On Tue 2003-02-18 at 14:47:42 +0100, [EMAIL PROTECTED] wrote: > Am Dienstag, 18. Februar 2003, 08:39:33 Uhr MET, schrieb Chuck Shirley: > > I wish them the best of luck, but it's like asking microsoft to play > > "Russial Roulette" with an auto-loading pistol, if you ask me... > > :-) > > > Of course it would make for some very inexpensive clustering systems, eh? > > Not according to this test: http://www.shadowflux.com/xbox.html > (google cache: > >http://216.239.51.100/search?q=cache:0v2HbYP42lsC:www.shadowflux.com/xbox.html+xbox+cluster&hl=en&ie=UTF-8 > )
<rant on>
Hum. Yeah. IMHO, that "test" is completely useless. He mainly laments
about setting up (incl. modding) being hard. There is only one
paragraph with results and those are incomplete.
He complaints about the MPI program not working well, but doesn't show
the source. He says the results of the MPI program were inconsistent,
but doesn't know why. What is a non-reproducible test worth? From what
is left of my experience with MPI clusters from university time, the
main reason for inconsistent behaviour is when the program logic is
flawed.
All in all, I only learned that he has a lack of knowledge of
scientific working (remember, the 3 Xbox'es were founded by his
College), than anything about Xbox as cluster. :-(
</rant off>
Sorry about the rant, but he goes out to test Xbox as clustering
solution and then concentrates on the part (modding) which is supposed
to be made obsolete by the very project that he used as basis for his
test (Xbox Linux). Even if he doesn't believe this will become
reality, I would have expected him to at least mention that
possibility.
Bye,
Benjamin.
msg93041/pgp00000.pgp
Description: PGP signature
