Why not .13??
How come motorola has not put a .13micron process into their processors? If they could do that their processors would be running circles around Intels and AMD's processors. They could reach over 3GHz with a little advancement. What do you guys think?? <img src="confused.gif" border="0">
[ 02-27-2002: Message edited by: Master ]</p>
[ 02-27-2002: Message edited by: Master ]</p>
Comments
I think Apple will place a higher priority on increasing bandwidth and reducing latency on the motherboard (including the frontside bus). That will do wonders for performance even with the current processors.
I'm much more fond of Motorola's approach to processor design than Intel's, and marketing concerns notwithstanding I hope they stick to their approach. Elegant, cool-running, efficient processors give Apple a lot more options, both in terms of industrial design and in terms of software design.
[ 02-27-2002: Message edited by: Amorph ]
[ 02-27-2002: Message edited by: Amorph ]</p>
[ 02-27-2002: Message edited by: Amorph ]</p>
<strong>I'm much more fond of Motorola's approach to processor design than Intel's, and marketing concerns notwithstanding I hope they stick to their approach. Elegant, cool-running, efficient processors give Apple a lot more options, both in terms of industrial design and in terms of software design.</strong><hr></blockquote>And of course moving to a smaller fab process would be entirely consistent with cool-running, efficient processors.
I've heard rumors that they'll move to .15 next? Who knows?
The 7455 scales to 1.8GHz? Sweet.
I sincerely doubt that most 1 GHz G4s can really be pushed to 1.2 GHz or 1.33 GHz. Mine are probably running 70 C or hotter under full load.
Two 1 GHz G4s almost eat up as much power as a single 2.2 GHz P4.
I think that 1100 MHz part is slightly different from the 7455s in my desktop though.
Of course. The difference is 100MHz.
See, if you have a reasonably compact instruction pipeline you don't need a lot of those optimizations.
As far as bandwidth goes, Mot did roll out RapidIO, and it's a no-brainer that their CPUs will be supporting it soon.
<strong>How come motorola has not put a .13micron process into their processors? If they could do that their processors would be running circles around Intels and AMD's processors. They could reach over 3GHz with a little advancement. What do you guys think??
</strong><hr></blockquote>
Yeah and while they are at it why don't they flip that magic switch and go straight to .09 micron or something even smaller. There are technical issues involved here. You can't imagine the complexities in both product design and process operation to move to a smaller geometry. Tools cost money, engineering takes time.
<strong>I was wondering when Eskimo would come in and inject some reality into this thread... </strong><hr></blockquote>
So was I.
According to most of the hype-mongers on this board, the magic switch is going to be used to produce 2GHz G5 procs by this summer. <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
"...You can't imagine the complexities in both product design and process operation to move to a smaller geometry. Tools cost money, engineering takes time"
well, I surley can... and Eskimo is right... Even if no architecture changes are made, a process shrink is a LOT of work... and if your going to shrink the process, it makes sense to improve the architecture while redoing a lot of previous work.... which compounds things...
on another note, the G4 is kind of stale in architectural advances... Intel's current performance is good, and for a company that has created probably the worst abomination of a architecture ever, they are putting some very good new ideas about architecture into their newer and future chips. The PPC line of processors needs much more than DDR ram, rapid I/O, a die shrink, and larger caches to be at the cutting edge of design. Multi-threading, higher memory bandwidth, and architecture that can further exploit ILP are all on the plate for Intel in the very near future... I don't see what Motorola's response could be. AltiVec is one of the finer implimentations of SIMD in a register, however, I am not entirely convinced the transistors coundn't have been put to better use... IBM has good architecture in the Power4 - for today, but just try to price one out... I'm not saying that Apple is dead, or that PPC is either - I'm not even saying they are less than adequate now, or in the future. However, things in the microprocessor world can move quickly, and I don't forsee Motorola catching up, or keeping up... I just hope we all have long happy lives, so we can see what happens...
They'll probably introduce those technologies when they become relevant.
See, if you have a reasonably compact instruction pipeline you don't need a lot of those optimizations."
hmm... a well designed/simple architecture is not a substitute for technological advancement. The things I describe - that motorola is not acting on - ARE optimizations, and DO make things execute in shorter time. Oh, and mistake no mistake about it - Those technologies became relevant the moment it was shown that they shorten execution times... they are already relavant.
Reguarding the wild card that Motorola has some breakthrough architecture under raps... and it will be unleashed on us the day apple ships an box - forget it. It takes many years to incorporate some of the things I described into a design - motorola hasn't even mentioned them once. If you guys are interested - I would highly recommend both 'computer architecture a quantitave approach' and 'computer orginization and design' as well as the technical papers on upcoming IA64 architectures. Motorola is not being remotely revolutionary, and only barely evolutionary... I hope their simplicity/lack of new ideas/beaten path approach benifits them in some unforseen way.