I want to go faster (in Seti!)

[Jorden]

Errrm guys, we wanted to go faster, not back in time

[jsb]

to fast...

[Avalon]

Hey. Knows somebody are there clients
that use 3D-NOW! and SSE?

The PC will crunch 2-3 times faster!

[CharlesWA]

Check out the thread I started called "Optimization questions". Maggi has posted some sample code for allowing you to "cache" work units if you're not on-line all the time. I think that might be a more effecient way to run SETI than having several instances at once. Of course if you have a dual processor machine, then run two instances (1 per processor).

I'm not quite sure which processor setup would benefit SETI the most. There are discussions about the caches size/speed of the Celeron vs. PIII somewhere in this forum, you might want to check them out.

------------------
A few computers, some with Matrox stuff...I'll add details later.

[jms]

The P3E is nice, I am running a 500E@765 ~ 5 hours a WU. Much faster than the Celeron 366@550 I had before. Some in the NG alt.sci.seti said something about the client benefitting enormously from the high FSB when overclocking the P3. You should probably also turn off Spread Spectrum Modulated, and CPU Cache ECC checking in the BIOS.

Jan M.

[Pace]

I want to go faster (in Seti!)

I have a P2-350, 256Mb..blahblah...Win2k. Anyway, I was wondering if you guys have any tips relating to what's below or indeed any in general.

When my computer is running overnight I usually run quite a few instances of Seti CLI - as with 1 or 2 the machine processes them and then sits idle for a few hours till I get up in the morning. Is this an effective way of 'caching' my WUs? Or does multiple instances slow it down (if my %'s are high I run upto 9)?

From DOS/Win9x/NT4/W2KPro what version will crunch WUs fastest? - and I my even add Linux to my machine so if this is significantly faster then let me know.

How long has v2.4 been out? And how often do updates come out - and most importantly are they any faster?

I am planning (still planning if anybody's seen my other queries regarding this upgrade!) to get a P3-600E to OC to 800 - is this good for Seti or would a Dual Celeron 500 be better? (Seti isn't the main factor in this choice but it is a factor).

Thanks in advance.

Paul

[Kruse]

Hmm... 5 hours seems awful slow for a 765 MHz machine! Is this because the coppermine is worse than regular PIII?

I have a vanilla PIII-450 running at 558, and it nails most WUs in just over 6 hours, and some in 5,5. You do the math.

[Pace]

Thanks guys, but Spread Spectrum Modulated? Please help Jan!

Paul.

[CHHAS]

Spread Spectrum Modulated is a BIOS setting, don't know what it does though.....

------------------
P3 500, 224 MB ram, G400 16SH, SB Live Value
Maxtor DM 40+ 30GB, IBM Deskstar 16GP 10GB, Maxtor 4320 13 GB
8 hours avg on the last 30+ SETI WUs

[Assimilator1]

Kruse

We've also found this ,we believe that its down to the cache size.The SETI code (?)doesn't fit into the L2 cache of the coppermine ,but does fit into the Katmai L2 cache ,hence the speed difference.
BTW the Celeron FCPGA seems to perform worse than the older Celeron by a large margain

err & I would hardly call 5 hrs to do a WU slow though!

[jms]

Pace,
Spread spectrum modulation is basically causing the fsb to slightly vary over time. This is done because if the system is running at one speed, it creates "noise" at that frequency, and can disrupt other devices.
Some say this comes with a performance penalty, so you should probably turn it off.
Jan M.

[CHHAS]

I tried setting my RAM to CAS 2 instead of CAS 3, this caused my old 32MB stick to refuse detection, but didn't cause any unstability.

My SETI WUs gets done ½ hour faster now
(avg 7½ hours on the last 5)

------------------
P3 500, 224 MB ram, G400 16SH, SB Live Value
Maxtor DM 40+ 30GB, IBM Deskstar 16GP 10GB, Maxtor 4320 13 GB
8 hours avg on the last 30+ SETI WUs

[Kruse]

Go CHHAS!

I was considering upgrading my PIII450 to a PIII700 Coppermine, but now I'm not so sure. Too bad the cache size makes such a big difference. Guess the choice is between a little bit of fast cache or a lot of slow cache. SETI seems to prefer the last (meaning a vanilla PIII700 would most likely outperform a Coppermine PIII700 when it comes to SETI).

Guess I'll just have to wait for a 512 KB Coppermine...

[Rags]

Hi there Kruse,

My P3550E actually outperformed my P3 450 overclocked to 560 in WU's. I don't know what my averages are now, maybe EES can show another chart with last weeks averages.

Rags

[Ees]

Right. Last weeks fastest seti@murcers:

<pre>+-------------------+------+--------+---------+
| name | wus | hours | average |
+-------------------+------+--------+---------+
| Rags | 33 | 48.50 | 1h 28m |
| Manex | 44 | 126.00 | 2h 52m |
| Guyver | 40 | 175.32 | 4h 23m |
| Lunti | 15 | 87.66 | 5h 51m |
| Eye MD | 17 | 100.12 | 5h 53m |
| Greebe | 56 | 350.64 | 6h 16m |
| Mark F. | 84 | 525.96 | 6h 16m |
| Jorden | 6 | 37.57 | 6h 16m |
| hominid skull | 55 | 359.40 | 6h 32m |
| Jakob Kruse | 36 | 236.12 | 6h 34m |
| paulcs | 40 | 262.98 | 6h 34m |
| Atelier Media | 65 | 438.30 | 6h 45m |
| acobra | 13 | 87.66 | 6h 45m |
| minsoo | 12 | 82.35 | 6h 52m |
| VSA | 60 | 438.30 | 7h 18m |
| Kruzin | 17 | 125.42 | 7h 23m |
| KvHagedorn | 54 | 403.27 | 7h 28m |
| Paul L Gerspacher | 35 | 262.98 | 7h 31m |
| CHHAS | 19 | 142.84 | 7h 31m |
| sleepingdragon | 16 | 120.40 | 7h 32m |
+-------------------+------+--------+---------+
</pre>

Eeeeeek! 1h28m?

Martin