DES Keyrate Rapidly Growing
David Hallowell
writes "As you probably know the EFF's Deep Crack machine
which won the last DES challenge is working with
distributed.net for this contest. There's
a graph
which shows the performance of deep crack (labelled DC)
and the rest of the distributed.net effort (labelled DCTI)
which is composed of everyone running the rc5des clients.
The keyrate from the combined rc5des clients is double
that of Deep Crack compared to that of last time when it
was less than Deep Crack (which has a constant keyrate).
The combined rc5des + deep crack keyrate is also shown."
Also, Team Slashdot is #2 in
the top 100 if you're curious.
You are definitely doing something wrong. I have the same processor, and under Linux it gets around 2.5Mkeys/sec! Maybe rc5 client is not getting scheduled because of something else running.
Something my UltraSPARC systems can crunch on... my Ultra 5 managed 4.4m keys/sec even while I work on it and play MP3s in the background, et al.
:(
So frustrating that the RC5 is inefficient on SPARC systems
...Steve
if those numbers are right we should exoust the
key space in about 5 days
The most striking thing for me is the poor showing of Team EvangeList. My theory - the lack of a quick-switch client for the Mac meant a lag of several hours before most people started crunching DES blocks. It took me three hours before I remembered to force a switch.
I think I'm going to throw in the towel on distributed.net after this, at least until they switch to OGR or Mersenne primes or whatever the new plan is. RC5-64 - who cares? It's just infinite monkeys typing and getting it sooner or later. I can't get myself worked up over SETI either.
What I'm listening to now on Pandora...
If you subtract out Deep Crack's (toad.com) numbers from the #1 team, we're the fastest software-based team.
Plus, I joined Team Slashdot just now (with keys already submitted) because it hasn't been possible to join a team for most of the past 3 weeks while the DES-TESTs were running.
Software sucks. Open Source sucks less.
Hmmm... maybe so... :)
fetch and flush on my clients... now they're crunching RC5.
83 Ultra 10's will do that, I guess :)
oh well, back to work.
[Jan 19 15:36:12 UTC] Completed one DES block 0033EDCB:80000000 (8*2^28 keys)
spoke:~/rc5> uname -a0.00:19:40.70 - [3,637,619.83 keys/sec]
Linux spoke.nols.com 2.2.0-pre7-ac4 #5 Sun Jan 17 14:35:16 PST 1999 i586 unknown
Run top for a second and kill the process that's eating up your cpu. Don't leave top running though, it's a cpu hog too.
Who are they? Where did they come from? And , most importantly, how the hell did they scrape up three times our kkeys/sec?
Aren't you dead?
Great, people I know were already saying that
d.net wouldn't make it against Tha Big One, but
now there seems to be a chance !
---- Boring sig.