Found this interesting.

I was cruising a website, looking at links, and saw one that grabbed my attention. It was a link to amdmb.com and they had a review of the new Tyan Tiger SMP AMD motherboard. So I figured that I would take a look at it and see what them benchmarks said.

Man was I suprised. I figured that it would blow my 800's@1002 out of the water. It didn't.

This system was two 1.2 Ghz AMD's.

Here are the tests that I looked at.

CPU Multimedia
AMD:
Floating Point: 15821
Interger: 13388

My Intels:
Floating Point: 13409
Interger: 10961.

Not bad, I expected to lose this. AMD's always bench better in Multimedia so this didn't suprise me. The next one did!

CPU Benchmark.
AMD:
ALU: 4321
FPU: 3295

My Intels:
ALU: 5401
FPU: 2682

Why, oh why, are my Intels smoking this AMD SMP rig in ALU (MIPS)? It shouldn't happen. The AMD's have more cache, more speed, are touted as being faster than PIII's, higher FSB....you name it.

I'm wondering if AMD has blown it when it comes to SMP?

well, what chipset does this MB have??I've heard that the MB's aren't that good, slowing it all down...so,I think at the beginning the AMD Dual-setups won't be comparable, it's the FIRST time they have DUAL-cpus (or am I wrong??)
well, as far as I remember, there are dual-socket7 boards, for running 2 pentiums...

I don't know what chipset they had. But come on, it's a $400 Tyan board....

Kingslayer said:

I don't know what chipset they had. But come on, it's a $400 Tyan board....

Click to expand...

yeah kingslayer, that is just plan SAD.. bet at 1.10ghz I could woop it.. or really close ..

Just another reason to stick with a dual intel system if you asked me

VP6's RULE!

-Trek

Yeah, I'm going to try and take these 800's up to 1100. I don't think the whole system can handle it though. We'll see. I'm going to be moving my computers to a smaller room in the house and I'm going to route the air conditioning to the computer case via dryer exhaust hose.

I just dont thing the NIC will take it. But I would bet that at 1100 it would smoke that AMD.

Intels cache architecture is better than AMDs cache architecture. The internal connection to the L2 cache is wider and faster than AMDs. This accounts for your PIIIs outperforming the dual AMDs in Integer performance. On the other hand the AMD chips do have a strong FPU when it comes to legacy X87 floating point code. That shows when the AMD system beats your system in FPU performance. So, your results look exactly as they should.

I'll tell you a secret. Depending on which benchmarks you use it's possible to do a comparision between 2 different CPUs and get whatever results you want. Because the architecture of AMD and Intel cpus is different they both have slightly different performance characteristics in different areas. So, while one may be stronger in one area there are areas the other one may be stronger. Alot of these benchmarks you've been seeing on websites are done using floating point intensive applications which make use of legacy X87 instructions that perform very well on the AMD strong floating point unit. I'm not saying the AMD isn't a strong CPU. It is, but most of it's strength in in the floating point unit and it has weaknesses in some areas as well.

Also AMD smp implementation may be less effecient than Intels. I'm sure you recall hearing about how AMD was gonna use seperate connections between each chip and the norhtbrige and how some people mistakenly thought this would be much better. Actually it isn't. It isn't because it makes it more difficult for each CPU to snoop the others cache. Furthermore it presents no real advantage as far as bandwidth is concerned because the connection to the memory is what determines that and both cpus are still gonna be contending with each other for access to the same memory bus . So, all that talk was basically just BS. If yu look at some of the recent comparisions between AMD MP and P4 Xeon systems you'll see that the AMD smp implementation doesn't appear to be as efficent. There were a few benchmarks where a single AMD beat the single P4 but then the dual P4 Xeon would slightly edge out the dual AMD configuration. Even when the dual AMD system did come out the winner the margain of victory would often narrow when going to a dual configuration for both. That is a single AMD would beat the single Xeon by a larger margain than the dual AMD would beat the dual Xeon. This shows that the AMD smp implementation isn't as efficent as Intels. This is probably why they only recommend the new MP processors for SMP. They probably have enhancements to help out here.

Some really good information there sp!! This thread really sheds light on the SMP battle, I was sure AMD had it all locked up but I guess not!