Question About G4 Processor???

[jumpman25]

It seems like almost everyone is complaining about how the G4 processor is out of date, and how apple needs to get a G5 into the powerbooks or people will stop buying them. I thought the G4 was a good processor. I read an article a while back stating that a G4 800MHz processor was faster than a P4 1.7GHz. This would mean that the 1.5GHz processors in the current powerbooks would be about the same as a 3.0+ P4 processor (not bad for a notebooks). The current PentiumM dothan chips are getting a lot of buzz lateley and a 2.0 dothan is said to be about the same as a 3.0 P4. I may be missing something here, but it looks to me like people are forgetting that the powerbooks are laptops, and people are expecting top of the line desktop performance out of them. From what I have seen, it looks like the top of the line G4 is right up there with the top of the line PentiumM. I'm interested in hearing what you people have to say on the subject. Are people making a big deal where none should be made? I'm wondering if the powerbooks were advancing so fast in the previous years that becuase they are slowing down or releasing smaller speed bumps, people feel that they are slow.

[maya]

Originally Posted by jumpman25

It seems like almost everyone is complaining about how the G4 processor is out of date, and how apple needs to get a G5 into the powerbooks or people will stop buying them. I thought the G4 was a good processor. I read an article a while back stating that a G4 800MHz processor was faster than a P4 1.7GHz. This would mean that the 1.5GHz processors in the current powerbooks would be about the same as a 3.0+ P4 processor (not bad for a notebooks). The current PentiumM dothan chips are getting a lot of buzz lateley and a 2.0 dothan is said to be about the same as a 3.0 P4. I may be missing something here, but it looks to me like people are forgetting that the powerbooks are laptops, and people are expecting top of the line desktop performance out of them. From what I have seen, it looks like the top of the line G4 is right up there with the top of the line PentiumM. I'm interested in hearing what you people have to say on the subject. Are people making a big deal where none should be made? I'm wondering if the powerbooks were advancing so fast in the previous years that becuase they are slowing down or releasing smaller speed bumps, people feel that they are slow.

When MR members here say that the G4 suck we do not mean clock speed we are talking about the overall package. Most important the bus speed.

G5 chips has a much faster and larger FSB when compared to the G4 chip, if the G4 chips has as good of a bus as the G5 chips, people here would not be complaining.

intel chips have a faster FSB when compared to the G4, though they chips are crippled in other ways.


Does this sort out your question.

[jumpman25]

Originally Posted by maya

When MR members here say that the G4 suck we do not mean clock speed we are talking about the overall package. Most important the bus speed.

G5 chips has a much faster and larger FSB when compared to the G4 chip, if the G4 chips has as good of a bus as the G5 chips, people here would not be complaining.

intel chips have a faster FSB when compared to the G4, though they chips are crippled in other ways.


Does this sort out your question.

Thanks for the quick reply! Yes, this helps out. Although I thought the PentiumM ran on a 200MHz FSB verses the G4's 167MHz. Doesn't seem like that big of a difference. So I guess people are just comparing the G4 to the G5 rather than to the PentiumM, as there does not seem to be that much of a difference speed wise between the G4 and the PentiumM???

[neilrobinson]

Originally Posted by maya

intel chips have a faster FSB when compared to the G4, though they chips are crippled in other ways.

hmm they dont understand PPC instructions?

i aggre that their is too much G4 bashing going on! its not all about clock speed etc. my G4 400 is faster than my P3 733 for web browsing, email and everyday tasks like that, when i comes to movie rendering, large photoshop files and quark it does lag a little but we are talking about 1991 here! the prod is that a lot of people are using their portables for MAJOR tasks which maybe their where not quite build for.

[ddtlm]

jumpman25:

I read an article a while back stating that a G4 800MHz processor was faster than a P4 1.7GHz.

Depends entirely on what your test is. In general, however, the P4 is faster.

This would mean that the 1.5GHz processors in the current powerbooks would be about the same as a 3.0+ P4 processor (not bad for a notebooks).

Bad assumptions there. Consider: the 1.7ghz P4 was a generation 1 core, with 256k L2, and a 400mhz FSB. The 800mhz G4 was a generation 1 "long pipe" G4, with 256k L2, 2MB L3, and a 133mhz FSB. Time passes. The 3ghz P4 is a generation 2 or 3 core, has a 800mhz FSB, and 512k or 1MB of L2. The 1.5ghz G4 is still the original core (simply made smaller), with 166mhz FSB, and 512k of L2. (No L3.) ...pause... So the P4 nearly doubled clock, it doubled FSB, doubled or quadrupled L2, and went through 1 or 2 new cores. The G4 also nearly doubled its clock, increased FSB by only 25%, doubled its L2, lost its L3, and didn't change design.

Intel and AMD change and evolve their products. Moto... sometimes releases a die shrink. Probably sometime soon they'll have G4's with on-die memory controllers and whatnot, which is great, but the core will probably be exactly the same as it has been since Apple's first 733mhz G4 (Moto 7450), which was introduced in January 2001.

Seriously, the best G4 you can buy today is somewhat less than twice as powerful as one you could have gotten 4 years ago.

Although I thought the PentiumM ran on a 200MHz FSB verses the G4's 167MHz.

The P-M has a 100mhz QDR bus, do a 400mhz data rate. It'll be moved up to 533mhz soon enough (133mhz QDR).

The big weakness of the G4 vs the P-M are:
1) FSB throughput: years behind.
2) L2 cache: new P-M's have 2MB on die.
3) Double-precision floating point: AltiVec can't help here.

[SiliconAddict]

Originally Posted by ddtlm

jumpman25:
Bad assumptions there. Consider: the 1.7ghz P4 was a generation 1 core, with 256k L2, and a 400mhz FSB. The 800mhz G4 was a generation 1 "long pipe" G4, with 256k L2, 2MB L3, and a 133mhz FSB. Time passes.

[yada yada yada]

Very succinctly put. Actually that was one of the least convoluted comparisons I've ever seen

[Mav451]

I have a few questions:

1) If Pentium M's have double precision, what exactly do the G4's have? Single precision?

2) And does the amount of precision have any sort of bearing on the whole FLOP argument?

[jumpman25]

Originally Posted by ddtlm

jumpman25:


Depends entirely on what your test is. In general, however, the P4 is faster.


Bad assumptions there. Consider: the 1.7ghz P4 was a generation 1 core, with 256k L2, and a 400mhz FSB. The 800mhz G4 was a generation 1 "long pipe" G4, with 256k L2, 2MB L3, and a 133mhz FSB. Time passes. The 3ghz P4 is a generation 2 or 3 core, has a 800mhz FSB, and 512k or 1MB of L2. The 1.5ghz G4 is still the original core (simply made smaller), with 166mhz FSB, and 512k of L2. (No L3.) ...pause... So the P4 nearly doubled clock, it doubled FSB, doubled or quadrupled L2, and went through 1 or 2 new cores. The G4 also nearly doubled its clock, increased FSB by only 25%, doubled its L2, lost its L3, and didn't change design.

Intel and AMD change and evolve their products. Moto... sometimes releases a die shrink. Probably sometime soon they'll have G4's with on-die memory controllers and whatnot, which is great, but the core will probably be exactly the same as it has been since Apple's first 733mhz G4 (Moto 7450), which was introduced in January 2001.

Seriously, the best G4 you can buy today is somewhat less than twice as powerful as one you could have gotten 4 years ago.


The P-M has a 100mhz QDR bus, do a 400mhz data rate. It'll be moved up to 533mhz soon enough (133mhz QDR).

The big weakness of the G4 vs the P-M are:
1) FSB throughput: years behind.
2) L2 cache: new P-M's have 2MB on die.
3) Double-precision floating point: AltiVec can't help here.

So where would you say the G4 stands compared to the PentiumM??? Also, you say the G4 has long pipes. I thought that it had shorter pipelines and this was why it was more efficient than a pentium which has longer pipelines. I know that the G4's are a little behind in the FSB department, but I would think if they were that incredibly slow, apple would have been forced to stop production on the powerbooks as nobody would buy them. I know OSX is very good at memory management, and thus runs much smoother than windows, and the whole virus/spyware thing, I can see how this helps keep the powerbooks going (not to mention they look nice :-)), but speed is important, and considering how fast technology changes, if they have changed that little in the last 4 years, I would think they would be long gone. I just feel like I am missing something here.

P.S. I thought freescale made the apple processors. Is freescale part of moto or something???

[ddtlm]

Mav451:

If Pentium M's have double precision, what exactly do the G4's have? Single precision?

They both support both single and double. The thing is, AltiVec only supports single precision floating point math, so the G4 looses its strongest asset and has to rely on its scalar floating point unit. (Remember the G5 has two scalar FP units, not one.)

And does the amount of precision have any sort of bearing on the whole FLOP argument?

Not really. Single precision is typically gona be a higher flop score.

jumpman25:

So where would you say the G4 stands compared to the PentiumM?

I'd say the P-M is a better chip, but if Apple ever ships that upcoming G4 with on-die memory controllers, that will go a long ways towards closing the gap.

Also, you say the G4 has long pipes.

There is a long-pipe and a short-pipe G4, the short pipe one being the one closley related to the G3, and like the G3, it is no longer offered in an Apple product. Both designs are pretty short compared to G5's, Athlons, or most anything else.

I know that the G4's are a little behind in the FSB department, but I would think if they were that incredibly slow, apple would have been forced to stop production on the powerbooks as nobody would buy them.

FSB bandwidth is less important than a lot of people think. Its a lot like the "megahertz myth" with CPU clocks. Some generalizations follow: 200mhz DDR (100mhz x2) wasn't any faster than PC133. 533mhz DDR2 (133mhz x4) really isn't any faster than DDR400. A P4 at 1.4ghz wasn't any faster than a P3 at 1.0ghz. You get the pattern: when technology is redesigned for higher clockspeeds, often more steps are added so that latency (access to memory, or an instruction through the processor) isn't lowered much. Why? Because its bounded by the speed electrons can move.

So, the 166mhz FSB on a G4 is kinda lame, but at the same time it can still get small things from memory more-or-less as fast as a G5's 1.25ghz FSB, because in both cases the limiting factor is more-or-less how fast the electrons can move.

This is not to say that fast FSB's are a waste. A fast FSB could be compared to a train, capabile of hauling huge amounts of stuff at once, but only in the case where all the stuff is well-ordered. A fast FSB is great for streaming video, as an example. However, if your dealing with random data, or not very much data, the old G4 FSB is fine.

Is freescale part of moto or something?

Yeah, Moto spun off Freescale. Old timers like me still say "Moto".

[jumpman25]

Originally Posted by ddtlm

Mav451:


They both support both single and double. The thing is, AltiVec only supports single precision floating point math, so the G4 looses its strongest asset and has to rely on its scalar floating point unit. (Remember the G5 has two scalar FP units, not one.)


Not really. Single precision is typically gona be a higher flop score.

jumpman25:


I'd say the P-M is a better chip, but if Apple ever ships that upcoming G4 with on-die memory controllers, that will go a long ways towards closing the gap.


There is a long-pipe and a short-pipe G4, the short pipe one being the one closley related to the G3, and like the G3, it is no longer offered in an Apple product. Both designs are pretty short compared to G5's, Athlons, or most anything else.


FSB bandwidth is less important than a lot of people think. Its a lot like the "megahertz myth" with CPU clocks. Some generalizations follow: 200mhz DDR (100mhz x2) wasn't any faster than PC133. 533mhz DDR2 (133mhz x4) really isn't any faster than DDR400. A P4 at 1.4ghz wasn't any faster than a P3 at 1.0ghz. You get the pattern: when technology is redesigned for higher clockspeeds, often more steps are added so that latency (access to memory, or an instruction through the processor) isn't lowered much. Why? Because its bounded by the speed electrons can move.

So, the 166mhz FSB on a G4 is kinda lame, but at the same time it can still get small things from memory more-or-less as fast as a G5's 1.25ghz FSB, because in both cases the limiting factor is more-or-less how fast the electrons can move.

This is not to say that fast FSB's are a waste. A fast FSB could be compared to a train, capabile of hauling huge amounts of stuff at once, but only in the case where all the stuff is well-ordered. A fast FSB is great for streaming video, as an example. However, if your dealing with random data, or not very much data, the old G4 FSB is fine.


Yeah, Moto spun off Freescale. Old timers like me still say "Moto".

Thanks for taking the time to explain all that.