G3/G4 on 90nm?

[MrCommunistGen]

Just an idea, probably already mentioned somewhere, but couldn't IBM or someone do a die shrink on a G3/G4 and then seriously bump up the clock speed so that Apple's low-end/portable market can get some better (faster) processors? Also, shouldn't Apple be using the G4s that support L3 Cache on the PBs instead of the one w/o L3?
God I love speculation!

-MCG

[OutThere]

IBM only makes the G3, which, until further notice, Apple has phased out. I assume that Motorola could work on their processors, but they have not had a good record in innovations for Apple. The G4s have been topped out at 1.42 for ages now, and if we see some G5 powerbooks soon I think that the G4s won't see any new big changes. The G3 though, I believe has some sort of future. I think that IBM could definitely crank the G3 up to 1.5+ Ghz, and slap on an Altivec/Vector Processing Unit of some sort, and turn the G3 from a 'dead' technology into something that would take the place of the current G4, and also finally rid Apple of Motorola. I personally think that the G3 has a future that people haven't considered yet. I think that the G4 is the truly dying technology.

[tpjunkie]

about a week or so ago, motorola announced 1.5 ghz G4s...apple has been known to utilize faster than announced chips, and its possible that they may use these newer g4s in the final revision of G4 powerbook, before switching to the G5.

[ingenious]

Originally Posted by tpjunkie

about a week or so ago, motorola announced 1.5 ghz G4s...apple has been known to utilize faster than announced chips, and its possible that they may use these newer g4s in the final revision of G4 powerbook, before switching to the G5.

that is just overclocking tho

[Catfish_Man]

Originally Posted by Calebj14

that is just overclocking tho

************. It's a new core. It also has new power management features and draws less power (which isn't possible for a simple overclock). Also, by definition, manufacturers cannot overclock, since they decide what the rated frequency is. The "overclocked G4" thing is either a myth, or an isolated case of Apple running G4s faster than they were rated (which I doubt, but is possible).

In response to the original post, I think if they bump the G3 to 90nm (I can't quite remember what the 750GX is on, but I think it's 130nm) they'll probably add some stuff to it, since it's already tiny. The quick and easy addition is a bigger cache. Other possibilities would be a second core, a memory controller, or possibly a vector unit. With a vector unit and an on chip memory controller a hypothetical 750xx chip could be a very nice mobile chip. With some tweaking they might hit 1.5GHz or so. With some serious redesign like the 7410->7450 transition it might get faster than that.

[DrBoar]

To get substantially above the current speed the G3/G4 would need to get to at least 2 GHz and perhaps have DDR as well. Will never ever happen, or so late that it will recieve as eager wellcome as the 68060 did, as we all remember

What we could hope for is a substantial price drop in the G4 CPUs and hence the G4 upgrades, but time is running out. A year from now all G4s will be outdated for many new CPU demanding games. The very slowest CPU I could find at DELL is a 2.6 GHz Celeron and outside AltiVec and SMP that one will outrace any G4 there is.

I expect that my 1.2 GHz G4 will fail in new games within a year, time to move on

[leftbanke7]

From a marketing standpoint, I think it would be a decent enough of an idea to modernize the G3 and give it some fancy pants name telling the public it's a mobile technology chip (as Intel is doing). Calling it a G3 would make it seem too old and calling it a G4 would give it all the negative connotations that the MotoSlowa chips have. Possible names?

iChip
PPCMobile

Damn, I can only think of 2. Any other suggestions?

[bousozoku]

It would have been a good idea to manufacture the G4 on a 90 nm process but it's useless to take an (poor) old design and spend more money than it's worth to modernise things.

If IBM decides to add the VMX vector unit to the G3, building it with a 90 nm process would be good and give them room to keep the floating point stronger than the current G4s of equal clock speed. However, it's not as if they make big money from the G3 line as is but compactness and great heat dissipation keep it active.