Folding Forum

The power of NVidia Saturn V shows I think where much of the GPU work goes. In April alone it's done more than three times what I've done in over fifteen years and I'm only just outside the top 150 of all time.

I do worry what happens when this dies down to those corporate superusers - NVidia, Amazon, CERN. Of the top ten users by 24 hour ppd average, only Anonymous existed before April.

Neil-B wrote:I have a suspicion that complete GPU Projects get turned around much quicker than the CPU ones which might exacerbate the lack of GPU Projects situation .. my gut tells me that some very large GPU folding resources ramp up when GPU Projects become available/plentiful and ensures the Projects complete as quickly as possible.

You're probably right. When one of my slots is empty, it's usually a GPU slot (one CPU slot, two GPU slots).

Do we still have an excess of users, relative to WUs? Would more users right now help?

I ask as well because I suspect that it has something to do with longer-term retention. People who sign up and don't see it doing anything are probably more likely to uninstall. (I suppose F@H could empirically examine this, since it knows when people stop being active, and also knows what WUs people did, presumably.)

My oldest slowest machine has been folding CPU wu for about a month. It's probably the lowest spec that could reasonably be used. It's on it's own username and is not signed up for QRB. It's returned 33 wu. It's currently lying about position 736,000 of 2.4 million usernames on the daily stats txt file sorted by points. I thought I'd read somewhere that that covered stats of users who had been active in the last 50 days, but I can't find a source for that. Looking down the bottom of the list there are approx 800,000 names who have under 1000 points, many of which have only folded 1 wu.

From those bits of info, I'd guess that a big percentage of recent joiners haven't been doing so too seriously. Even if I'm wrong on the 50 days bit,and the files are showing all time, I think that the same applies, just that if that is the case, there have always been a lot of folk dipping in a toe and going home again. Not sure if this helps the discussion at all.

Many many users never bother to configure a username and passkey and fold anonymously which is why Anonymous is the largest producer by far.

HaloJones wrote:The power of NVidia Saturn V shows I think where much of the GPU work goes. In April alone it's done more than three times what I've done in over fifteen years and I'm only just outside the top 150 of all time.
I do worry what happens when this dies down to those corporate superusers - NVidia, Amazon, CERN. Of the top ten users by 24 hour ppd average, only Anonymous existed before April.

It's like you're posting for me. Completely agreed. Maybe it's selfish because the ends are justified or whatever, but it makes me uncomfortable to see what had been a balanced ecosystem for many years get turned over.

schertt wrote:

HaloJones wrote:The power of NVidia Saturn V shows I think where much of the GPU work goes. In April alone it's done more than three times what I've done in over fifteen years and I'm only just outside the top 150 of all time.
I do worry what happens when this dies down to those corporate superusers - NVidia, Amazon, CERN. Of the top ten users by 24 hour ppd average, only Anonymous existed before April.

It's like you're posting for me. Completely agreed. Maybe it's selfish because the ends are justified or whatever, but it makes me uncomfortable to see what had been a balanced ecosystem for many years get turned over.

I understand this, Saturn V; thanks for all the work and I hope you stick around. I have been involved in the project for 11+ years and sometimes there have been sudden burst donors and this has been beneficial on the whole, this time is different with many corporate/institutional type donors which has disrupted the previous culture of small donors driving the project long term. Maybe they will stay, or at least stay enough, but I am worried that work unit/servers shortages may drive away people who could have been long term small donors. Having said that I remain optimistic, enough to have just ordered a open box GTX1660ti to add to my arsenal for the immediate future even though my 2 current cards are not currently fully engaged.

Much as I am in awe of Saturn V's power, I'd much prefer them to be taking the excess once the "ordinary" folders have been given work to do. For me, it's more beneficial to have a million individual folders than half that and one corporate equivalent to half a million

We know that many people are unable to get work to keep their systems occupied and that can lead to them abandoning the project. When the CR benefit of participating is gone and NVidia stop, those lost individuals won't come back.

At this point in time tbh I'd prefer the science completed by the fastest route possible … and if the big "corporate" donations are making this happen then I don't honestly care if I don't get any WUs … For long term strategic planning yes I agree home users should get priority (but we could then debate whether people with a single slow bit of kit should get priority over those who invest in bigger farms of kit) however for now I am truly grateful for the support of the Saturn V, CERN, EC2Spot style massive donations because science is being done way quicker than it would be without them … It may cause some long term damage - or people may understand (or be helped to understand) that the current situation is truly exceptional and stuff just needs to get done - and if they understand this they will hopefully either stick around or come back once the current crisis is over.

Neil-B wrote:At this point in time tbh I'd prefer the science completed by the fastest route possible …

I'll say this makes sense. A lot of people have come into this project very recently with the mindset of a global pandemic so speed seems like the only important thing here. But the problem is what happens when one entity begins to control a majority of said science's progress. The ends may be reached but aren't fully justified.

The "one entity" is actually quite a few large institutional/organisational/corporate/individual contributors each contributing at scale that prior to COVID-19 would have been extremely unusual - It isn't so much that these contributors are controlling the science's progress rather that they are giving it a hell of a push at the moment when it most needs it.

Some of these large contributors may hopefully find ways to contribute (maybe at a lesser scale as their kit returns to its usual functions) "post covid-19", just as I hope many of the individual donors who have joined over the last months "to fight covid-19" choose to continue to fold … and I understand concerns about the impact of temporary contributors … but only time will tell.

I honestly believe that accepting all help to progress the science as fast as possible is not the wrong thing to do, and won't in the long term significantly damage FAH especially as the risks of doing this will no doubt be being carefully considered/monitored by the FAH Consortium and so I reckon that in this case the ends do justify the means.

FallibleCanine wrote:Do we still have an excess of users, relative to WUs? Would more users right now help?

Nobody every likes to admit that "we don't want your help right now" but realistically, there is a lot of underutilized hardware out there; which is another way of saying we have an excess of users relative to WUs. There seem to be plenty of CPU assignments but a lot of users classify them as "not worth the trouble" because of their relatively low PPD rewards. I look at the science they're doing and I think that the overall science would benefit if we had more people doing that class of work. Thus we could use more "small" donors ... we've got plenty of "big" corporate donors who are gobbling up the GPU workload. I realize this contradicts the global answer that I gave initially.

Like somebody said earlier, I have no employment contract with the FAH Consortium. I'm just a volunteer who has been here a long, long time.

bruce wrote:

FallibleCanine wrote:Do we still have an excess of users, relative to WUs? Would more users right now help?

Nobody every likes to admit that "we don't want your help right now" but realistically, there is a lot of underutilized hardware out there; which is another way of saying we have an excess of users relative to WUs. There seem to be plenty of CPU assignments but a lot of users classify them as "not worth the trouble" because of their relatively low PPD rewards. I look at the science they're doing and I think that the overall science would benefit if we had more people doing that class of work. Thus we could use more "small" donors ... we've got plenty of "big" corporate donors who are gobbling up the GPU workload. I realize this contradicts the global answer that I gave initially.

Like somebody said earlier, I have no employment contract with the FAH Consortium. I'm just a volunteer who has been here a long, long time.

After being told in multiple threads, in multiple forums (including here), by multiple people that F@H CPU work is not important science and the low PPD backs that up, I’ve moved the last of my CPUs over to Rosetta@Home. If it is in fact important work, can someone official maybe state that? Ideally this statement would come with a revisit to CPU PPD, but I get that messing with points right now is probably low on the priority list.

JimboPalmer wrote:Until the servers can keep up with the current volunteers, I feel adding more frustrated volunteers is misguided.

As I said in my previous post, I see really important science being done by the CPU client.

JimboPalmer wrote:Until the servers can keep up with the current volunteers, I feel adding more frustrated volunteers is misguided.

Let's consider Endgame's post as agreeing with JimboPalmer. I fully understand the frustration.

I wonder if there's a way to bias the Assignment logic so that the "little guy" running a CPU slot gets an assignment even if it adds a slight delay to one of the corporate GPUs.

Seems the simply answer is to increase the points for cpu work