Posts by cristipurdel

1) Message boards : Number crunching : Rosetta and Android (Message 77733) Posted 14 Dec 2014 by cristipurdel Post: After reading the previous threads, it is still not clear to me how much RAM the current app is requesting. Now my Galaxy 1 is saying that it needs 488.28 MB RAM and 808.70 MB disk space. Are these numbers accurate or is just my phone requesting too much? When I am checking http://wuprop.boinc-af.org/results/ram.py?plateforme=all&tri=2&sort=desc it says that Rosetta Mini should require 62.4 MB RAM and 338.1 MB disk space. Why is the discrepancy so high?
2) Message boards : Number crunching : Rosetta and Android (Message 77568) Posted 9 Oct 2014 by cristipurdel Post: so again don't blame android or the android version of boinc. I don't blame use of android version of boinc in general. For some protocols (some projects) is good enough, but not for rosetta. Android version has great potential, but up to now is only marginal for computational power. I agree, android is not yet up-to-par when running Rosetta. There's a 11x performance loss when running the standard abinitio-relax folding protocol on a 96 residue protein sequence, compared to our in house machines. The average float and int ops BOINC cpu benchmark values of active hosts compared to Windows is around 2.4x and 1.7x less, respectively. I'm guessing the rosetta performance loss is mainly from memory bandwidth but we'll have to profile to see what is so slow. But the true metric, IMO should be not just CPU throughput but CPU throughput/energy consumption. I don't really have an idea of what that value might be. Glad to found an answer as to why some WUs are failing on my androids. I agree that the performance/power is the true metric to watch out for. I have around 6-7 androids running boinc, and probably I will add 4-5 more, since I find the crunching during charging... priceless ...