Folding Forum

OK, thanks for that. After running for a while, the eta exceeded the maximum allowed, so I just paused it and uninstalled it. IntelAtom N455, at least on this crappy little netbook, will try to work, but is definitely not suitable for purpose.

2 Weeks in, 4 old laptops are still plugging away. I've had 1 cpu fail - replaced with a spare, otherwise working away available 24/7 and folding more than not. I added a headless Win10 box about 10 days ago which seems to be returning 1 low points wu a day. It has 4 core of some sort, which I'm guessing is running on somewhere 1-3 of those as there is a recognised GPU which doesn't seem to be picking up any work. It's headless as an HDMI/VGA adaptor failed just after I'd installed FAH, and I can't access anything to do any better setup until a new one arrives.

According to the stats txt file, I've folded 64 WU. I think I've had 2 fail, and know of 2 more which have exceeded Timeout, which doesn't sound like a horrible rate to me.

I'm still unsure whether folding slowly is helping or hindering at this stage, but I think that as long as the system is server constrained, folding faster would just reach server overload faster/more often, leaving the fast folders unable to find work just as often.

It would be nice to see what the overall picture is in terms of end to end process, including the preparation of wu by scientists, pool of unfolded wu (particularly whether this number is rising or falling), and at the other end of the process, number of fully folded projects still waiting for a scientist to do something with them. All luxuries at the moment, but unless you can identify your pinch points in the overall process, efficiency is hard to achieve.

Not in any way meant as a criticism. This is by a big margin the most impressive IT project I've ever been a (microscopic) part of.

It is kind of crazy in magnitude isn't it … Very easy to get tied up about little bits when actually just stepping back and looking in amazement at what has been created it breath-taking … In this case it ought to be "look after the pennies and the gold bullion will look after itself" … Thanks for being part of it all

peterjammo wrote:It would be nice to see what the overall picture is in terms of end to end process, including the preparation of wu by scientists, pool of unfolded wu (particularly whether this number is rising or falling), and at the other end of the process, number of fully folded projects still waiting for a scientist to do something with them. All luxuries at the moment, but unless you can identify your pinch points in the overall process, efficiency is hard to achieve.

Those kind of luxuries sometimes refresh me when I've spent 12 or 14 hours answering mostly the same 6 questions that everybody has to ask when they first join this forum. (ore new landing page has helped. SOME people are now using the search function.

The work being done by our technical people are the most visible, concentrating on throughput limitations such as server bandwidth or bugs that delay the points getting added to the stats DB.

Behind the scenes there are an increasing number of scientists generating unfolded WUs for distribution. They are very busy, too. Part of their job is to dream up new things that need research. but just resupplying the servers with new WUs for active projects can be a challenge. If they fall behind, we can't see it. If there are no more WUs for your CPU or your GPU on any of the work servers, you get the same message as when all of the servers are running at maximum bandwidths. That doesn't happen very often but when alarm bells go off, those people do respond. [Oh, and the message isn't intentionally misleading to hide that human limitation, The fact is that the assignment server is telling you "I can't find a server with WUs that can satisfy your request" is true, whether each WS has no remaining usable bandwidth or that WS is down or that WS has run out of WUs. It just means a different person has to respond.

Notice the word "increasing" in that paragraph. I'm seeing many old friends who ran FAH projects years and years ago and have been away. They can hit the ground running without need for much of a skills upgrade pick an interesting thing to study and start building WUs ... provided we have the server capacity for them to manage. (Some of them even bring their own server.)

One step that you've left out is the data reduction. When a project has been active for XX days, the researcher faces a (virtual) pile of completed WUs that are not yet useful to the world-wide scientific community. I don't know enough about this step to comment, but it's still an essential step for each scientist.

In normal times, the final step would be to write a research paper to be submitted for peer-review and publication. That'll probably happen, but I expect that some of the world-wide scientific community can also make use of the output of the data reduction step above. I've seen some recent video presentations that are similar in nature to a traditional scientific paper or TED talk and they seem to be a more rapid method of getting useful information out to the public.

For example, we're running a lot of simulations of various ligands against interesting binding sites on the virus. This is well in advance of any actual drug trials. I suppose that promising results may be exported by telephone before the preparation for a video or the typing of a formal conclusion. Given the nature of the COVAID crisis and the speed at which discoveries are being made, it's hard to say what's actually happening on the back end of the FAH's research and the front-end of whatever scientific team can use what we discover.

Not in any way meant as a criticism. This is by a big margin the most impressive IT project I've ever been a (microscopic) part of.

Amen

peterjammo wrote:OK, thanks for that. After running for a while, the eta exceeded the maximum allowed, so I just paused it and uninstalled it. IntelAtom N455, at least on this crappy little netbook, will try to work, but is definitely not suitable for purpose.

I've got an N455 running on a M/B with a GPU slot on it. It's really useful to drive a GPU (and it can still do some projects before they expire). Is there a way to hook up an eGPU?

peterjammo wrote:...Once completed, paused the running wu, removed the now idle "cpu" and restarted with setting -1. Only 1 core folding. Paused and set to 2, now folding using both. No change to eta or estimated TPF, but maybe that won't update until the next wu starts...

Do note that the ETA will change after an increment of 3% so it takes time. Moreover, WUs from previous projects will display the old value since your client "knows" how long they used to take so will display the old information. Once the WU has incremented by 3% the new ETA will slow change and the updated value will be remembered for future WUs from the same project.

bruce wrote:Is there a way to hook up an eGPU?

I don't know. The cheapest eGPU is way over my budget.

It really is a horrible little netbook. It belonged to my daughter and she only used it a short time before it got terminally stuck trying to update Win 7. I resurrected it several years on to run Xenial Pup for my 4 yo grandson to watch youtube videos. Several letter keys have died now (common fault apparently and no 2nd hand keyboards about) so I used an airmouse board to set up FAH on Mint 19.3. The wireless won't work on Mint, and I've run out of ethernet cables. Really only good for the bin.

Oh and on the long post above from Bruce - that fills in a whole lot of detail about the big picture. Thanks for taking the time to write it.

With a stay-at-home order from our Governor I sometimes take the time to be verbose.