So, what's our target this year? Top 5 would be great.
Considering we were 8th last year, then 10th this year would already be a big improvement. This is because in contrast to 2017, this time three "financial services" ventures are registered as "teams" in this challenge. (Byteball.org with 14 M boinc-PPD daily average, BiblePay with 6.6 M, Gridcoin with 6.4 M. I took these averages from
Free-DC. Note, IBM's average is 8 M boinc-PPD. Multiply with 7 to arrive at WCG-PPD. The challenge stats are counted in WCG points.)
Of the 7 teams which finished in front of us in 2017, the following ones are now (14 hours after the start) still behind us: HardOCP, Planet 3DNow!, TechPowerUp!, L'Alliance Francophone.
I do wonder whether or not SETI.USA made goals for themselves in this year's Birthday challenge, unlike 2017. They were the only real team who succeeded to get into the Thor challenge 2018 final round.
And to name just one more strong team of those who are still behind us, there is Decrypton. They did not participate in the Birthday challenge 2017, but they much outpaced us in a month long
12/2017 challenge.
Erm,
@Markfw I'm interested in result comparison between your 1950X and 2990WX, do you have any?
Well, I did some quick checking. One of my 1950x's is doing 4:40 cpu time on mapping cancer markers. The 2990wx is doing 3:80.
When I measured MCM performance on my own hosts in November 2017, I found considerable variations between tasks on one and the same host:
- Run times varied between 3 h and 5 h.
- Granted credit per hour run time varied between 20 and 50 BOINC credits/hour.
- Samples with a size of more than 50 tasks still had coefficients of variation in excess of 10 %, for run time and for credits per run time, respectively.
If this variation is still going on in current MCM batches, then an average of points-per-day from a considerable number of tasks needs to be taken, before comparisons between different hardware can be made.
(Furthermore, to compare different hardware, use the same OS. Some WCG subprojects have a considerable performance difference between their Linux and Windows applications. I suspect MCM doesn't differ a lot per OS, but I haven't actually checked.)