UBBFriend: Email This Page to Someone!
Get 3DNow! Message Board
Team 3DNow! SETI
Putting 3DNow! into SETI@Home
profile | register | preferences | faq | search
|
UBBFriend: Email This Page to Someone!
|
Get 3DNow! Message Board Team 3DNow! SETI Putting 3DNow! into SETI@Home
|
| Author | Topic: Putting 3DNow! into SETI@Home |
| Armin High Priest of Athlon |
 posted July 12, 1999 19:22  
I contacted the chief scientist for the project to get in touch with their programming team to check out the possibility to add 3DNow! code into the SETI@Home client to tap into the power of our CPUs which, among other things, should give us a nice speed gain over Intel teams <evil grin>. Seriously, if SETI@Home can do with single precision FP we should be able to put 3DNow! in. That would be good advertising for 3DNow! for other scientific projects with heavy FP needs as well. I will ask AMD for full developer support if SETI@Home cares to get back to me about this. IP: Logged |
| kris_john Celeron Worshipper |
posted July 17, 1999 20:48
Well done on your negotiations to optimise SETI@home for 3Dnow! Hopefully a bit of friendly rivalry between the Intel and AMD camps will provide a much faster SETI@home client! For a while it looked like the code wouldn't be released to anyone... BTW: if you want a bunch of other optimisation tips, try http://zap.to/clubteam Kris. IP: Logged |
| ToM_1st Follower of Athlon |
posted July 23, 1999 13:17
When will the 3DNow! optimization be ready ?? We it shorten the time as mentioned (Factor 3-4) it would be easy to get in Top 50 !! Hurry up programming guys !! :-)) IP: Logged |
| scott-boy Celeron Worshipper |
posted July 27, 1999 08:46
Strange thing is the floating poitn needs in the program are not that high. I get the same rusults on an AMD k6 compared to the K6-II all running mormal with 26hr time. (programs shut down except seti and explorer) Friends with celerons and pII at same speeds get similiar results(depending on packet size)The higher mhrz PC's seem to get it done faster. Now the Athlon should do a much better job. Rember that the seti program is a small program that removes certain info form the relating frequency. it uses basic math processing. Look closly at the program (and size). Tweaking things can help. But my old K6 is jsut at fast at completing the tasks as the k6-II with 3 times the memory and many times the cache. It is confusing me why, but it happens. The k6-II benchmarks so much higher... I stil lwant to know more about using the linux or other versions on windows 98 SE..email me with more info and instructions... IP: Logged |
| Armin High Priest of Athlon |
posted July 27, 1999 10:25
The key influences on SETI@Home performance are x87 speed (same on K6-III and K6 at same clock) and much more than that the speed of the memory that has to deliver the data blocks to the x87 unit (this means the whole daisy chain of L1-L2-(L3)-main memory). You will likely be able to beat a faster K6-III using a K6 with overclocked memory (i.e. 112 Mhz vs. 100) until our optimized algorithm can use prefetch to ask for data that is not yet needed, which will parallelize the operations of memory and x87 (or then 3DNow!) unit. IP: Logged |
| fredro Follower of Athlon |
posted December 10, 1999 11:25
what ever happened to this idea? IP: Logged |
| All times are ET (US) | next newest topic | next oldest topic |
|
|
|
|
|
|
All submissions are copyrighted by their respective authors and are not for re-use in any form without their explicit written consent.
Powered by: Ultimate Bulletin Board, Version 5.37
© Madrona Park, Inc., 1998 - 1999.