The PIII Intel's Secret Weapon?!

[topgunn]

I found this article over at Tom's to be a great read. Intel's future may not be as dim as it had begun to look over the last year and a half.
http://www.tomshardware.com/cpu/20050525/index.html

 

[Electric Monk]

AFAIK Intel deliberately removed half (or more, it's been a while) of cache from the last PIII release because it was performing too well against their new favorite PIV. I believe the 1.5 GHz PIII was showing up a 2.0+ GHz PIV.

It's nice to see the better architecture resurrected in the Pentium M.

The other cute thing is that IBM and Intel have switched places. Intel is using the cycle efficient P-M while IBM has the high clocked Cell. Of course Cell is way better then Netburst ever was, but still…

 

[mgargan1]

Well, Tom's Hardware did a guide a while back on the 1.26GHz PIII "tualatin". Which was the latest core of that processer. It was based on the 130nm design and had 512k of cache. The 1.26GHz PIII was on par with a 1.8GHz P4. So a 1.4 (the fastest clockspeed) is gonna be on par with a 2.1GHz P4 with 256k of cache.

 

[Platform]

What is it with CPU's......for every new generation that comes they become less sufficient.....

G4 clock = better than G5 at same clock speed.
PIII clock = better than P4 at same clock speed. ​

 

[caveman_uk]

What is it with CPU's......for every new generation that comes they become less sufficient.....

G4 clock = better than G5 at same clock speed.
PIII clock = better than P4 at same clock speed. ​

Maybe because the marketing people have got everyone believing that the faster the processor clock speed the faster it will be so the engineers make processors that can run at faster clockspeeds regardless of how efficiently they do it.

And of course a P4 must be better than a P3 as it's got a 4 in it

 

[Platform]

Maybe because the marketing people have got everyone believing that the faster the processor clock speed the faster it will be so the engineers make processors that can run at faster clockspeeds regardless of how efficiently they do it.

And of course a P4 must be better than a P3 as it's got a 4 in it

Well Intel yes.........but apple too

 

[caveman_uk]

To be honest whilst the G4 and G5 are comparable at the same clock speed, I very much doubt that we'd be up at 2.7GHz if we'd waited for Motorola/Freescale to get the G4 there. The current G4's only reach the speeds they do because they have no L3 cache and that is what made the last of the G4 powermacs so competitive....

There's a lot of people thinking that a G5 powerbook will automatically be much better than a G4 one. Unless there's a real difference in clock speeds I think people are going to be disappointed.

 

[Pedro Estarque]

G4 clock = better than G5 at same clock speed.
​

This is simply not true. The G4 is a nice chip ( I use them every day ) but the G5 has a much better Floating Point unit, it is capable of doing more instructions per cycle and is not crippled by a 167 bus. A 1Ghz G5 would blow my 1Ghz G4. check this out Unless by "better" you mean cooler, and even this would be questionable. We have never seen a 1.5 Ghz G5

The P4 was an architecture intended to clock really high with bigger pipelines and a nice branch prediction ( above 4Ghz ) and that high frequency would compensate for its design "failures". However, it didn't scale as nice as intel imagined.

 

[caveman_uk]

This is simply not true. The G4 is a nice chip ( I use them every day ) but the G5 has a much better Floating Point unit, it is capable of doing more instructions per cycle and is not crippled by a 167 bus. A 1Ghz G5 would blow my 1Ghz G4. check this out Unless by "better" you mean cooler, and even this would be questionable. We have never seen a 1.5 Ghz G5

Interestingly on that chart, I see the results for the Dual 1.4G4s as very competitive for the dual G5 2.0s (chips running with a much higher FSB and 50% higher clock speed). The only time the G4s really suffer is on the graphics intensive tests which says a lot more about the AGP system than it does the CPU core.

I would take a dual G4 1.4 with 2MB L3 cache over anything less than a dual G5 2Ghz any day. Those dual G4s are a lot cheaper as well as a more flexible form factor.

 

[Platform]

This is simply not true. The G4 is a nice chip ( I use them every day ) but the G5 has a much better Floating Point unit, it is capable of doing more instructions per cycle and is not crippled by a 167 bus. A 1Ghz G5 would blow my 1Ghz G4. check this out Unless by "better" you mean cooler, and even this would be questionable. We have never seen a 1.5 Ghz G5

The P4 was an architecture intended to clock really high with bigger pipelines and a nice branch prediction ( above 4Ghz ) and that high frequency would compensate for its design "failures". However, it didn't scale as nice as intel imagined.

I know that....just that the actual performac you showed was not at the same clock speeds....[upgraded] And in matter of effieiency the G4 IS better But in BIG caclulations and thing the G5 is better

 

[Pedro Estarque]

Interestingly on that chart, I see the results for the Dual 1.4G4s as very competitive for the dual G5 2.0s (chips running with a much higher FSB and 50% higher clock speed). The only time the G4s really suffer is on the graphics intensive tests which says a lot more about the AGP system than it does the CPU core.

I would take a dual G4 1.4 with 2MB L3 cache over anything less than a dual G5 2Ghz any day. Those dual G4s are a lot cheaper as well as a more flexible form factor.

The dual 1.4 G4 is very close to the dual 2.0 G5 in Photoshop because it's one app that uses AltiVec as heavily as possible, one thing that the G5 always sucked at. Its altivec implementation was a last minute add on. And of corse, one massive L3 cache to compensate for the embarrassing 167Mhz bus.
You would get similar results with any highly Altivec optimized apps such as BLAST. Unfortunately these aren't the majority of them.

 

[superbovine]

To be honest whilst the G4 and G5 are comparable at the same clock speed, I very much doubt that we'd be up at 2.7GHz if we'd waited for Motorola/Freescale to get the G4 there. The current G4's only reach the speeds they do because they have no L3 cache and that is what made the last of the G4 powermacs so competitive....

There's a lot of people thinking that a G5 powerbook will automatically be much better than a G4 one. Unless there's a real difference in clock speeds I think people are going to be disappointed.

why make a faster one that cost more when you can sell a slower one for the same price?

I suspect you are going to correct, once the benchmarks come out. i think the big test will be video encoding and audio encoding with freescale.

 

[Xapplimatic]

What is it with CPU's......for every new generation that comes they become less sufficient.....

G4 clock = better than G5 at same clock speed.
PIII clock = better than P4 at same clock speed. ​

I remember the embarassment of Intel that P3s were faster than P4s, given..

But a G5 slower than a gee-4.. um.. NO.. remember 64-bit versus 32.
To start with that's twice the bandwidth at any given same clockspeed, not to mention the buss is multiples of times faster.. You can engineer a test to advantage a G4 in one area, but you can't fool me and say that any 32-bit slug-speed FSB G4 is gonna give anything even close to similar overall system speed than a Hyper Transport-equipped 64-bit G5.. It just isn't possible. The G5 will kick the G4's buss every time.. (and I don't even have one yet ;(

 

[superbovine]

I remember the embarassment of Intel that P3s were faster than P4s, given..

But a G5 slower than a gee-4.. um.. NO.. remember 64-bit versus 32.
To start with that's twice the bandwidth at any given same clockspeed, not to mention the buss is multiples of times faster.. You can engineer a test to advantage a G4 in one area, but you can't fool me and say that any 32-bit slug-speed FSB G4 is gonna give anything even close to similar overall system speed than a Hyper Transport-equipped 64-bit G5.. It just isn't possible. The G5 will kick the G4's buss every time.. (and I don't even have one yet ;(

http://developer.apple.com/document...ing/index.html#//apple_ref/doc/uid/TP40001064

Myth #4:
Myth: My application will have much faster performance if it is a “native” 64-bit application.
Fact: This is true for some other architectures because the number of registers and the width of registers changes between 32-bit and 64-bit mode. However, the PowerPC architecture does not have either of these limitations. It was designed for 64-bit computing from the beginning, and supports 64-bit arithmetic instructions in 32-bit mode. Thus, on PowerPC architectures, software does not generally become faster (and may actually slow down) when compiled as a 64-bit executable.

 

[Platform]

I remember the embarassment of Intel that P3s were faster than P4s, given..

But a G5 slower than a gee-4.. um.. NO.. remember 64-bit versus 32.
To start with that's twice the bandwidth at any given same clockspeed, not to mention the buss is multiples of times faster.. You can engineer a test to advantage a G4 in one area, but you can't fool me and say that any 32-bit slug-speed FSB G4 is gonna give anything even close to similar overall system speed than a Hyper Transport-equipped 64-bit G5.. It just isn't possible. The G5 will kick the G4's buss every time.. (and I don't even have one yet ;(

Ok, but we can't even use 64Bit today so that is nearly irrelevant

But as I said in some apps that require a lot of CPU power the G5 wind because of the FSB, but normal apps the G4 can be just as fast if not faster...don't know where but I saw somewhere the 1.67 beating the SP 1.8 PM