How are AMD 64's faster than P4's?

[mAsTAd]

I know they are faster in gaming because of its onboard mem. controller but how about in opening large programs and starting Windows? I don't get how P4's have higher clock speeds and A64's are faster. Thamks.

 

[araczynski]

more efficient designs (i.e. more efficient/intelligent use of resources/cache/pathways) akin to the new nvidias (slower clocks) when compared to the new ATIs (faster clocks).

its amazing how ingenuity steps up to the line when you're the underdog.

 

[3chordcharlie]

More clock speed doesn't always mean faster - A64 can perform more instructions per clock, so they 'catch' and in many cases pass the P4.

Think of the difference between a 2.4ghz P4 and a 3ghz celeron - the 'slower' chip is actually MUCH faster.

 

[FullRoast]

mAsTAd - I guess you got the short answer, more efficient. If you'd like to know more, there is some good information here. The article is specifically about Opteron's, but much of the internal architecture details still apply to the Athlon 64.

But, to confirm what has already been said:

The difference in pipeline architectures is what makes a clock-for-clock comparison between the Xeon and Opteron invalid (much like the Pentium 4 to Athlon XP comparison was invalid on a clock-for-clock basis). The Xeon's architecture allows it to reach high clock speeds at the expense of doing less work per clock cycle, the appropriate comparison ends up being one of cost and real-world performance, not one of clock speed.

 

[Algere]

Or i.e. if the Athlon 64 does 15 instructions/MHz aka IPC (mentioned above) while the Pentium 4 does 10 instructions/MHz. This is a roughl example of why the Athlon 64 does more than a Pentium 4 @ equivalent CPU frequencies.

Athlon 64
[ 2,000MHz(2GHz) x 15 instructions = 30,000 instructions ]

Pentium 4
[ 2,000MHz(2GHz) x 10 instructions = 20,000 instructions ]


Therefore it would take a Pentium 4 @ [ 3,000MHz(3GHz) x 10 = 30,000 instructions ] to perform as well as the Athlon 64 @ 2 GHz.

Obviously other factors come into affect such as FSB frequencies, cache, marchitecture, etc. which means you can't exactly say processor A is faster or equivalent to processor B on IPCs alone.

 

[Lonyo]

My petrol engine at 6,000 rpm is only as fast as my diesel at 3,000 rpm!
How!

There's a lot more than any figure of speed involved, be it rpm or cycles per second (Hz).
A CPU is basically an engine, and rpm, or bhp, aren't the be all and end all in terms of performance in an engine, neither are GHz the be all and end all in CPU's.

 

[Arsynic]

Clock speed alone doesn't make one processor "faster" than another. For example, a 50 foot Catepillar dump truck going 40mph can move the same amount of dirt faster than a Ferrari going 150mph. This is because CAT can transport the whole load in one trip while the Ferrari has to make 10 trips back and forth to do the same amount of work.

While the differences aren't as pronounced in the CPU industry as my illustration was, the concept still applies. The A64 just has a bigger "payload" than the current P4's. The way Intel advertises, you'd think they were selling CATs that go 150mph.

 

[jamesey]

ive used this analogy to explain the AMD speed to people

imagine 2 trucks with 100 boxes each. One is owned by Intel, the other by AMD. Each has a man that must move all the boxes out.
The Intel man can move his boxes 1 at a time. he can go in and out quickly, but he only does one at a time.
The AMD man can move his boxes 2 at a time. he's slower, but he gets more done per trip.
In the end, they will perform about the same.

I guess now with 64 bit, the analogy can be that the AMD man carries bigger boxes

 

[howdyduty]

LOL. Could be AMD just smells better.

Actually, I think Algere has the right idea. Nice analogy Algere.

 

[corkyg]

All these analogies and reason are good. Here's another simple one to ponder. The real reason is bandwidth. My analogy is we have a 16 bit CPU, and let's call that a single lane highway. Vehicles, no matter how fast they are, have to be in single file. Now, we go to a 32 bit structure, and we have a two lane highway where cars can go two breast. AMD has built a 4 lane highway (64 bit) and that allows four abreast traffic. Given 100 cars, going from A to B, which structure will get the most cars there the fastest? It's almost a no brainer - the 4-lane highway.

 

[DotheDamnTHing]

so why are p4s better at encoding (at least compared to amds before socket939)

 

[iwantanewcomputer]

Originally posted by: Lonyo
My petrol engine at 6,000 rpm is only as fast as my diesel at 3,000 rpm!
How!

There's a lot more than any figure of speed involved, be it rpm or cycles per second (Hz).
A CPU is basically an engine, and rpm, or bhp, aren't the be all and end all in terms of performance in an engine, neither are GHz the be all and end all in CPU's.

good analogy dude

 

[JBT]

Originally posted by: DotheDamnTHing
so why are p4s better at encoding (at least compared to amds before socket939)

Have use seen the 90nm 939 AMD bench's??? Intel is going down hill fast once AMD has these puppies out.

 

[DotheDamnTHing]

i wrote in my post that the p4s are better at encoding compared to amd BEFORE s939....and i would like to know why

 

[iwantanewcomputer]

Originally posted by: DotheDamnTHing
i wrote in my post that the p4s are better at encoding compared to amd BEFORE s939....and i would like to know why

higher clock speed mainly, but the new stuff on 939 outweighs it

 

[Wahsapa]

p4's have specific instructions for encoding which are superior to amds, however socket 939 changes all that. i think the p4's were better at enconding because of sse3(the new a64 cores just got this instruction set)? not sure tho.