Zagen30 wrote:7im wrote:
RD, Average core count drops the last few months as people transition from core 15 and 16 which don't need a dedicated core to core 17 which does need a dedicated core. I can't prove that, but seems plausible.
That doesn't explain the vast majority of client or core loss.
You are correct. But clearly, I was not addressing you with that statement, nor was I adressing the question about the large drop in the number of active clients.
Zagen30 wrote:The project's recent high point over the summer looks like this: it went from 130k CPUs/256k cores on March 5, when core 17 was released, to a high of 315k CPUs on June 21, a high of 535k cores on July 13, and a cores/CPU high of 2.008 on August 6. We're talking double the cores and more than doubling the number of active CPUs. The number of core 17-capable GPUs (AMD + Nvidia Fermi) during this time actually dropped from 20k on June 21 to 16k on August 6. Core 17 GPUs did experience a surge in the next couple of months, hitting a high of 24k total on October 4. The number of Windows CPUs+cores was off a bit from the summer height, in the range of 250k CPUs and 480k cores. I realize these are somewhat lagging indicators, but that surge was in large part during the summer, as new clients show up in the stats immediately.
Since then, the project has lost around 90k clients and around 190k cores. In that time, the number of core 17-capable GPUs fell to a low of 15.5k on December 11, though it has surged back up to over 20k in the last week or so. Now, some of these GPUs may have been getting core 17 WUs for the first time when core 17 was moved to full FAH on November 7, and thus leading people to remove a core, but it's nowhere near enough to explain such a massive loss of CPU power. Even if you assumed that every single one of those GPUs had just dropped a core (ignoring that many had already done that in the past several months), that's only around 20k cores, barely over 10% of the total number of cores lost.
On[e] thing that might have contributed to the drop in Windows machines is the ending of core 78. I don't know how many set-and-forget machines there were out there running 6.23 or older that couldn't get any more uniprocessor work. These would have stopped showing up in the stats starting October 15 (if they turned in work right after the servers were taken down on August 26). There was a loss of around 50k machines and 95k cores between October 15 and December 15; how much of this was uniprocessor-related and how much was donor dissatisfaction is kind of hard to say from our end. 95k sounds like a lot, but maybe PG has data to confirm it. This also doesn't seem to account for the other 100k cores lost since mid-December to the present, as I'd think by 60 days after core 78 machines first stopped appearing in the stats, pretty much all the uniprocessor-only clients would have been passed out of the active client count.
There must be others like me who have given up SMP folding because the GPU points are just so much better. When I got my first 780, I still kept my 3770k going, but once I added a second one, the 15-20k PPD it was producing was peanuts compared to the 170-180k for each GPU (not even factoring in my 4P bigadv box at 900k). I know I'm well above normal, but I've seen some new people being advised to skip SMP folding if they have a powerful card or two, and some must have reached that conclusion on their own.
Ah, finally, someone who can start to see a picture larger than the one just past his nose.
I mentioned core_78 in this thread already. Please note that although they assigned the last core_78 WU at the end of August, you have to take in to account that core_78 projects have a typical deadline of 1-2 weeks that you would need to add to the 50 day tick on the client count. But some larger 78s have deadlines as long as 3 months. So there would be an initial drop that's large in the 7-9 weeks after the end of August, but there could have been some still folding as late as December 2013, and then dropped off, if there was a server still around to collect it after that long.
This is an example of what I mean by "bigger picture."
Et al, anyone claiming to know why the client count dropped without a 4 page explanation of all the variables just doesn't have a leg to stand on. Could be people don't like V7. Could be the crappy economy. Could be everyone is moving to mobile devices. Could be an alien invasion. Could be GPUs. Could be an AMD stats glitch. Could be the increasing cost of electricity. Could be Global warming, so they turned off their FAH heaters. Could be Bitcoins. Could be all of that and MORE!
Correlation is not Causation. Do NOT assume that listing a client count drop and then pointing a finger at the need for more transparency has anything do with anything. It doesn't, and you can't prove it, so stop doing that it already. It just makes one look inept, if not worse.
Don't get me wrong. I am not saying PG is perfect. Far from it!!! Lot's of room for improvement. But if you want something to improve, don't say the sky is falling as your justification for more transparency, or whatever. It's not a winning strategy in any game.
Instead, state your case in a more logical and less emotional manner, if possible. Ask for improvements, demands won't work. Leave out the absolutes, because they are guaranteed to be absolutely wrong. You're welcomed to say what the change will do for you, or what you will do as a result, and suggest what it might do for others, but don't assume to know that for everyone. You don't, and neither do I. And if you think that PG doesn't know the consequences of making a change that they've already made twice before, you would be wrong again.
Fight the change, or embrace it, but do it smartly. Baseless complaints and ultimatums don't work.
