ODLK1 - Calculated RAC for CPUs
In the past months I noticed regular requests from Gridcoin team members on Steemit, looking for the RAC output of their systems to determine which Gridcoin project to crunch.
This gave me the idea to gather information for all whitelisted BOINC projects and calculate per CPU the RAC (Recent Average Credit) and potential earnings.
ODLK1
After a few specific requests, I’m presenting in this episode an overview of 1,000+ CPUs with the calculated RAC and daily earnings for ODLK1. Similar to previous episodes, I have only included those CPUs for which I have gathered sufficient information based on sample set of 12,000+ hosts and 2,000,000+ work-units.
After reviewing 15 whitelisted projects, ODLK1 is currently on position 10 in terms of average earnings and as such a below average earning project (see table below). Furthermore, it has the lowest average RAC score of all projects on the list.
Info on Github
Due to the large number of CPUs and projects, I have stored the output separately on Github. The main file can be downloaded in Excel format or OpenDocument format.
The file contains the following worksheets:
- Worksheet 1: Average daily earnings per project
- Worksheet 2: Daily Gridcoin earnings per CPU for all projects
- Worksheet 3 to 17: CPU calculated RAC and GRC earnings per project
- Worksheet 18: Threads per CPU
- Worksheet 19: GRC per 1k RAC for each project
I will continue to update the file with more projects and refresh the earnings so it is best to download the latest version from Github.
Average daily earnings per project
In case you don’t know your CPU, you will likely achieve the maximum earnings for your system by selecting one of the first projects in table on the first worksheet (see also table above). Don’t confuse the numbers in the table with actual earnings, it is merely an index to compare the earnings potential between projects. Obviously, the project ranking will change when computing power is shifting across projects so I will continue to update the table to reflect this.
The ranking of projects is established in the following way:
- Select those CPUs which have results in all projects (this applies to 312 CPU types at the moment)
- With the selected CPU types calculate the average earnings for each project
- Sort the projects from highest to lowest average daily earnings
Calculated RAC
I used the following formula to calculate the system RAC:
Breakdown of the formula:
- Σ points of sample set, is the amount of points credited by the project for the returned results
- Σ time of sample set, is the total time it took to process all returned WUs
- Σ CPU threads, is the number of CPU cores x CPU threads.
Three assumptions as part of these calculations:
- Multithreading is enabled in the BIOS
- The calculated RAC is based on running one BOINC project, 24/7 on all cores
- The higher the number of gathered WUs per CPU, the more accurate the RAC value
Previous episodes
The table below presents a summary of all projects introduced so far including high-level details per project.
| Project | CPUs | Hosts | WUs | Information |
|---|---|---|---|---|
| Asteroids@home | 1,303 | 19,462 | 2,885,452 | Raw data p1 (Size: 21 MB), Raw data p2 (Size: 17 MB) |
| Citizen Science Grid | 1,118 | 14,368 | 1,657,693 | Raw data (Size: 24 Mb) |
| Cosmology@Home | 1,092 | 17,758 | 2,124,977 | Raw data p1 (Size: 16 MB), Raw data p2 (Size: 12 MB) |
| Enigma@home | 746 | 3,775 | 2,143,679 | Raw data (Size: 17 Mb) |
| LHC@Home | 1,279 | 17,727 | 2,189,428 | Raw data p1 (Size: 14 MB), Raw data p2 (Size: 15 MB) |
| NFS@home | 1,016 | 56,273 | 1,666,526 | Raw data (Size: 19 Mb) |
| Numberfields@home | 1,009 | 20,885 | 1,542,956 | Raw data (Size: 23 Mb) |
| ODLK1 | 1,010 | 12,340 | 2,098,300 | Raw data (Size: 21 Mb) |
| PrimeGrid | 931 | 9609 | 2,030,902 | Raw data p1 (Size: 20 MB), Raw data p2 (Size: 25 MB) |
| Rosetta@home | 1,585 | 57,240 | 2,042,631 | Raw data p1 (Size: 18 MB), Raw data p2 (Size: 17 MB), Raw data p3 (Size: 19 MB) |
| SETI@home | 1,747 | 122,277 | 2,213,442 | Raw data p1 (Size: 20 MB), Raw data p2 (Size: 24 MB), Raw data p3 (Size: 19 MB) |
| SRBase | 640 | 6,997 | 1,732,240 | Raw data (Size: 15 Mb) |
| theSkyNet POGS | 1,155 | 13,046 | 2,285,013 | Raw data p1 (Size: 8 MB), Raw data p2 (Size: 19 MB) |
| TN-Grid | 905 | 9,468 | 2,238,497 | Raw data (Size: 21 Mb) |
| Yoyo@home | 953 | 6,617 | 1,704,412 | Raw data (Size: 17 Mb) |
Thanks for reading. Follow me for more BOINC & Gridcoin related articles.
Image source: Header, Gridcoin logo, BOINC logoData source: BOINC project websites
Upon closer inspection, there is an obvious pattern in the highest magnitude units: I found that TNGrid and SRBase do not have available work units all the time. This means that not all the crunchers can get work, so the payout is split among fewer people, who then get higher rewards.
This is an interesting feature of how Gridcoin works and, if your goal is to maximise magnitude/earnings, should be taken into account when selecting projects to work on.
On the other hand, a strong case can also be made about joining the projects you personally believe in, independently of small differences in payout....
I think your observation is valid and this Excel file on Github can help anybody trying to optimise the output of their system.
Personally I like astro projects and I don't have a preference for a specific project so based on the table above I'm now crunching Cosmology@home on my laptop. I'm actually applying a mix between maximising the earnings but still crunching the projects I like.
Great post! Thanks for analysing ODLK. I somehow thought that it was offering good magnitude. I know realise it doesn't, so I am tempted to move to TN-Grid or SRBase. Unless everyone else thinks the same as well, in which case I should stick with ODLK and benefit from the reduced competition. Prisoner's dilemma!
Thanks @p-a. I think people have the impression that due the low team RAC, ODLK1 provides a good return but this is not the case because it has also a low credit distribution.
Anyway good luck with contemplating your next move! :)
Again all these numbers are fluid and meaningless , its valid for the moment in time you calculated but since the number of crunchers along with hosts machines it changes before the days end so these calculations are all invalid for every project people waste their time to create them.
I'm updating the project tables on weekly basis using the earnings for a system having a RAC of 1000 (GRC per 1k system RAC). I'm applying the same method for system earnings as in my weekly updates and there you can clearly see these fluctuations.
So the snapshot in this article might be static but the Excel file on Github remains up-to-date. This is why I mentioned above to always download the latest Excel from Github.
I suggest that you check this file and let me know your views.
You just planted 0.42 tree(s)!
Thanks to @parejan
We have planted already 7146.18 trees
out of 1,000,000
Let's save and restore Abongphen Highland Forest
in Cameroonian village Kedjom-Keku!
Plant trees with @treeplanter and get paid for it!
My Steem Power = 22158.13
Thanks a lot!
@martin.mikes coordinator of @kedjom-keku