Suggested Changes to F@h Website

[Jesse_V]

I would just like to point out a few things in the F@h website that I think ought to be changed. Hopefully here we can reach some sort of consensus as to the exact changes, but I thought I should just bring some things up.

1) Screen saver. As far as I know, the screensaver client is no longer available. That was v4, and we're now moving into v7. On this page http://folding.stanford.edu/English/Download it says "The v4.x Screensaver client runs like other screensavers, except that it also runs our calculation in the background. However, a separate Screensaver version is being phased out, and that feature has been rolled in to the graphical clients." I suggest removing this statement, as it seems to imply the screensaver exists when it mostly likely doesn't anymore. Also applies to http://folding.stanford.edu/English/FAQ-main and perhaps to statements throughout http://folding.stanford.edu/English/FAQ-ATI

2) Beta warnings. On the same page as #1, there's this bold warning about backing up your hard drive, and whatnot. I understand that the High Performance Clients use a lot of resources, and can cause some irritating lag at times. In some rare cases there is a possibility that those clients could cause the CPU/GPU to become so hot that they fry, and then it would be good to have a backup of your hard drive anyway. At the very least that beta warnings statement should be removed from the Downloads page, as it is just repeated when they click on the HP Clients link anyway. It also seems overly harsh. I do think "As in the use of any beta software, please make sure to back up your hard drive before installing." is important and should be kept. In my opinion, I think everything else in that bold statement should go, especially because there's already statements about "We are not responsible for computer damage" in the Policy FAQ. All the important statements about HP Clients are stated in regular font anyway. The entirety of this statement applies to this page as well: http://folding.stanford.edu/English/DownloadWinOther . Also, this page: http://folding.stanford.edu/English/Guide#ntoc5 describes the HP Clients really well, and I humbly suggest that something like that be the norm instead.

3) supercomputer comparison. Japan's K computer was recently unveiled, and it runs at 8.162 petaFLOPS. Currently we are at 4.102 native and 6.400 x86 petaFLOPS. So the statements about a supercomputer being really slower than us no longer apply. However, there are still valid arguments such as the expensive of running a supercomputer, and that you have to share it. Also the high-speed bus is not really needed, since clearly protein folding is doing just fine with minimilistic communication. So supercomputers are expensive, shared, and not appropriate. Since 2007 we have been far ahead of any supercomputer, it was only recent that we were beaten. Thus, statements about the slowness of supercomputers isn't accurate at this moment. This applies to these pages: http://folding.stanford.edu/English/FAQ-Press#ntoc4 http://folding.stanford.edu/English/FAQ-main#ntoc8 and http://folding.stanford.edu/English/FAQ-Diseases#ntoc14 and the third paragraph in http://folding.stanford.edu/English/FAQ-Petaflop . But we will beat the K computer, so any fix may only be temporary.

Those are the main issues I've noticed, although there are just some minor updates that could be made in other pages. I hope that when the v7 client goes public there won't be so much scary beta warnings on its page, but rather a calm explanation as to the risks involved as mentioned in #3. Thanks in advance for any explanation, suggestions, or applied changes you have for these.

[7im]

1. Fixed the download mention. The main FAQ needs more work than a quick fix. It's somewhat aged.
2. No change, IMO. Too many people already ignore the beta warning, as common place and as direct as it is. Lessening that warning would only make the problem worse.
3. Please do a little more research on how "supercomputers" get their TFlops ratings. There is a BIG difference between theoretical max TFlops, and actual sustained throughput. FAH is actual throughput. You also came to an incorrect conclusion about high speed bus needs. FAH to a Cluster is apples to oranges.

[Jesse_V]

1. Fixed the download mention. The main FAQ needs more work than a quick fix. It's somewhat aged.
2. No change, IMO. Too many people already ignore the beta warning, as common place and as direct as it is. Lessening that warning would only make the problem worse.
3. Please do a little more research on how "supercomputers" get their TFlops ratings. There is a BIG difference between theoretical max TFlops, and actual throughput. FAH is actual throughput. You also came to an incorrect conclusion about high speed bus needs. FAH to a Cluster is apples to oranges.

1) Thanks. I see the change, and what's there is very nice. I'm happily surprised that changing it was that fast. Thanks!

2) At least on my machine the beta clients seem more stable than the warning implies them to be. yes, they do have "rough edges" and whatnot, but that the beta warning seems to be more appropriate perhaps if you run them as a Beta Tester or something. You seem to be implying that many donors have had some serious problems with them on their machines. That statement must be based your experience on this forum. I still feel like it shouldn't be listed twice though. Listed once in bold should be enough for a "well we told you so". Its not like its in fine print or anything. "Please consider the use of these clients carefully" is a very appropriate statement.

3) I was under the impression that supercomputers get their rating from the LINPACK benchmark. The LINPACK is out-of-date but since its always been used and always gives a single value. The LINPACK does some math calculations and judges the speed of the system based on how fast it can complete those calculations. Theoretical speed is when they add up all the components and try to predict their performance. This of course is never actually attained, so that's why they benchmark. Doesn't the benchmark measure throughput, at least in terms of FLOPS? At least benchmarking works fairly well for F@h. My information about the high speed bus came from http://folding.stanford.edu/English/FAQ-main#ntoc8 and I thought that the high-speed bus isn't needed since we all get a WU which we then return when completed. If we were analyzing massive amounts of data sets at once, then I could see the high-speed bus being handy. But we run hundreds of parallel simulations of tiny individual proteins. So that's where I'm coming from for #3. I understand F@h and a supercomputer are drastically different in terms of design and components, but I'd be happy to learn from you why my conclusions are wrong. Thanks.

EDIT: oh I see your "research" link now. Some interesting stuff there. Thanks for pointing it out. The problem is that the K computer is still faster than us. According to this page http://www.top500.org/lists/2011/06/press-release the K computer uses CPUs instead of GPUs. So therefore the LINPACK is likely more accurate than it was for testing Tianhe-IA. The LINPACK has problems, but it's standard and its result can obviously make a supercomputer the top of the list. People very likely just look at that petaFLOP result and think of being able to compare throughput/speed. In my opinion, the statements on the website should be carefully reworded. They are still expensive and provide more features than just raw CPU power. Plus, if F@h had been run on a supercomputer, the PG wouldn't have been able to write the book on large-scale distributed molecular dynamics.

[7im]

2. Please don't make me point out the many "I told you so posts" even with all the beta warning redundancies, boldness, and bright red text. Yes, the current betas are a little more solid lately. However, the beta warning is general purpose. PG is not going to reword the warning each time the current beta gets better or worse than before.

I'm sure the download page will be revamped as we get closer to the V7 release. I'll ask that any redundancies or harshness be reviewed/updated.

3. You've read the http://folding.stanford.edu/English/FAQ-flops right?

Yes, that research link talks about a GPU supercomputer, but you didn't seem to grasp the main idea that related to comparing FAH to a supercomputer. FAH is real world sustained throughput. Linpack, used on the supercomputers, tests in short bursts, shows peak performance only, not sustained performance. Apples to oranges.

These quotes make that point clear, with comparing FAH with ANY supercomputer.

"... it's not clear that the Linpack benchmark which pegs the machine as the world's fastest is a useful indicator of its performance in real-world applications.

The Linpack benchmark is one of those interesting phenomena -- almost anyone who knows about it will deride its utility," says Dunning. "They understand its limitations but it has mindshare because it's the one number we've all bought into over the years."

"There's how well is it going to perform on the Linpack benchmark, and then how well will it perform on a broad range of science and engineering applications," says Dunning. "We know of no project in the U.S. that will perform as well as [this supercomputer] on a broad range of science and engineering projects."

Clearly, Linpack is a legacy test method kept around long past it's usefulness for the sake of continuity and expediency, not for it's efficacy as a performance predictor on data intensive real world tasks.

Also see the post by Dr. Pande about Flops and Supercomputers...http://foldingforum.org/viewtopic.php?p=67416#p67416

Here is a good article from Nature.com discussing FAH vs. Supercomputers... http://www.nature.com/embor/journal/v5/ ... 00108.html Dr. Pande is quoted here as well.

Your homework assignment is to write 2 paragraphs (~150 words or less, each P) comparing and contrasting distributed computing vs. super computing, as related to FLOPS comparisons. No footnoting required. Just 2 simple paragraphs that might fit well at the end of that FLOPS FAQ.

[VijayPande]

PS We're working on a site overhaul, especially for the front page and the download page. The download page in particular is really crazy complicated and the front page could also be made more friendly for newbies. These are the highest priority for our web team. My hope is to have this ready in time for the v7 release completing to the point where we can put it on our main download page.

[Jesse_V]

PS We're working on a site overhaul, especially for the front page and the download page. The download page in particular is really crazy complicated and the front page could also be made more friendly for newbies. These are the highest priority for our web team. My hope is to have this ready in time for the v7 release completing to the point where we can put it on our main download page.

Ah thanks. Hope to see some really nice pages! Hopefully everyone's suggestions will be considered, but I'm confident your team will make it look really great. May I also take this opportunity to add my congratulations on your latest v7 release, and how much I approve of and appreciate such client unification. Looking forward to early next year when all this client-choosing craziness can finally start being done with!

In the meantime I'll start doing some studying for Professor 7im...

[7im]

Kidding about the 2 paragraphs. But you'll need to understand the differences there as the basis for a deeper discussion on bus speeds.

[Jesse_V]

Kidding about the 2 paragraphs. But you'll need to understand the differences there as the basis for a deeper discussion on bus speeds.

Of course. Clearly my suggestions are wrong, but as long as the website statements about the F@h-supercomputer comparison remain accurate and up-to-date, that's all I'm shooting for there. Perhaps when the site gets overhauled those pages will get reexamined, but in the meantime I didn't want those statements to sound obsolete. I made the mistake of seeing two FLOP/sec counts and thought that was a good metric to compare computing systems. For the others that do this, I want to make sure that the website explains things properly. A valid point is "well if this research is so important, why don't they use a supercomputer?" and I just wanted to make sure that argument was effectively countered. Thanks for the insights.

[Jesse_V]

One more thing about the website I found: on this page http://folding.stanford.edu/English/FAQ-Uninstall#ntoc1 the links of "United Devices" and "Entropia" don't appear to work properly. Suggest removal, and possibly replacement with "..., like PrimeGrid, SETI@home, World Community Grid, and many others."

Thanks. I wonder just how much will be overhauled. In any event, it will be a big task!

EDIT: on the same page, specifically this section, http://folding.stanford.edu/English/FAQ-Uninstall#ntoc5, the "illegal installations" link does not work. Less than half of the page is about "uninstalling", but I think its quite likely that that page will be significantly changed in the coming few months.

[7im]

There are people who find this page because they come across an unauthorized installation of FAH and just want it gone because they don't know what it is.

But when you explain what FAH is about, they may still uninstall it, but then they also understand it wasn't a virus and didn't do any damage. They may also become interested in the project, and run it on a personal machine. That's why the page is more than just how to rip it out.

P.S. Dead links replaced.

[Jesse_V]

There are people who find this page because they come across an unauthorized installation of FAH and just want it gone because they don't know what it is.

But when you explain what FAH is about, they may still uninstall it, but then they also understand it wasn't a virus and didn't do any damage. They may also become interested in the project, and run it on a personal machine. That's why the page is more than just how to rip it out.

P.S. Dead links replaced.

Ah. Thank you for explaining! In that case I would suggest having uninstall things first, and then the "What is Folding@home anyway" topics afterwards. Its named "Uninstall" after all. Do you agree with its organization?

Thanks for replacing the deadlinks. May seem weird, but this is one of the reasons I enjoy distributed computing. Donating money to the American Cancer Institute is great, but what do you get in return? In the immediate sense, probably nothing. Do cancer/Alzheimer's/Parkinson's/etc work on your computer essentially for free, and you get this great community, get updated and in-the-know as to exactly what science is going on, how our progress compares to others, talk to others doing the same thing you're doing, and you get to change the website to boot. Awesome.

[7im]

I agree with the layout of that page to the extent that the Table of Contents is at the top of the page, and visitors can click on any link they want to skip down to any topic they want.

[Jesse_V]

Also, should be a quick thing for you to fix, but here: http://folding.stanford.edu/English/FAQ-Diseases#ntoc11 the title of "Osteogensis imperfecta" is spelled wrong. Should have another "e" in "Osteogensis" and "imperfecta" ought to match the capitalization in the description, so that it becomes the correct "Osteogenesis Imperfecta"

[7im]

That, and 2 other mispells, fixed. Total papers updated to 95.

[Jesse_V]

That, and 2 other mispells, fixed. Total papers updated to 95.

Awesome. Thanks so much!

By the way, new section: http://en.wikipedia.org/wiki/Folding@ho ... gnificance