I regularly read questions from Gridcoin users who would like to know how much they can earn with their system and which BOINC project would be the best option.
Earnings are derived from the RAC output of BOINC projects and are mainly determined by the type of CPU (or GPU but this article will focus on CPUs only). Because there are so many different CPU types in the market, it would be impossible to build an overview with actual RAC output for all of them, even if you would have access to all hardware and software.
An alternative method is to calculate the RAC based on returned project work-units. Even though it doesn’t provide a 100% accurate RAC value for all systems, which is impossible due to a variety of reasons, it would still provide a reasonable number that can be used as a baseline to calculate the earnings and select a project.
Last summer, I started capturing WU results on various project websites to calculate the RAC output per CPU. To ensure I wouldn’t overload any project server, I only sent a (page view) request every 10 seconds (or maximum 8500 requests / 60Mb per day) and have build-up sample set which I’m now using to calculate the RAC for a wide range of CPUs.
Calculated RAC
I used the following method to calculate the system RAC:
The formula is built with the following elements:
- Σ points of sample set, is the amount of points credited by the project for the returned results (column 4 in tables below)
- Σ time of sample set, is the total time it took to process all returned WUs (column 3 in tables below)
- Σ CPU threads, is the number of CPU course x CPU threads. I assumed in my calculations that multithreading is enabled in the BIOS (column 5 in tables below)
Two other important factors:
- The higher the number of returned WUs, the more accurate the RAC value (column 2 in tables below)
- The calculated RAC is based on running SRBase as only project on a sysetm, with 100% CPU utilisation and 24/7
Calculated RAC for CPUs crunching SRBase
In the coming weeks I will publish a series of articles for most whitelisted projects with an overview of CPUs and their calculated RAC.
Today I will start with SRBase, for which I have gathered 1,700,000 WUs resulting in 2 tables with 620 CPU types in total. There are even more CPU types in the market but I have omitted any for which I didn’t capture results.
Next to the calculated RAC, I have included a column with GRC earnings as well but because this is changing regularly, you can always take the latest value from my weekly project stats update on Steemit (last version here).
I have put every effort in providing accurate RAC values but the output is as good as the results taken from the SRBase website.
Looking forward to your feedback and suggestions where I’m open to make any modifications to the calculation presented above.
And here are the tables:
Phenomenal work, thank you! This must have taken a lot of work. I've been thinking about trying to relate the performances of different kinds of processors on different projects and this is exactly the kind of data I need. I'm sure many others will find it useful as well.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Thanks @ilikechocolate, I hope indeed that this article can help people to determine how much RAC ouput their CPU can produce.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
E5-2650L v3 has 12 cores/24 threads.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
I have updated the table, thanks for the feedback @ilikechocolate.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Such a big table!
Xeon E5645 has 6 cores / 12 threads, not 4.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Thanks @hotbit, I have updated the table. :)
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
very interesting approach! far more accurate that the method described a few days ago by cautilus
i have a working prototype of a mining calculator based on that method. basing it on this one would be preferable.
I just think a joint effort in collecting the data and/or making it usable would be nice!
Are you planing a publicly accessible database/API for this information?
or github repo for the code or the data?
I could offer some hosting/database usage and some coding time (mainly js, node.js, html, css, sql and elasticsearch etc.) ...
one additional feature i have build in in my calculator: I use a benchmark database with roughly 3000 CPUs. If i want to calculate an estimated reward for a CPU i don't have boinc data for, i calculate the rac per benchmark point for every CPU i have boinc data for (for a given project) and then use an average of that data to calculate the estimated rac of that CPU based on the benchmark points.
anyways, thanks for this great work!
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Thanks @holger-will, I will make the WUs available on github shortly and share it with the community. Concerning the code to fetch the data, it is based on Autoit and I think it is too crude. I have other ideas how to automate the process, which I will share with everybody in the near future.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
would it be usefull to have a dedicated screen scraper for exactly this type of data? My code is based on node.js and it could easily be extended to support the data you used, and can be published via npm for example...
Looking forward to your ideas!
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
My RAC was considerably lower on Ryzen. That's odd.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Thanks @ragnrokdel, can you share which Ryzen CPU you are using and RAC you're achieving? Besides is SRBase the only project you're running 100% of the time (no heavy game usage or other CPU intensive apps)?
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
This raises one question... on http://srbase.my-firewall.org/sr5/top_hosts.php the top host is a Ryzen 7 1800X with a rac of 384,196.47 ... how is this even possible?
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
The calculated RAC for this host is 733,599 but it will never achieve this because of the results it is producing:
Another intersting element is that it is using Darwin as compared to the rest of the of the top 100 using Win/Linux.
I won't judge here but it would be interesting for the SRBase project owner to investigate this.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
based on your table the calculated RAC for that type of processor is 49734...
If you look at the WUs of this host, there is something very fishy about it... the CPU times are waaaaaayyyy to short compared to everyone else on the list. And the host is on top and has no active WUs
the Wu versions (v0.21) seem strange, there is no v0.21 all other host are crunching on v0.20 !?
ohhhh... and it acquired all it's points on four days!!!!! thats for 127156 WUs only for this host...
this would be interesting for the gridcoin pool as well as the host is apparently mining for the pool and "stealing" MAG from everyone!
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
all that said, this is either a bug in SRBase or the SRBase developers making themself some GRC ;-)
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
ok... the assumption that there is no v0.21 was wrong... there is! sorry!
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit