AMD vs. Intell

[Puer Aeternus]

Okay folks....All you AMD users read up!! I am being humiliated by a couple 1 gig p3 computers that Skyhook is running SETI with.
his system is below:
_____________________
Skyhook's rigs*
Iwill BD133U with a PIII-S 1.26 @ Default
512meg PC133 @ 222/5-7
ATI Radeon 8500(retail) @ Default

Box 2:
The same except:
GA-60XET-C motherboard
______________________
Now I have been using Skyhook as a reference as I pulled in WU's and moved up the ranks..now that rascal has passed me by one wu and may just whup my fanny good!! his average is 10-12 wu's a day....mine is 6-8. How the heck did his 2 rigs beat mine?
As you can see below I have my rigs in my sig. How can a 1 gig p3 beat a t-bird 1.4 @ 1.6ghz? And is a p3 1 gig faster than a Duron 1 gig..I can accept that.

*Ican tell Skyhook has a good sense of humour and hope he does not mind the competition and good intentions of my post I dont mind you beating me...but I do mind you beating me your Honda civic vs. my mustang..LOL(Just Kidding Sky, you have 2 nice rigs there!! )

 

[Yodums]

Well, for one your not on DDR so it's going to slow you down quite abit, and your motherboard won't take advantage of that chip since it's a 266fsb chip. The P3-S chips w/ 512K Cache crunch pretty fast. Around 4hours stock I'm guessing (TC's crunched 3.3hours overclocked I'm guessing). And he has two so that's like 10-12 a day!

 

[TC]

A P3 is nothing to laugh at. In many ways it is superior to a P4, although it will never be able to ramp as high in clock speed. I had two, sold one to Morpheus and Mictlan gets the other. I still have one here and clocked around 1500 they crunch just as fast as my XP's running in the 1600-1700MHz range. I'm attaching a screen shot of my seti queue result log. The P3S is named Tualatin2 and is clocked at 1481MHz. The A7M266D has a pair of XP's clocked at 1700MHz. The Tiger is a pair of 1500+ XP's @ default. Local host is a dual 2GHz Xeon running 4 clients - two per cpu in hyper threading mode.

 

[Yodums]

A P3 is nothing to laugh at. In many ways it is superior to a P4, although it will never be able to ramp as high in clock speed. I had two, sold one to Morpheus and Mictlan gets the other. I still have one here and clocked around 1500 they crunch just as fast as my XP's running in the 1600-1700MHz range. I'm attaching a screen shot of my seti queue result log. The P3S is named Tualatin2 and is clocked at 1481MHz. The A7M266D has a pair of XP's clocked at 1700MHz. The Tiger is a pair of 1500+ XP's @ default. Local host is a dual 2GHz Xeon running 4 clients - two per cpu in hyper threading mode.

Hey I got 3.3hrs exactly! Great chip! They seem to be more responsive than the P4.

 

[TC]

Well keep in mind the P4 here is running two clients simultaneously on each processor. That would more than double the average time on a regular cpu, i.e. the P3 would be taking about 8 hours a unit if I opened another client.

 

[SkyHook]

HONDA!!!

I may be a Volkswagen Bug(original edition), but I ain't no HONDA!

Actually our conversations have caused me to stay on my toes and were and still are a welcome part of this whole experience. I know my chasing you has given you the opportunity to get advice from some of the other team members on improving your systems for more output and it has shown in increased production by yourself over the past couple of days.

In the spirit of friendship I am promising you now, I will set on 499 for as long as it takes for you to catch up so you can get 500 first.

Your turn my friend!
SkyHook

 

[LandShark]

the newer Tualatin core PIII is much faster than the Coppermine core. at least for SETI.

the fastest PIII under my sig. crunch just under 6hr/WU. despite the fact that it's running @ 1.32G (120x11). a stock 1.26G PIII-S would crunch much faster than it.

 

[TC]

the newer Tualatin core PIII is much faster than the Coppermine core. at least for SETI.

the fastest PIII under my sig. crunch just under 6hr/WU. despite the fact that it's running @ 1.32G (120x11). a stock 1.26G PIII-S would crunch much faster than it.

You know what, I just remembered they added hardware pre-fetch just like the XP's and P4's. Totally forgot that. That accounts for the speed increase - not just a die shrink and faster clock. I wonder if the tualatin would like DDR more than the old core?

 

[SkyHook]

Sure Helps

The 512k of full speed L2 @ 0 Latency seems to be a big part of what makes these PIII-S's take a liking to SETI. But what I've really noticed is how smoothly they handle multi-tasking. You can move freely from one thing to another without any sign of hesitation or feeling of "catching up". Once I get some other hardware issues ironed out I can't wait to see what these two systems can do when I put the spurs to them. And eventually I would like to pair them up on a good duallie board.

SkyHook

 

[SkyHook]

Hey TC

TC, take a look at this guy's members name on our team:

Hedgehog (Dual PIII Tualatin@1540 - 1GB DDR266)

Seems like he could maybe answer your question about the Tullies and DDR. However I have no idea who he is.

SkyHook

 

[Greg M]

The PIII-M's are nothing to sneeze at either. My 1Ghz PIII-M laptop crunches WU's in just a hair over 5 hours.

 

[Mictlan]

I'm also running a PIII-S with DDR.....(and waiting to have a dual setup soon).

I've done a few test with my old setup. I had a BD100+ (with the 440BX chipset) and cunched a WU in 7:30, managing to drop that time to 6:30 running at 115 MHz FSB. I then used a PIII 850 with the DVD266u-RN (using a VIA Apollo266T chipset)mobo and I was getting a 7:15 h with the DDR mobo. That is 4% decrease in time!!!!! Maybe not much, but I wanted to go with a dual tualatin support then it came as an added bonus.

When I upgraded to the PIII-1.26S chip, it just blowed my expectations big time. I was hopping for 4 h/W in average, but I'm getting 3.5 h/WU......I'll post my times next Tuesday, when I have enough of them to give you some hard data.....and compare it with the PIII-S without DDR.

Skyhook: as you are running with SDRAM could you crunch the benchmark WU so we have a good idea of how different are DDR and SDRAM. I'm crucnh it next Monday night.

 

[SkyHook]

Willing to Help

I would be more than willing to help and provide benchmark test info, if someone could be so kind as to help this braindead moron understand exactly how to do that test.

Right now I have SETISpy running one instance of the CLC on each machine. SetiSpy is pointed to and gets and delivers WUs from my SETIQueue. That has all been working perfectly for me and I would prefer not to mess with any changes there. If I were to not run my normal setup on one machine for a time and instead placed another CLC in a completely different folder and ran it just as a commandline client would that be a good first step and what to do from there? If it's simple, I could even run the test on each machine seperately as they have different MBs just to see how they compare. Sorry for being a pain, just don't want to screwup what has been a very successful setup for me so far.

THANKS,
SkyHook

 

[Mictlan]

check the information here. It's the second post in this thread.

Or go here.

The only problem is that I haven't hear from the guy that was taking care of the benchmark result.

 

[SkyHook]

Something For Comparison

I realize these aren't the BenchMark WU, but they are a pretty good representation of how this machine is performing with some different ARs.

ADDENDUM: Just to clarify, these are with NO overclocking, strictly default settings.

 

[Morpheus]

Juan...

I have the PIII-S running on the TUSL2-C with SDRAM... would be ineterested to see if the DDR really helps at all.... (I just don't think the chip can take advanatage).... I am also pumping mid-3 hr WUs out regularly...

I have gotten VERY high FSB out of this chip, but need a better HSF... 1.7 GHZ would be fun

I'm also running a PIII-S with DDR.....(and waiting to have a dual setup soon).

I've done a few test with my old setup. I had a BD100+ (with the 440BX chipset) and cunched a WU in 7:30, managing to drop that time to 6:30 running at 115 MHz FSB. I then used a PIII 850 with the DVD266u-RN (using a VIA Apollo266T chipset)mobo and I was getting a 7:15 h with the DDR mobo. That is 4% decrease in time!!!!! Maybe not much, but I wanted to go with a dual tualatin support then it came as an added bonus.

When I upgraded to the PIII-1.26S chip, it just blowed my expectations big time. I was hopping for 4 h/W in average, but I'm getting 3.5 h/WU......I'll post my times next Tuesday, when I have enough of them to give you some hard data.....and compare it with the PIII-S without DDR.

Skyhook: as you are running with SDRAM could you crunch the benchmark WU so we have a good idea of how different are DDR and SDRAM. I'm crucnh it next Monday night.

 

[Mictlan]

RIght now I'm running it at default speed, as I'm seeing if my DDRAM holds it with the more agressive settings. I've been getting almost all normal range WU into the 3:45 range. High AR have gone to the 3:20 or so. Next Tuesday I'll post the times, as I'll have a ggod amount of units crunched with the chip (about 50).

 

[SkyHook]

Interesting

Going by your last statement compared to the numbers I posted which cover a fairly wide span of ARs, it would appear that your DDR might be getting you a 20-30 minute advantage on average per WU. Given 6-7 WUs per 24hrs. you could be picking up close to an extra WU per day just from your memory.

SkyHook