PDA

View Full Version : top speed



Whity2
10-25-2003, 04:31 AM
what is your top speed in sieveing guys?
mine is 404 kps on AMD XP +2000 (underclocked to +1800)

larsivi
10-25-2003, 07:28 AM
I've had 417 kps on the low range with an Athlon 1.33 GHz

MikeH
10-25-2003, 11:48 AM
445 kps on the dual range with an XP 2100+.

Nuri
10-25-2003, 12:24 PM
516 kps on regular range with XP 2400+

rosebud
10-25-2003, 12:30 PM
474 kps on dual range with a P4 2.6 GHz

Keroberts1
10-25-2003, 04:11 PM
does any one have a dual amd mp 2600+ system? I wish i wanna know what those do. I can only guess but I'm guessing fast.

ceselb
10-25-2003, 04:12 PM
214 kps on a PIII700@800. That's my fastest box btw. :rolleyes:

Nuri
10-25-2003, 04:27 PM
Originally posted by rosebud
474 kps on dual range with a P4 2.6 GHz

Wow!

All I can get at a P4 is one third of this.

To be precise; 135 kps on regular range with a P4 1.4 Ghz and 150 kps on dual range with a P4 1.7 Ghz, which I am using for other projects. But, if there's a way to speed them up at sieve, I'd be happy to shift these two to sieve as well. BTW, my test results are for version 0.37, sse2 client. Other clients perform at least 20% worse.

Which version of the client are you using? Or, may be it's due to some setting or a specific hardware component. So, could you please give some specific details as well?

Thx.

rosebud
10-25-2003, 06:21 PM
Originally posted by Nuri

Which version of the client are you using? Or, may be it's due to some setting or a specific hardware component. So, could you please give some specific details as well?



I use the sse2 client v0.37, too, no special settings.
I think the speed is due to the 800 MHz PSB (Processor System Bus, but don't ask me what that presicely is) and fast memory.

On my another P4 (2.4Ghz) I can't get more than 260 kps.

I'm sorry that I don't have a magic recipe to speed things up :geezer:

Nuri
10-25-2003, 07:17 PM
Thanks for the info. I'll look into those two components as well on my next upgrade.

ceselb
10-25-2003, 10:08 PM
157 kps on a PIV-1.5 with 400Mhz fsb and fast (rdram) memory.

priwo
10-26-2003, 02:50 AM
267 kp/s on a PIV-Celeron 2000@2480 (124MHz FSB) with version 0.37 sse2 client
367 kp/s on a PIII-Celeron 1300@1560 (120MHz FSB) with version 0.37 cmov client

both dual range

and the PIII needs only 150W electricity, the PIV needs 250W

Mystwalker
10-30-2003, 06:49 PM
Just (shortly) tested my P4 (3 GHz, PSB800) with Dual Sieve range:

~ 2x 260 kp/s = a bit over 500 kp/s

Really seems like it hugely depends on memory bandwidth when using a P4.

Another observation:
Using DualCheck range only, I score ~ 2x330 kp/s. So the speed decrease of Dual Sieve is almost neglegtible when compared to the much bigger n range. :eek:

MJX
10-31-2003, 06:30 PM
ceselb :157 kps on a PIV-1.5 with 400Mhz fsb and fast (rdram) memory. 100 kp/s on a celeron 400 OC@411MHz
more than 160kp/s on a celeron 700MHz (can't remember exactly, but faster than a p4 1.5MHz :smoking: )
concerning speed,sieving don't seems to take much care of L2 cache size that is really a problem with other factoring methods (like ppsiqs). Older celeronA may be relatively as fast as p3 or athlons (with sdram) and much faster than p4...

Mystwalker
11-07-2003, 09:53 AM
Currently, my system runs at 2x275 kp/sec = 550 kp/sec. :elephant:


Older celeronA may be relatively as fast as p3 or athlons (with sdram) and much faster than p4...

I guess you mean the non-A Celeron, don't you? The A-type already sports L2 cache and is comparable to the P3 considering speed/frequency.

MJX
11-11-2003, 05:30 AM
I guess you mean the non-A Celeron, don't you?

in fact, I was not talking about oldest celerons that doesn't have cache but about the fact that Sob sieving speed is the same with 128,256 or 512kb L2 cache (and maybe with 64kb too)... Other sieving computations (as the sieving phase of ppsiqs factorization method) are many times slower with small cache mem because all the sieve has to fit in it...

This is not the case with SoB and it is still fun to join the sieve even with old cpus!:geezer:

Best wishes.

Philipppe.

maniacken
12-29-2003, 11:26 PM
just for a laugh, but i think i win the speed competetion.

a snipet from the SOB.dat file

pmin=162026480078699 @ 488 kp/s
pmin=162026490078767 @ 4294967104 kp/s
pmin=162026500078823 @ 489 kp/s


boy i only wish sieving went that fast.

specs

AMD 2100+ non OC
256 mb pc2100

kenny

cedricvonck
12-30-2003, 03:39 AM
280 kp/s on a P4 2.6 (version .39 of the client)
190 kp/s on a P4 1.6 (version .40 of the sse2 client)

Mystwalker
12-30-2003, 12:53 PM
Originally posted by maniacken
boy i only wish sieving went that fast.

Well, when I started sieving (which was a couple of weeks after it went public), the best guys had like 10 kp/sec sieving speed. So I think increasing the performance by factor 50 within 1 year is nevertheless quite impressive, isn't it? ;)

Keroberts1
12-30-2003, 02:59 PM
I just downloaded the new client version .40 and it seems ot have given me a speed increase but i didn't expect it to so i don't remember what speed i had before. I'm quite certain it is faster now though. Things get faster all the time it seems. I'm getting almost 500 kp/s wiht an AMD 2000+

The sp[eed it runs at seems ot be alot more consistant too. It gives the same steady speed constanly when not being used as opposed to varying slightly in either direction by about 5-10 kp/s.

mklasson
12-30-2003, 04:51 PM
I rewrote the timing routines in conjunction with the bsd port, so that may explain the more stable rate. It shouldn't be any faster though, so if it is then it's quite likely just an artifact of the new timing code.

Keroberts1
12-31-2003, 12:48 AM
I tried reinstalling the old client and there is a definate 5% speed increase on my computer. At least thats what the programs report as the speed. Perhaps this is just a reporting differance. we'll see ho the completion stats come out in the end. I do expect them to be better because the computer is already ahead of where i expected it to be by this time today. I can't wait to install this version on my other machines after the new year.

maniacken
12-31-2003, 02:55 AM
mine has gone up a little
489 to 493 every little bit helps though


kenny

MJX
01-01-2004, 04:58 PM
the same for me : 106 to 108 kps...:D

b2uc
01-01-2004, 09:37 PM
Originally posted by mklasson
I rewrote the timing routines in conjunction with the bsd port, so that may explain the more stable rate. It shouldn't be any faster though, so if it is then it's quite likely just an artifact of the new timing code.

And the FreeBSD version rocks Sir Mikael :)

alpha
01-13-2004, 03:51 PM
559kp/s on an XP2200+ overclocked to an XP2400+ equivalent (1.8GHz -> 1.95GHz).

Deoje
01-15-2004, 11:56 PM
645kp/s on Amd Athlon Barton 2800+ @ 2187MHz

Does Proth sieve run faster on linux boxes, and can anyone recommend a good linux distro for a linux newbie. How about SUSE 9 Personal or Pro? I have never done anything with linux but I would like to try it out.

pixl97
01-16-2004, 12:12 AM
No, the linux client seems slightly slower. Check out this thread.

http://www.free-dc.org/forum/showthread.php?s=&threadid=5202

pixl

Death
02-04-2004, 04:47 AM
yesterday I've changed my computer from 864MHz Celeron to 1.7 Celeron.
but I didn't notice any speed increase.
avg was 180 k/s became 180 k/s
what's happend?
sse2 looks slightly faster then cmov
and no optimization does 120 k/s
what can I do?

ceselb
02-04-2004, 05:48 AM
That's a celeron II, based on the P4 architecture. It won't be any faster than that. (I've told you already).

It will be better for PRP now though, maybe you should switch to the SoB client on that box instead.

Death
02-04-2004, 07:37 AM
I can't understand how doubled CPU frequency doesn't give any speed gain. =(

And alas, I can't connect to sob.pns.net =( to recieve WUs.

ceselb
02-04-2004, 09:00 AM
Because the P4 really sucks at lots of things and truly rocks at a few others. With 2.5+ Ghz and fast memory the downsides aren't all that bad.

(Don't get me wrong, I like the P4)

Mystwalker
02-04-2004, 07:23 PM
Plus, the Celeron counterpart of the P4 is really slow compared to the P4... :(