Seti packets faster on AMD?

[BFD Deadeye]

OK, this has got me baffled, and I was hoping someone could shed some light on this for me.

First off I am NOT trying to start yet another AMD vs. INTEL thread. I think those are ridiculous because they are both excellent CPU makers.

With that said, I had a 1.8 GHZ server machine I was waiting to deliver for a few days running SETI packets so as to monitor CPU temps. The client wanted a desktop case, which of course in stock form, hasn't got much in the way of fan space. It was running around 53-54 Celsius processing said packets, and taking 20-30 hours to complete them.

On the other hand, my 1.4 GHZ T-Bird takes an average of 12 hours to complete a SETI packet, but with my many case fans (2 in, 3 out + PS fan) it only runs 45 Celsius at peak.

Would the cooler temps in the T-Bird account for the faster SETI packet processing? Does SETI send packets at random, or does it know the CPU speed via some kind of hardware ID, so it sends out bigger packets to faster CPU's? If none of the above, what could account for the processing time difference? TIA.

[HAL9000]

The AMD's are faster number crunchers than the P4's Actually, my Coppermine Celeron 1100 and 1000 machines crank out faster RC5 packets than my P4 1.7.

I recently sold an XP1800+ to an astronomy prof here at the university just for the sole reason of number crunching and we both decided that it was in his best intrest to go with the AMD because of the fact he could get more done in less time.

As for temps, I think that's just coming down to you're getting better airflow in the AMD machine. I run the distributed.net client and purr along in the low 40'C range under full load.

[doctorgonzo]

SETI packets are all the same size. Some may need more processing than others, due to the presence or absence of noise and signals. But something really sounds weird if you are getting that much difference between the two processors. Are they set up the same way? The graphics take up a lot of processing power, so if the T-bird is set up to go to a blank screen after a minute and the Pentium is not, that could do it.

20-30 hours on a Pentium 1.8 GHz is a looong time, though. My 1.3 GHz T-bird crunches out packets in 8-9 hours on average. It is the better FPU on AMDs, I reckon.

[BFD Deadeye]

Thanks for the replies, fellas. I'll attribute the difference to the AMD faster number crunching abilities, combined with the extra heat due to lack of case fans in that Intel. The heat must slow it down a tad.

[HAL9000]

Nah.. I don't think the heat is really slowing it down unless you hit an overheat situation, which you are not. I don't have airconditioning in my place and don't notice any variance when the temps go from a comfy 72'F in the house in the winter to in excess of 100'F in the summer.

[glc]

It used to be the exact opposite. I had a 200 MMX that took 29 hours to crunch a unit and a K6-2 running at 417 that took 34 hours. Now it's 11 hours with a P3-800.