LINUX Beta Test App 4.58 available
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250620213
RAC: 34650
19 May 2006 12:47:15 UTC
Topic 191249
(moderation:
A new Beta Test App is available for Linux at our usual Beta Test Page. It features an assembler-coded "hot loop", and has been built with a somewhat unified build process that will probably work on other platforms, too. It's not surprisingly faster than the 4.55 App, but still a bit, and I'd like to have some feedback regarding certain properties (e.g. validity of results) to see in which direction to proceed.
Happy testing!
BM
BM
LINUX Beta Test App 4.58 available
)
I think you need to change the md5sum on the page as well.
duke out
RE: I think you need to
)
I just checked on two machines that the one listed there (0f231bd3c057d3e2fff2ed95aa2cc696) is correct. Your browser might have cached an old version of the page, or you might have had a download error.
BM
BM
2 WU - mixed 4.40 and 4.58
)
2 WU - mixed 4.40 and 4.58 crunched fine, 1 validated 1 pending
2 WU - 4.58 only crunched fine, 1 validated, 1 pending
early yet but 4.58 seems to be munching away fine on my AMD XP2400+ / SUSE 9.1
Thanks Bernd.
7 WU's done with Albert 4.58
)
7 WU's done with Albert 4.58 in my Athlon Thunderbird 1.33 Ghz (no SSE).
The app is 5 % faster than 4.55 in this machine.
5 WU's with granted credit.
1 WU valid but still pending for granted credit.
1 WU marked with a "validate error" and pending for credit
(link to result)
I suppose this one may be caused by the validator error anounced in E@H main web page. I got a similar "validate error" recently with Albert 4.55, too (link). And this is inusual, because in this non-SSE processor Albert 4.55 has worked very well, with no validate errors.
This new test app 4.58 is 36%
)
This new test app 4.58 is 36% faster than the 4.55 on my Pentium M 1.70GHz, Linux 2.6.15-1-686!
Up to now I got 10 valid results, 6 pending.
Thanks Bernd!
Good evening, 4 WU's
)
Good evening,
4 WU's returned successfully. 3 with pending credit and 1 validated.
Approx. crunch time on a Pentium IV 1,7 GHz :
4.55 : +/- 20100 sec
4.58 : +/- 11100 sec
Cu,
Bernard.
Hello ! First Wu validated
)
Hello !
First Wu validated with Albert 4.58 on my PIII 800Mhz.
The time has dropped from 46.000 seconds with Albert 4.40 to 16.000 seconds with 4.58. Thats impressive ! Turbopower.
Hi, 13 WUs valid with
)
Hi,
13 WUs valid with 4.58, 7 pending, none invalid.
Speedup from 8900 sec/WU to 5700 sec/WU (4.55 to 4.58)
PC used: A64 3200+Clawhammer on SuSE 9.1 (HostID:512894)
A fresh WU-series on another similar machine is being
calculated at present at 10 minutes/result, the fastest
speed, I've had on that PC had been around 1 hour/result.
Seems this piece of assembler is doing magic... :-)
RE: A fresh WU-series on
)
Looks like you had got some small units. They are not very loved by the guys with very fast machines, because they were only allowed to crunch 32 units per day (the curse of 32). It is been discussed to manage the distribution of the small units to slow computers. I don´t know how this would be handled in S5-run.
On my 800MHz P3 coppermine
)
On my 800MHz P3 coppermine the v4.58 app seems to be 27-30% faster than v4.55
Work unit 8712621 is kind of interesting. My 800MHz P3 running v4.58 took just 29.4 seconds longer than a hyperthreaded 3.00GHz P4 running v4.37
Nothing travels faster than the speed of light with the possible exception of bad news, which obeys its own special laws.
Douglas Adams (1952 - 2001)