's the case, you're only here to recruit for some other project, and recruitment is prohibited on this forum.

On the previous page, someone requested the source of the rumors. That question is still applicable today. If the facts can be verified and they apply to FAH, we'd all be very interested. If the facts cannot be verified or they apply to something other than FAH, the Mods need to deal with the topic in another way.

codysluder wrote:How do you plan to make ATI 2-3 more efficient than NV?

This is not my idea. This is the plan of the man who created a wrapper for DNECT.

I am not going to recruit anyone. Here are my stats http://folding.extremeoverclocking.com/ ... =&u=442804.
Unfortunately, PG can not cope with the customer at the ATI graphics cards. It is also true that many people just use these cards and that they are frustrated by the fact that outdated NV cards have better PPD from modern ATI cards. My goal is to draw attention to this problem that PG may react in advance. If the wrapper comes from the beta phase is likely that users of ATI cards will go to BOINC.

codysluder wrote:On the previous page, someone requested the source of the rumors. That question is still applicable today. If the facts can be verified and they apply to FAH, we'd all be very interested. If the facts cannot be verified or they apply to something other than FAH, the Mods need to deal with the topic in another way.

On the previous page is a link to the Polish forum of BOINC. Even if I knew the source of wrapper, and so I do not give anyone. I spent too much of my time, money and effort for the F@H project.

I would never go to BOINC, it is forbidden by my religion

And my ATI card is currently having a rest ... it's in my gaming machine and I only use it when I play ... until the OpenCL client saves us

toTOW wrote:I would never go to BOINC, it is forbidden by my religion

Once I had that kind of attitude, but things happen and times change.

My goal is to draw attention to this problem that PG may react in advance. If the wrapper comes from the beta phase is likely that users of ATI cards will go to BOINC.

However, you can't react in advance to something that has been going on for some years now.
Hence, you are about 2 years late to the discussion.

In a couple of months AMD 6XXX series cards will launch. There are rumors of improved shader design and 256 bit controller bus on lower end cards.
Combined with a new GPU 3 client there might be hope.

filu wrote:Unfortunately, PG can not cope with the customer at the ATI graphics cards. It is also true that many people just use these cards and that they are frustrated by the fact that outdated NV cards have better PPD from modern ATI cards. My goal is to draw attention to this problem that PG may react in advance. If the wrapper comes from the beta phase is likely that users of ATI cards will go to BOINC.

PG can cope with the customer with the ATI graphics card, it's ATI that cannot. The fact that CUDA is much more efficient that Brook means that NVidia is doing a better job of supporting the type of computing that FAH needs than ATI is. A BOINC wrapper for FAH will not change that fact. (Hopefully OpenMM/OpenCL will, but who knows when.) BOINC will still have to use the Gromacs/Brook software provided by Stanford for ATI just as they would have to use the same Gromacs/CUDA software that Stanford provides for NV.

Wrappers do not change the analysis software or its interface with your hardware. BOINC will not change those pieces either. BOINC is a tool allowing a distributed computing developer to use a standard structure for servers and points and for a user interface, but it's NOT a tool that defines the analysis that's done, that's the responsibility of the project itself. Stanford had already developed a structure for servers and points and a user interface before BOINC was invented so there's no particular advantage to discarding it and replacing a customized tool with a general purpose tool.

If someone does develop a wrapper, I don't think Stanford would mind (since there would potentially be more people folding) but it will not solve any differences between ATI and NVidia.

And since there's no way to compare points or results, how can you say whether running some other projecct on BOINC is either more effective or less effective than running FAH on a particular GPU? The only comparison that's accurate is the comparison between FAH/Brook/ATI and FAH/CUDA/NVidia so claiming that BOINC/FAH/Brook/ATI will "steal" donors from FAH is ludicrous.