posted December 09, 2000 16:29         

I haven't been able to crack the 4 hour time yet, but have gotten close.

It's a 900 MHz T-Bird core Slot A on FIC SD11 board with 100 MHz memory. AMD 750 chipset, board is a year old. First generation Slot A Athlon board and these are the best times it's done in SETI 3.0 command-line client in Windows 98SE:

12/7/00 11:55:57 AM 26se00aa.11918.23601.934640.180 1.738 2.670 11.500 2.173 4.112 99.99%
12/8/00 3:57:52 AM 26se00aa.11918.21152.153388.41 23.771 18.550 11.783 2.172 4.130 99.99%
12/8/00 3:18:49 PM 25se00aa.7185.23314.842330.33 15.919 26.430 11.903 2.172 4.207 99.99%
12/9/00 9:12:02 AM 25se00ab.6988.30274.592332.77 11.236 26.630 11.582 2.173 4.129 99.99%

(Taken from SetiSpy log file)

Anybody on 3dnow running a 1.2 GHz T-Bird?

If so please post your best times. I'd like to see a sub 4-hour SetiSpy log.

Cruncher

IP: Logged

posted December 09, 2000 17:01         

I added a new SETI account under the name SpeedyG+, so that's the name you see on the SetiSpy snaps.

SpeedyG+ will be a GHz+ machine when it's finished. It's still an idea right now because it is based on the 760 chipset, which AMD is still working on getting the bugs out of.

The idea is a machine with above 1 GHz processor speed and 1 GB of RAM minimum, with above 1 GigaFLOP performance.

But the 900 MHz T-Bird core Slot A is already running 1.2 GigaFLOPs, which fascinated me when I first saw it in SiSoft Sandra but there it was, so the SpeedyG+ concept should probably be a machine close to 2 GigaFLOPs when running DDR memory (hence the "G+" )

Come on AMD, get those critters out already!

Cruncher/SpeedyG+

IP: Logged

posted December 10, 2000 14:06         

IP: Logged

posted December 10, 2000 15:43         

How fast can these things go? Are 2 hour times possible for desktops, and what kind of AMD powered machine is it going to take to do that?

This is getting insane. I remember with a K6-3 thinking an 11-hour time was fast, now I do 4 hour times all day long, and here we are talking about 2 and 3 hour times.

When will it be possible to get a 1-hour time I wonder? What AMD powered machine is going to be able to crack that nut... what AMD processor and what type of memory is going to be able to do the 1+ hour SETI work unit?

Hey Danball have you tried that WU test at Ars Technica Lambchop? You ought to run it on your system and post the time over there in their stats board (if you haven't already):
http://www.teamlambchop.com/bench/benchfile.htm

I started doing it and it looked like it was going to be a low 4 hour time WU, and I didn't finish it 'cause it was in the 900 MHz T-bird time range on Ars Technica already. But if you got a screamer, maybe you get a 3-hour time with the WU and put it up on their board that an OC'ed Duron 600 did it

It's nothing but Intel freaks over at Ars Technica anyway, so a little Duron going nuts on that WU would wake 'em all up and make 'em blink twice I think

SpeedyG+/Cruncher

IP: Logged

posted December 10, 2000 16:02         

The Alpha is still the King at 2:27, but somebody put up a P4 time at 2:30 on a *supposed* "stock" 1.5 GHz P4 (I'll believe that time when I see about 5 other similar times for "stock" P4s up there, given that almost every machine on that page is OC'ed).

Interesting to see Roelof of SetiSpy pushed a T-Bird to a 2:52 time in Win2K. Interesting because he is a die-hard "Intel is the Undisputed King of Microprocessors" type (and Budweiser is the King of Beers! really, they are! Heineken and Moosehead suck! really, they do!)

Looks like the race to 1-hour is already on!

Cruncher/SpeedyG+

IP: Logged

posted December 10, 2000 23:50         

IP: Logged

posted December 11, 2000 01:01         

[This message has been edited by DanBall (edited December 11, 2000).]

IP: Logged

posted December 11, 2000 11:33         

[This message has been edited by DanBall (edited December 11, 2000).]

IP: Logged

posted December 11, 2000 18:50         

They say to email the results to "Max" mailto:maxathome@ntlworld.com or "RB" mailto:ratbastard@mindless.com until they get the "issues" on the submission page worked out.

Cruncher/SpeedyG+

IP: Logged

posted December 13, 2000 06:10         

wtemp.sah file shows this on the work unit: cpu_time=14470.790000

SetiSpy log for the work unit is this:

12/13/00 2:44:09 PM 11ap99aa.3187.8034.286082.173 5.879 20.850 6.718 2.186 4.020 99.99% 5 24.626 0 0.000 0 0.923 0 0.000

Cruncher

IP: Logged

posted December 15, 2000 12:27         

Trying to copy from SetiLog, but format is messed up, so here is the data.

3.859hrs. angle 6.853
3.908hrs. angle 2.620
3.857hrs. angle 7.813
3.793hrs. angle 7.325
3.850hrs. angle 4.157

All with WinNT CLI 3.0 under Win98SE

IP: Logged

posted December 15, 2000 18:02         

I'm in a bit of a debate right now over whether Ramdisk has any advantage over fast hard disks.

I ran tests using Ramdisk in Windows 98SE and IBM 7200 RPM ATA100 hard disk in Windows 2000, WinNT CLI 3.0 in both systems. I used the Ars Technica SETI "benchmark" WU in both systems. These are the results:

In 98SE with the Ramdisk cpu_time=14470.790000.
In Win2K from the hard drive cpu_time=14622.836606.

From SetiSpy's log for the different operating systems, with and w/o Ramdisk:

Windows 98SE Ramdisk
12/13/00 2:44:09 PM 11ap99aa.3187.8034.286082.173
5.879 20.850 6.718 2.186 4.020 99.99%

Windows 2000 IBM 7200 RPM drive
12/14/00 2:31:11 PM 11ap99aa.3187.8034.286082.173
5.879 20.850 6.718 2.186 4.062 99.99%

IP: Logged

posted December 15, 2000 22:04         

IP: Logged

posted December 15, 2000 22:54         

IP: Logged