AMD Wins "Best CPU Manufacturer Award"

[Duvie]

Originally posted by: nserra
You are all crazy.

Who says Mhz doesn't matter is a moron. It does matter!!!
What you can?t do is to compare two different Processors at the same Mhz level. It?s not right. AMD64 vs AMD XP vs PentiumM vs Pentium4 at 2.0Ghz is it right?
And now Pentium4 3.2Ghz vs Pentium4 3.8, the 3.2 is faster at what? The only possibility of a slower Ghz be faster than the higher one is at the board level (Single channel, dual channel, ?)

the prescott at the lower 2.8-3.2ghz range is often times slower then nortwood by more then 1%...the fact is a faster cpu can be slower at the cpu architecture level...It has to do with the 31 stage pipeline of the prescott versus the 22 stage pipeline of the northwood. ...Another example where the pipelibne (longer reversed progress was when P4 first migrated from the P3 1.1ghz to the P4 1.3-1.4ghz willamettes.....The pipeline I believed was doubled....The newer Dothan chips are like going back to that shorter pipeline and look at how a 2.0ghz Dothan wipes the florr with chips 1ghz plus more in speed....relook at your argument...

I may have interpreted what you were saying wrong, and if I did apologize...

 

[apairoflegsanyone]

Originally posted by: Duvie

Originally posted by: nserra
You are all crazy.

Who says Mhz doesn't matter is a moron. It does matter!!!
What you can?t do is to compare two different Processors at the same Mhz level. It?s not right. AMD64 vs AMD XP vs PentiumM vs Pentium4 at 2.0Ghz is it right?
And now Pentium4 3.2Ghz vs Pentium4 3.8, the 3.2 is faster at what? The only possibility of a slower Ghz be faster than the higher one is at the board level (Single channel, dual channel, ?)

the prescott at the lower 2.8-3.2ghz range is often times slower then nortwood by more then 1%...the fact is a faster cpu can be slower at the cpu architecture level...It has to do with the 31 stage pipeline of the prescott versus the 22 stage pipeline of the northwood. ...Another example where the pipelibne (longer reversed progress was when P4 first migrated from the P3 1.1ghz to the P4 1.3-1.4ghz willamettes.....The pipeline I believed was doubled....The newer Dothan chips are like going back to that shorter pipeline and look at how a 2.0ghz Dothan wipes the florr with chips 1ghz plus more in speed....relook at your argument...

I may have interpreted what you were saying wrong, and if I did apologize...

I think what the poster is trying to say is that despite the architecture improvements, the p4 needs the mhz increase to become a beast because it was design for that since day 1. So basically, mhz does matter in this case.

 

[Zebo]

Think Zebo really ment was that Prescott was slower in some apps then

No it's 3x slower in the modeling we do, It's Cusom made proggy and it takes literally 3x longer to chrun out results. Same goes when I compress 20 gigs of data w/ winrar. I'm not talking about these Intel designed or bankrolled benchmarks the review sites use but heavy duty math. It also has 3x slower mem latency you can see all the time as A64's.

Vee could chime in why this is so better than me, but the bottom line is intel P4 does well in apps written for it's optimizations and not much good in anything else. Dothan is just the opposite.

 

[Trikat]

:thumbsup:

 

[TuxDave]

Originally posted by: Zebo

Think Zebo really ment was that Prescott was slower in some apps then

No it's 3x slower in the modeling we do, It's Cusom made proggy and it takes literally 3x longer to chrun out results. Same goes when I compress 20 gigs of data w/ winrar. I'm not talking about these Intel designed or bankrolled benchmarks the review sites use but heavy duty math. It also has 3x slower mem latency you can see all the time as A64's.

Vee could chime in why this is so better than me, but the bottom line is intel P4 does well in apps written for it's optimizations and not much good in anything else. Dothan is just the opposite.

If a programmer chooses not to use the optimizations in their programs, then is it the fault of the processor? Essentially, if I build in a microcode that let's you compute a dot product of large vectors in one cycle but you choose to iterate it yourself, who's fault is that?

 

[Redstorm]

Originally posted by: nserra
You are all crazy.

Who says Mhz doesn't matter is a moron. It does matter!!!
What you can?t do is to compare two different Processors at the same Mhz level. It?s not right. AMD64 vs AMD XP vs PentiumM vs Pentium4 at 2.0Ghz is it right?
And now Pentium4 3.2Ghz vs Pentium4 3.8, the 3.2 is faster at what? The only possibility of a slower Ghz be faster than the higher one is at the board level (Single channel, dual channel, ?)

I'm not saying your right and im not saying your wrong but...

For any given processor architecture increasing the frequency increases performance this is why we over clock our boxes to get better performance, you cannot however use frequency as a measure of performance when comparing differing architectures.

When you are comparing different processor architectures its Instructions per clock that you should be looking at to measure performance.

Instructions per clock is why an Athlon 64 at 2.4GHz beats a Pentium 4 at 3.6GHz, if the Athlon was able to run at 3.6GHz Intel would be out of business as it would blow everything Intel has out of the water.

AMD deserve the award just for integrating the memory controller onto the chip thus eliminating the FSB bottle neck that plaques Intel solutions to this day.

Look at the quad server example using Opterons and Xeons. Every time you add a Xeon it has to share the FSB with the other CPU's as you add CPU's the situation keeps getting worse all have to share the same FSB (Bottle neck). However with an Opteron system every time you add an Opteron you get another memory controller and bank of memory thus increasing the bandwidth available to the CPU's. Simply put Opterons scale better.

The sooner Intel admit defeat and follow AMD's lead the better for all of us, otherwise Intel will keep falling further and further behind in the performance stakes.

 

[clarkey01]

Originally posted by: TuxDave

Originally posted by: Zebo

Think Zebo really ment was that Prescott was slower in some apps then

No it's 3x slower in the modeling we do, It's Cusom made proggy and it takes literally 3x longer to chrun out results. Same goes when I compress 20 gigs of data w/ winrar. I'm not talking about these Intel designed or bankrolled benchmarks the review sites use but heavy duty math. It also has 3x slower mem latency you can see all the time as A64's.

Vee could chime in why this is so better than me, but the bottom line is intel P4 does well in apps written for it's optimizations and not much good in anything else. Dothan is just the opposite.

If a programmer chooses not to use the optimizations in their programs, then is it the fault of the processor? Essentially, if I build in a microcode that let's you compute a dot product of large vectors in one cycle but you choose to iterate it yourself, who's fault is that?

Its more "These benchmarks are made around Chip A more then Chip B.. And Chip B still wins"

And when chip B wins, even by slender margins, you still get people go " ah but its only 3% faster". Ah well.