Message boards :
News :
Should the fixed credits be higher?
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
Send message Joined: 13 Apr 19 Posts: 26 Credit: 11,397,920 RAC: 17 |
So I have decided to double the credits. This will get it closer to the PrimeGrid credits and I think this will make most people happier. If necessary I can always put it back later. Yes the credit change has been implemented |
Send message Joined: 6 Jul 20 Posts: 1 Credit: 23,936 RAC: 0 |
I Agree too... |
Send message Joined: 31 Mar 20 Posts: 1 Credit: 4,311 RAC: 0 |
I seem to have a much higher number for numberfields than other BOINC projects,or Science United projects[/img] |
Send message Joined: 8 Jul 11 Posts: 1342 Credit: 514,709,901 RAC: 579,646 |
I seem to have a much higher number for numberfields than other BOINC projects,or Science United projects[/img] I don't see that you've completed any tasks recently. Are you talking about an account under a different username? Either way, credits may be higher than some projects, but should be less than PrimeGrid. That's the happy medium I tried to find. |
Send message Joined: 25 Oct 18 Posts: 15 Credit: 112,744,248 RAC: 7,764 |
" Either way, credits may be higher than some projects, but should be less than PrimeGrid. That's the happy medium I tried to find." Still comfortably less than PrimeGrid Eric e.g. for an nvidia gtx1660ti on Linux ... ... PrimeGrid pps-sieve around 1.3M credits per day ... NumberFields GetDecic around 46.5K credits per day That is 28 times higher and, rightly or wrongly, that REALLY does matter to a great many people when they are deciding which project to run. PS: I would be happy to provide you with timings to back up these figures |
Send message Joined: 8 Jul 11 Posts: 1342 Credit: 514,709,901 RAC: 579,646 |
" Either way, credits may be higher than some projects, but should be less than PrimeGrid. That's the happy medium I tried to find." I was referring to the CPU. As you point out, you can't even compare the GPU credits. The Numberfields GPU app is about 10x faster than the CPU app (depending on your CPU/GPU combo). A highly efficient GPU app could probably raise that to about 100x faster. That would get us much closer to the PrimeGrid GPU credits. I don't want to artificially increase the credits for the GPU just to compensate for an inefficient app - it feels like cheating. I will attempt to make the app more efficient when I find some time. |
Send message Joined: 25 Oct 18 Posts: 15 Credit: 112,744,248 RAC: 7,764 |
Thanks Eric, I look forward to that new GPU app as I really did enjoy running NumberFields again during the Pentathlon. Yours has always been a very well-managed project in my experience |
Send message Joined: 19 Aug 11 Posts: 1 Credit: 9,624,913 RAC: 0 |
I finally hit 25M at Universe and switched back to NumberFields. After running NumberFields the last two days, I can attest the credits should be higher here if you want more long-term crunchers. That may or may not matter to the administrators here, but even running three decent Nvidia GPUs along with my 3 CPUs, I'm not getting close to what Universe alone was granting per day, and that was on CPU only. |
Send message Joined: 27 Dec 19 Posts: 6 Credit: 623,103 RAC: 0 |
You should add redirect from https://numberfields.asu.edu/ to https://numberfields.asu.edu/NumberFields https://numberfields.asu.edu/ only displays "It Works" |
Send message Joined: 5 Jan 13 Posts: 43 Credit: 41,980,418 RAC: 56,628 |
It is interesting, when we back to Search 3? Only two searches remained there. |
Send message Joined: 8 Jul 11 Posts: 1342 Credit: 514,709,901 RAC: 579,646 |
It is interesting, when we back to Search 3? The plan is to close out column 10 in the sf7 search first. I updated this page to reflect this: https://numberfields.asu.edu/NumberFields/batch_status_2dim.html The final search for sf3 has a prohibitively large number of WUs (estimated to be ~20 million). I was trying to put this search off as long as possible to give me time to improve the GPU app. As it turns out, I have been working on this in my spare time over the last week. It is slow going - so far I only have about a 5% speed improvement. |
Send message Joined: 13 Mar 19 Posts: 10 Credit: 34,380,826 RAC: 25,503 |
5% are better than nothing :) |
Send message Joined: 13 Apr 19 Posts: 26 Credit: 11,397,920 RAC: 17 |
5% are better than nothing :) My thoughts exactly, great work Eric. I will be back to help as soon as I have reached a goal that requires all of my CPUs, will be late 2020 early 2021 :) |
Send message Joined: 8 Jul 11 Posts: 1342 Credit: 514,709,901 RAC: 579,646 |
5% are better than nothing :) I estimate it will take almost a year to complete sf3 at the current rate. To get a meaningful speedup I need much more than 5%. |
Send message Joined: 13 Mar 19 Posts: 10 Credit: 34,380,826 RAC: 25,503 |
Is it possible to make the code public available? In other projects there were some user who made optimized apps for specific cpus which are significant faster For example Rakesearch: https://rake.boincfast.ru/rakesearch/forum_thread.php?id=39 Maybe some of them could help here too |
Send message Joined: 8 Jul 11 Posts: 1342 Credit: 514,709,901 RAC: 579,646 |
Is it possible to make the code public available? So a fellow cruncher named Khalim helped me by uploading the code to github, but it is currently private. The plan was to make this public at some point. Once I finish my latest "improvements" and the code is in a better state, I will make it public. Khalim also helped with some testing scripts that will allow users to fine tune parameters for their specific GPU, and then these parameters can be put in a lookup table to benefit users with the same GPU. |
Send message Joined: 13 Mar 19 Posts: 10 Credit: 34,380,826 RAC: 25,503 |
Is it possible to make the code public available? Oh thats a very good news :) |