[spunky]

Hi, I just checked boinc/seti@home site to check on server status cause i have been having upload problems, then I find in the news section that they have released a newer version of boinc 4.13 with upload bug fixes.. Has anyone here been having problems with uploading using 4.11, and anybody know that this is a bug, and if it is, when will there be a newer version for us TeamMacNN'ers....

[Drash]

I was using the optimized 4.11 and was getting units rejected, something like bad name - can't quite remember the reason. Anyhow, I installed 4.13 and so far haven't had a unit rejected. I also found that the benchmarks weren't that much less than with the optimized 4.11. I'm still using the Altivec'd SETI 4.30 client though. If SETI does the actual work, does it matter whether BOINC gives good or bad benchmarks? SETI BOINC is a pile of cr*p compared with Classic seti, IMHO - there is so much stuff unexplained, two programs to do one job, fixes coming out like MS security updates etc. etc.

Ash

[davygravy]

Hi Drash,

I can understand the general frustration w/ Boinc.
*it is new, and CLI based instead of "double-click"-based
*it has a particular, picky directory (folder & file) structure
*it doesn't yet have all of the bells & whistles that the screensaver-based S@H Classic had,
*it seems less supported... (well, the support is there, but it's of a different nature)

Well, it does have advantages, (stuff that couldn't be said for S@H Classic)
*it is more flexible, allowing us to work on different projects,
*it is uptodate from an architectural standpoint (database, xml, etc.)
*it can be optimized for given platforms, like mikkyo & javalizard have done here,
*it is less prone to tampering w/ data (falsification of results) and is secure.
*though it does have multiple components, we can now cache WU's (the old system needed a 3rd party program to cache WU's)

I think it is an improvement, though still in development stage. A project of this scope and with the flexibility/portability is quite an undertaking. I think they are doing a pretty good job. Problems, yes, but pretty darn good.

davygravy

[spunky]

Hey I solved the problem, I just downloaded the newer one from boinc/seti website and replaced the 4.11 version 4.13, and it still is pretty fast usinf setiathome 4.30... I will Replace it later tho once the guys from TeamMacNN makes a newer one... cheerious!...

<img sac="http://www.hickorytech.net/~spunky/images/GoToHell.jpg">

[mikkyo]

The boinc_4.13 optimized compiles are up on the beta page if anyone wants to try them out.
There are 2 versions, one requires XCode1.5 to be installed(free from Apple Developer - just make an account).
The gcc3.5 compiles are up to double the whetstone speed of the regular gcc3.3 compiles.
The gcc3.5 compiles will also work on Tiger Preview 10.4.

http://members.dslextreme.com/~reade...boincbeta.html

Enjoy.

[davygravy]

mikkyo,

Thanks for the tip. I'll probably give it a try, after I am certain that my "overflow" problem what I was having is sorted out. (... you mentioned you had a similar problem on a G5 w/ 1GB RAM, and though that seems pretty dissimilar to my B&W G4, I found I was getting those overflow errors on several of my machines...so I've since set them to what you said, and the results look encouraging... I haven't _yet_ seen another overflow since the new preferences went into effect - I need another 2-3 days to check to see how well those results validate)

When you say that the newer gcc3.55 compiles are running the whetstone benchmark at double the standard-issue's speed... does that translate into the client or worker being twice as fast? Just the client, right?

BTW, out of curiosity, ... How much (%) speed increase should we expect in practice? How will this affect our credits?

Thanks,

davygravy

[mikkyo]

The gcc3.5 are double the gcc3.3 optimized speed.
Faster clients mean less time not running workers, which mean more work.
The difference is small, but enough to matter.
The optimized workers(seti is the only one at this point) make the most difference.
Now a gcc3.5 seti worker might be pretty nice, too bad I can't build seti right now - it appears once again some source files are missing from their source tree.