20020821, 20:45  #1 
Aug 2002
1001000_{2} Posts 
4096 P4 help wanted
I am going to be running a 77.9 M on a fast P4, that has The highest quality PC2700 i could get my hands on(cl2 @190, CL2.5@220) I would like to see some benchmarks of 2.42.5 Ghz P4's(to eliminate processor speed as a variable) doing 4096 ffts. Also what ram timings, type of ram,and chipset are you using. Anyone who can help will be well appreciated. I will be posting my timings as I get them tonight. Thanks.

20020821, 21:20  #2 
Aug 2002
3^{3}·313 Posts 
Is it 4096K or 4608K? I ask because for the time I did mine (I gave up after 2 weeks!) I used a 4608K FFT...

20020821, 21:32  #3 
Aug 2002
2^{3}·3^{2} Posts 
from the status page at mersenne.org
[code:1]69100000 79300000 562,700 0 299,584 0 0 263,116 0.35 8,866,276 2.604 4096 [/code:1] 
20020822, 00:41  #4 
P90 years forever!
Aug 2002
Yeehaw, FL
3·11·239 Posts 
P4 2.329 GHz is 246 milliseconds
PC870 RDRAM Note: Version 23 will likely speed this up by 7% (or more). 
20020822, 03:04  #5 
Aug 2002
2^{3}·3^{2} Posts 
ok, P4 2.46ghz DDR308 cl2
244 ms(average 100 iterations) Lemme find a keyboard and i will try 4/5 (cpu/ram) and some other CL values. I should be able to get DDR380, 266, 333, and 400(2 systems) /edit version 22.8 knocked a ms off the time versus 22.7. That is nearly 22 hours saved. 
20020822, 05:31  #6 
Aug 2002
256_{8} Posts 
How to benchmark such high FFT size?

20020822, 05:39  #7 
Aug 2002
48_{16} Posts 
in the advanced menu choose time and enter 77900000 for the exponent size, and a number for the #of iterations(i use 100 as there are many things that can skew a benchmark, the larger sample size helps keep errors too a minimum)
Then take the total time and divide by the number of iterations for an average, or use the best time and post it. I like to use the average, but I am not picky. And thank you. 
20020822, 05:45  #8 
Aug 2002
2·3·29 Posts 
~224ms
P4 1.8@ 2520MHz RD1120 
20020822, 22:21  #9 
Aug 2002
101 Posts 
This thing is pretty linear until some 'outside' factor kicks in, such as the cache size limitation. Let t be the table size and s be the time in miliseconds, my XP 1800+ is roughly
s = 0.1t  4 It gives me ~400 ms for ~4000k table size but my test told me 466. Need another approximation once the 256k cache shows. For 512k cache, I suspect it is linear till ~4000k. Once we are in the 200 day range, a day or two wont matter as much. 
20020822, 23:10  #10 
Aug 2002
Rovereto (Italy)
3·53 Posts 
some data...
Average time on 1,000 it:
P4 1.6a@2.24 (16X140)=260.483 P4 2.0a@2.3 (20X115)=246.706 P4 2.26b@2.6 (17X153)=223.251 P4 2.0a (20X100)=294.332 P4 1.8a@2.52 (18X140)=234.151 Regards from Italy. Guido 
20020822, 23:22  #11 
Aug 2002
Rovereto (Italy)
3·53 Posts 
Re: 4096 P4 help wanted
[quote="Tasuke"]I am going to be running a 77.9 M ...quote]
Hi! My faster P4 should take less than 195 days for testing M77,900,000 wich is not much more than 150 days needed by my "old" TBird 0.9 Ghz for a 10mio digits exp. some time ago. So I'd like to test such a monster number as well... How to ask Primenet for exponents like these? Thanks and regards from Italy. Guido 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Most Wanted  rogue  FermatSearch  35  20210315 14:06 
New wanted ranges!  ET_  FermatSearch  12  20161109 13:31 
Since when was 4096 prime?  Gordon  PrimeNet  6  20150704 17:30 
Most wanted  kar_bon  Riesel Prime Data Collecting (k*2^n1)  15  20110809 16:50 
100 Most Wanted  Citrix  Factoring  24  20040222 01:05 