1) Message boards : Number crunching : Credit decrease. (Message 4782)
Posted 26 Aug 2021 by Christopher
Post:
That's too bad. I remember back in the day (circa 2016-ish) when 667 points was awarded for about 40 min computing on an i7-4770. Now a BH spin lasts over 1.5-3 hrs for the same amount of points. I am considering abandoning the project. You shouldn't have to constantly upgrade your hardware just to get the same ppd. The idea is that as you upgrade your hardware, your point accumulation escalates and you can progress through the logarithmic scaled badges. Right now there is just a regression in progress for old backers of the project. There's no psychological motivator to stay engaged if you just flat-line.
2) Message boards : Number crunching : Some task run longer than usual (Message 3751)
Posted 29 Aug 2019 by Christopher
Post:
You know, quite frankly I would agree to a certain compromise considering that a bit of deflation might be necessary when compared to certain projects on BOINC. I would propose increasing the points per WU to an even 1000 or 1500. Nice, round, easy, an increase, but still some deflation gets accomplished. idk just my two cents.
3) Message boards : Number crunching : Some task run longer than usual (Message 3744)
Posted 26 Aug 2019 by Christopher
Post:
Everyone here has made some good points, Krzysztof (pun intended). The reality is that the silicon keeps getting faster. It always was and it always will. New tech is gonna grind out points faster. It will always appear that you are "giving out too many points" unless you constantly deflate.

However, deflation has it's risk. Competitive crunchers will move to projects that offer better return. People who rely on gridcoin to offset their power costs will also move on (unless practically everyone leaves, then the magnitude will go back up). And then, there's the issue of new users who won't be able to "catch up" and get a nice badge unless they use cutting edge hardware.

Yes, ultimately, it's all about the science. But credits help people see at a glace how much compute was accomplished (and, yes, that means that new and expensive machines win the arms race adjusting for compute time). If rules keep changing then the credit system won't reflect anything anymore.

I would propose that credit remains tied to the amount of work accomplished. If you would like to reward senior users who have crunched out work over longer periods of time on slower machines, you could consider a revamped badge system (ex. badges for years working on the project). In the worst case scenario, if you really want to deflate credit per amount of work, then I would suggest doing that when new projects come out (ex. ultraviolet or whatever).

I do agree with you; however, that work units will need to be enlarged as the silicon gets more powerful. Logistically and computationally it is more efficient. But I don't agree with credit deflation at least when the end user isn't really upgrading their silicon. They just feel penalized that they aren't on 7nm.

Just my two cents. As for me personally, I will be diversifying my compute operation. Not just because of this, but because of the reality of projects going dark after some time (ex. denis or enigma). I hope that you can get enough people to keep crunching this in order to keep the papers flowing to the publishers and keep the enthusiasm alive.
4) Message boards : Number crunching : Some task run longer than usual (Message 3730)
Posted 25 Aug 2019 by Christopher
Post:
For now I will switch work to another project until this gets settled I guess...
5) Message boards : Number crunching : Some task run longer than usual (Message 3729)
Posted 25 Aug 2019 by Christopher
Post:
Is a credit increase planned? Proportionally, a WU would be worth just shy of 2000 right now...

Of course, this is based on compute time not number of operations. Perhaps the workload isn't optimized and is leaving some parts of the silicon idle during some cycles depending on architecture. Did anyone notice any difference in number of gigaflops reported in the properties section of the WUs?
6) Message boards : Number crunching : No WUs right now? (Message 3643)
Posted 2 Aug 2019 by Christopher
Post:
Yeah, I haven't gotten WUs for a few days now...
7) Message boards : Number crunching : No WUs right now? (Message 3617)
Posted 30 Jul 2019 by Christopher
Post:
Is it just me or are the WU now getting longer? I got a few today BTW...
8) Message boards : Number crunching : No WUs right now? (Message 3612)
Posted 29 Jul 2019 by Christopher
Post:
Yeah something is wrong. I am getting work very sporadically. I really would like to know what could be the reason.
9) Message boards : Number crunching : No WUs right now? (Message 3605)
Posted 26 Jul 2019 by Christopher
Post:
Actually, I am still having issues with getting WUs today...
10) Message boards : Number crunching : No WUs right now? (Message 3597)
Posted 25 Jul 2019 by Christopher
Post:
Never mind. I just got some WU. For some reason the last few days have been hit or miss when it comes to getting WUs on time before the previous ones complete. idk if there is some performance impact or slowdown affecting the main WU distribution servers...
11) Message boards : Number crunching : No WUs right now? (Message 3596)
Posted 25 Jul 2019 by Christopher
Post:
Hi guys,

I am just wondering if anyone is getting any new WU. I haven't gotten any over the last day (last 2 days had less and less of tasks in queue) while BOINC stats says that WUs are available. Using i7-4790 only. Manual update of project doesn't work. Nothing really changed on my end. Is there anything going on on the project's end currently?

Thanks.
12) Message boards : Number crunching : cheating? (Message 3329)
Posted 7 Feb 2019 by Christopher
Post:
Magnitude is calculated from the percentage of the total pie


Right. And the new credit system was giving variable points per WU (so some people would get disproportionately lower returns on average based on a wide range of variables). This (potentially significant) group of people would then end up with less of the "pie." Hence, this would give them less mag and could make them potentially leave, leading to less total compute power being devoted to the project (at least for the short term until a smaller "pie" would entice those whose computers and unique pan-project timing luck can get a better return on WUs). Either way, the risk of loss from so much flux isn't worth the reward especially mid-project.
13) Message boards : Number crunching : cheating? (Message 3298)
Posted 5 Feb 2019 by Christopher
Post:
I agree. Ultimately points should reflect the amount of work accomplished. A certain amount of calculations (can be stratified according to FP16, FP32, FP64, or INT) should yield a set amount of points. Even if projects utilize multiple participants computing the same WU: checking or re-doing someone else's work is still work and still uses the same amount of power (adjusting for architecture and performance of course). Equal pay for equal work. :)
14) Message boards : News : Problems sorted :) (Message 3297)
Posted 5 Feb 2019 by Christopher
Post:
Thank you for getting this sorted out. Now back to crunching... ;)

Also, regarding the android app: I am using pie (9.0) and already have BOINC all set up. Looking forward to it. 10 nm :D Even though BOINC only uses 2 little cores at most from background standby resume (unless you kill the app and start it again), it's still something. Perhaps, there is a way the application itself could exit and restart on BOINC resume so it uses big cores?
15) Message boards : Number crunching : cheating? (Message 3296)
Posted 5 Feb 2019 by Christopher
Post:
That's fine. As long as we get everything fixed with validation and move back to the 666 points per WU going forward, we can all just write off this incident as some temporary technical difficulty. Just please let us know when all the back-end stuff is done so we know when to hit "resume." Update/edit: I just saw your announcement. Thanks man.
16) Message boards : Number crunching : cheating? (Message 3290)
Posted 5 Feb 2019 by Christopher
Post:
I highly recommend switching back to the old credit system once any validation issues are fixed. A lot of volunteers will quickly switch to other projects that give more points per compute time (especially those running gridcoin...they won't even bother checking the forum but just switch to whatever gives them more magnitude), and that will massively impact the project. Ultimately, getting the science done and publishing (and, trust me, Poland NEEDS a boost in high impact journals) is much more important than getting credits in line with the rest of BOINC. The whole purpose of the credit system is to give an incentive for others to participate. Right now I have switched over my computing to SETI because I can get the same amount of points and run it faster on a GPU. As both a UC Berkeley grad and a student at a Polish university, the credit drop is making it harder to be patriotic right now.

If you want to do a credit system overhaul, I would suggest implementing changes after BHspin is complete and for another project or application (with appropriate early announcements). Pulling the rug from under your users overnight will just piss people off and put this whole project in jeopardy and massively delay the schedule, which I assume has already been planned out.

And, please, let us know when the credit system is back in a news post or something. I really would like to switch back and support Polish research once all of this blows over.







Copyright © 2022 Copernicus Astronomical Centre of the Polish Academy of Sciences
Project server and website managed by Krzysztof 'krzyszp' Piszczek