• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

intel faster, but why AMD better?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Arkaine23 said:
Current AMD CPU's have an extra ALU (Arithmetic-Logic Unit) and FPU (Floating Point Unit) by design compared to Pentiums.

i never knew this. but then why don't the XP chips crush the intel ones in Floating point operations benchmarks (like sandra)? even furthermore, with SSE2, PIV flops are almost double that of an XP chip?



Yodums said:
Their CPU Benchmarks are about the same however Intel's bandwidth greatly over ride's AMD which really helps the CPU.

the barton core with 512 kb L2 Cache should help close this gap somewhat, no?
 
Ya I have heard of this 1.8A too. That sounds really awesome. With that type of o/c (and on AIR cooling!) I think it will have better bang/buck. Maybe I will get one some day, but for now I am happy with my XP chip.
 
this was an explantion that was given on my cs forums and i think he explains it best these arnt my words but thought you guys would want to see what he had to say.

"In order for Intel to take the crown so they dont look bad to the public eye in the speed wars against AMD, Intel changed the number of bits the CPU processes in a single cycle from 8bits to 6bits. If intel made a p3 2.0GHz and you compared it to the p4 2.0GHz, the p3 would be faster because p3s still process 8bits per cycle.

p3 @ 2,000,000,000 cycles / sec * 8 bits / cycle = 1.6 * 10 ^ 10 bits per second
p4 @ 2,000,000,000 cycles / sec * 6 bits / cycle = 1.2 * 10 ^ 10 bits per second

Then that would utlimately translate down to the number of raw instructions executed per second which we dont need to get into right now.



AMD still processes 8bits per cycle (as the p3s do) and in order not to make themselves look bad in comparison to intels p4 which they "cheated" to appear faster, AMD gives you an equivalent speed rating as if it where a p4. That is the 2000+ number you see packaged on the box. The actual clock cycles, however, is rated lower as you have noticed 1.66 GHz for the 2000+ CPU. If AMD did only process 6bits instead of the 8bits it currently does and just increased the number of cycles as intel did, then the clock cycle speed would in fact be 2.0 GHz or the 2000MHz speed on the package.

p4 2.0GHz = 2,000,000,000 cycles / sec * 6 bits / cycle = 1.2 * 10 ^ 10 bits per second
AMD 1.66 (2000+) = 1,633,000,000 cycles / sec * 8 bits / cycle = 1.3 * 10 ^ 10 bits per second"

from [BC]Twitch.
 
wow nice explanation. i just feel it's all about how thier built. a bridge made from stonger steel will not always hold more weight if it has poor arciteture (sp?).
 
Ya I like your (3/P4 comparison. It makes sense. I was wondering why the P3s are more effevcient per megahert than the P4s in benchmarks. Wonder why Intel changed the architecture...maybe it was necessary in order to give it the mhz headroom???
 
those arnt my words, i dont know to much about how they work like that, thats the leader of my cs clan, he knows a ton about comps, and have always gotten good info from him, i just hope it clears up some of the confusion between intel and amd.
 
Ambiguous said:
this was an explantion that was given on my cs forums and i think he explains it best these arnt my words but thought you guys would want to see what he had to say.

"In order for Intel to take the crown so they dont look bad to the public eye in the speed wars against AMD, Intel changed the number of bits the CPU processes in a single cycle from 8bits to 6bits. If intel made a p3 2.0GHz and you compared it to the p4 2.0GHz, the p3 would be faster because p3s still process 8bits per cycle.

I don't buy that. Can that leader of you "cs clan" offer up a link for that?

Neither the P3, nor the P4, nor the Athlon process data in 6 or 8 bit chunks. Depending on what instruction they are executing, any of those chips will operation on a 16, 32, or 64 bit word (all bits in parallel). Additionally the P4 can handle a 128-bit interger through SSE2.

Your explaination does sound reasonable, it's just that the chips don't work that way.
 
Albigger said:
the barton core with 512 kb L2 Cache should help close this gap somewhat, no?

To me AMD doesn't benefit from cache as much as Intel does for example, the Morgan is basically low clocked speed XP's except without the copper core and 256KB L2 Cache but still pretty much would be equivilant to an XP at the same speed. The cache different is significant in size but in performance rarely means anything. In the SETI Bench the only thing stopping a Morgan from benching near an XP is the speed.

I'm still wondering whether the Bartons and Clawhammer are going to have SOI in them. All I basically heard was speculation so I'm not so certain on that. If it does have it, it will give AMD a huge upper hand.

Yodums
 
AMD needs to release something with a 4000 mhz clock/ 512 L2 cache and a FSB that has 2 instructions on both rising and falling action (intel "quad pumped" equivalent) before i buy another one.

I have an unlimited budget, so for me, Intel is the clear choice.

price/performance champ would be AMD currently.
 
Yodums said:


To me AMD doesn't benefit from cache as much as Intel does for example, the Morgan is basically low clocked speed XP's except without the copper core and 256KB L2 Cache but still pretty much would be equivilant to an XP at the same speed. The cache different is significant in size but in performance rarely means anything. In the SETI Bench the only thing stopping a Morgan from benching near an XP is the speed.

I'm still wondering whether the Bartons and Clawhammer are going to have SOI in them. All I basically heard was speculation so I'm not so certain on that. If it does have it, it will give AMD a huge upper hand.

Yodums

thanks for the explanation. but isn't SETI mostly number crunching, and not memory intensive? just wondering.
and i have read about SOI also, but it was a while ago. I'll have to go looking for some more articles.


runsalone said:

I have an unlimited budget

... all I can say is "must be nice" :)
 
Yodums said:
I'm still wondering whether the Bartons and Clawhammer are going to have SOI in them. All I basically heard was speculation so I'm not so certain on that. If it does have it, it will give AMD a huge upper hand.

I believe it's official that the Barton core WILL NOT use SOI, but Hammer will (if they can ever get the thing out).

Here's one story I found.

One news story doesn't prove it of course, but I recall Ed confirming it a few months ago.. It's too bad too, that would have been a sweet chip with no MB upgrade :(

originally posted by Albigger
thanks for the explanation. but isn't SETI mostly number crunching, and not memory intensive? just wondering.
and i have read about SOI also, but it was a while ago. I'll have to go looking for some more articles.

SETI loves memory bandwidth.. Folding however (usually, depending on what kind of protein & method of processing is being used) relies totally on FPS.
 
Gnerma said:

SETI loves memory bandwidth.. Folding however (usually, depending on what kind of protein & method of processing is being used) relies totally on FPS.

oh ok, i don't know because, well, ...

...yes its true, i don't use SETI.... :(

i used to do F@H, and i thought it was the same thing kinda. thanks for informing me.
 
OC-Master said:
Simple!

Clock for clock, the AMD system will always be faster than the Intel system because AMD's current line of CPUs can do 9 intructions per clock cycle where as Intel CPUs do only six intructions per clock cycle!

1 clock cycle = 1Hz 1000Hz = 1KHz 1000KHz = 1MHz 1000MHz = 1GHz


OC-Master

Note the "can do" part. It's not so easy to separate tasks so well. In most cases, an intel chip ends up doing maybe five, and an AMD does maybe seven.
That's still a solid advantage to AMD... But think about this: Clawhammer is supposed to separate these tasks better, so that AMD will do more like 8 - 9 per cycle. And then there's the whole 64-bit thing, which will make it another story altogether.
 
Anaxagoras1986 said:
Wonder why Intel changed the architecture...maybe it was necessary in order to give it the mhz headroom???

Although the increase to a 20 stage pipeline over the AMD CPUs's 10 did slow it down a bit Mhz for Mhz, it gave it a lot more stability at higher clock speeds.

The Pentium 3 pipeline would not be able to handle CPU speeds of over 2GHz, as it would be too unstable.

So Intel gambled on a 20 stage pipeline: early P4s sucked and and were very poor, but once they got the Mhz up they have the ignorance of a lot of computer buyers on their side, plus their chip can cope with 2.8GHz+ chips.

The Palomino core did 1.33Ghz up to 1.8Ghz. Then it was too unstable to release much faster, and so Tbred came into it.

However, the P4 Northwood core is so stable it can cope with lower speeds like 1.6GHz, all the way up to, and over 3GHz.

The stability of a 20 stage pipeline is begining to show. AMD have had to release three different cores for the 1.33GHz-2.13Ghz speed range. The Nortwood core has spanned over 1.2GHz so far, and will go further.

David
 
Gnerma said:


I believe it's official that the Barton core WILL NOT use SOI, but Hammer will (if they can ever get the thing out).

Yep, I think it's pretty much confirmed with the AMD Roadmap not having SOI Labeled on the Barton squares:

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_608,00.html

Albigger said:


thanks for the explanation. but isn't SETI mostly number crunching, and not memory intensive? just wondering.
and i have read about SOI also, but it was a while ago. I'll have to go looking for some more articles.

Gnerma is correct and that's why the P4 is better in the SETI area. They have that 3:4 divider and insanely high memory bandwidth.
 
Back