T-Bird or P3 for S@H?

[The Magicman]

Hi!

Me again...
I need a little advice which CPU to buy for S@H.

P3-700@1050MHz+, 150MHz+ FSB CAS3

or

T-Bird 900@1050MHz+, 117Mhz+ FSB, 140MHz mem CAS3

 

[Hellburner]

That's a hard one to call, with the old client the P3 would win hands down but with the new clients coming out now it's an entirely different ball game. There just isn't enough data to authoritatively say which one is best. The few mentions I've seen give a slight edge to the P3, but I think it's been mentioned that might be mostly a chipset problem...

 

[blade47]

I would guess the P3 but as HB said you never know until v3.0 actually comes out.

 

[The Magicman]

As I understand the P3 would win with a little margin (if any).
Would it be enough to satisfy a ~150$ price difference?
My guess would be: no

Thanx for the answers.

 

[xtreme2k]

well i think you can take it this way

in a generally intel optimised program

the P3 with high FSB kicks the Tbird


but when programs reoptimise for the AMD

Tbird kick ass

 

[Assimilator1]

xtreme2k

|>>>>>in a generally intel optimised program<<<<<

Are you sure? ,as fas as I know there is no optmisation for any platform ,the PIII's tend to do better (on BX boards anyway) because of the better memory performance(latency) of their boards.

If I'm wrong though ,I'd like to see the site you got the info from.

 

[blade47]

I agree with Assim1. It's not even so much the bandwith of the memory but the latency, as Assim1 stated. Since the new clients are less memory dependent the AMD's do better than they used to.

 

[Orange Kid]

>>>>>in a generally intel optimised program<<<<<


Talking -------- SSE vs 3Dnow

 

[Orange Kid]

Double Post

 

[blade47]

Didn't the seti people say they don't use sse or 3d now instructions? I may be wrong but I thought I read somewhere that the clients don't use either set of instructions.

BTW I don't see how they could be using SSE, because doesn't a P2 with the new client perform the same as a coppermine or katmai P3 at the same clock speed. To the best of my recollection the P2 didn't have sse. I thought it came about in the P3.

 

[Hydra]

Does Seti use the FPU? All I know is that it's doing FFT's and that those calculations don't involve floating point math (IIRC). Anybody with more info on this subject?

 

[BadThad]

YES, SETI uses the FPU!

Assim1 - Many programs are optimized natually for intel cpu's. It has nothing to do with SSE vs 3Dnow. I am not a programmer, but perhaps someone out there can enlighten us. I've read about this for years now but I forgot the rationale.

 

[Hydra]

I believe it has something to do with most compilers defaulting to optimizing for Intel. One of the fastest/best compilers at the moment is the Intel C/C++ compiler which, naturally, optimises for Intel.

Too bad we can't have the DNet source so you could try to get your compiler to optimize for your CPU. On the other hand, I believe DNet is not releasing the source because it could be a means to create a 'hacked' client.

 

[blade47]

Hydra, these guys are talking about S@H. Which is not associated with Dnet.

I also thought that RC5 ran better on AMD chips anyway?

Fear of hacking the client is why seti@home doesn't release the source code.

 

[Hydra]

Sorry, I'm rather single-minded ;-)

But the same is true for S@H right? If you have the source you probably could write a client that would return bogus results just to get a high ranking. Thus totally screwing up the contest.