Huh?
Snowy_River said:
While I don't dispute the fact that Moto has pushed the speed of the G4 up, they haven't done as good a job pushing the processing power up.
The G4 is essentially in the performance class of the Pentium III processor. So why hasn't Motorola kept up with Intel's chip performance when Apple has had continually shrinking marketshare and Motorola has gotten two or three dollars from Apple's market while Intel has received eighty dollars from the Windows side? That's what you get when marketshare is overwhelming on the other side.
That does not detract from the fact that Motorola has made a lot of improvements to the G4 far beyond simply Mhz upgrades. It's just that the company has not had the financial means to come up with an entirely new chip architecture exclusively for Apple's measly desktop marketshare.
Snowy_River said:
One of the main applications that I use, when the G4 first came out, was as fast or faster than the same application under Windows on the top end Intel or AMD processors.
That was before Intel came out with the Pentium 4. When the Pentium 4 first came out, Apple's sales literally plummeted and the stock price fell off a cliff like Wylie Coyote.
Snowy_River said:
Now, if we compare how the speed boosts, that you seem to be arguing exonerate Moto, stack up against comparable Intel machines (note that neither the Intel nor the G4 are absolutely the top of the line), and how the G5 fits in too:
Another example of how more money brings faster improvements to processors is by comparing Itanium to the Power4 chip. When the Itanium first came out it's performance was soundly trounced by the Power4. Now, the Itanium 2 is speeding past the Power4 like a rocket. Why? It's due to Intel having very deep pockets and deciding that they could lose money at that market for ten years if necessary. IBM is simply not keeping up with the pace of changes Intel is making to Itanium. Intel has moved Itanium well in front of the Power4 on SPECint and SPECfp with the Power4 running at 1.7 GHz and Itanium at 1.5 GHz with a 6 MB onboard L3 cache. Intel will up that to 1.7 GHz and 9 MB of L3 cache in the next few months, while IBM will move to the dual-core Power5. In 2005 Itanium will also have a dual core with 24 MB of L3 cache and in 2006 there will be multiple cores on one chip. IBM is simple running as fast as they can, but they seem to be on a treadmill not making much headway against Itanium.
Snowy_River said:
Gee, it looks like, despite Intel's failure to reach 3.9GHz, they are easily beating the G4.
The Pentium 4 has beaten the G4 from just about the beginning. That's again mainly due to Intel being able to design a chip specifically for one market, the desktop. Motorola could not do that and expect to make money at it, because nobody but Apple with their shrinking marketshare would buy it.
As for IBM being able to make a processor just for Apple's desktop market...we'll they don't have the wherewithall to do that either. The 970 is simply a cutdown version of the Power4 with Altivac tacked on.
Snowy_River said:
So, clock speed really doesn't tell it all. While I realize that this is a single application, it is, also, more or less a 'real world' test, not a contrived benchmark. Also, in case you were wondering, VectorWorks is not MP aware, so any boost that the dual 2GHz G5 got over the single 1.8GHz G5 above simple clockspeed was due to the system load being placed on the second processor (same with the dual vs. single PCs).
The G4 doesn't have to be as fast as the topend Pentium chips, that's what the G5 is for. With some changes to the I/O of the G4 and a move to a smaller process size, the G4 should be quite competitive with the upcoming Pentium-M Banias chip. Or do you seriously believe that the G5 can cover the entire desktop and notebook market for Apple? If there is another IBM chip design for Apple's notebook market, then where is it? People were speculating that IBM would take over with a update to the G3 and Apple would throw Motorola out. Well, that didn't happen, now did it. Apple will use the G4 for some time to come and in order for that to happen the G4 has to get some updates. Those updates are on the way and a Apple spokesperson said as much recently.
Snowy-River said:
If this chip comes to pass, then great! I'd love to see it. However, I've read too many times when Moto has planned something and it's fallen through or been seriously delayed. How long have they been planning to incorporate RapidIO into their PPC line? Originally it was going to be part of their G5, before they discontinued that development project.
RapidIO was to be incorporated into a replacement for the G4 according to Motorola's 2001 PowerPC roadmap, which is still on the companys website. Motorola ran into financial difficulties that squashed those plans. However, you must understand that a replacement for the G4 was still going to be designed also for the embedded market and not for the desktop only. So, it would still have had performance compromises for the desktop computer market.
The G4 can still be a competitive chip in the portable market with some design changes that Motorola seems to have given the green light to. The Pentium-M is not exactly a performance demon running at a top speed of 1.7 GHz. The G4 has the potential to compete against it in performance versus watts used.
You must also understand that a high performance monolith processor is just one of many ways to get chip speed. Another way is to take a much lower performing processor and stick two or more of them on one chip or stick thousands of them in boxes to make a super computer, as IBM is going to do with the 500 MHz PowerPC processors that will be used in the Blue Gene project. Intel is also planning on making Itanium chips using several lower performing, multiple processors in 2006. Why? Because lower performing processors are much more efficient than high frequency mono processors such as the Pentium 4. Intel is probably coming to the stark realization of how inefficient the Prescott version of the Pentium 4 is after boosting it to 31 pipeline stages from 20.
Snowy_River said:
Maybe they'll get their act together. Maybe not. But, please forgive me if I don't place my trust in vaporware from a company that doesn't have the greatest track record (even if they have managed to boost clockspeeds).
All the comparisons are made with what is being done on the Wintel side. Conveniently ignoring that the X86 side has over 95% of the market and buckets full of money to throw at it. Take a look at the history of the Power4s performance in comparison to Itanium. You will see the same thing happening there. Intel is simply improving Itaniums performance at a much greater pace than IBM can with the Power series of chips. Making a 970 spinoff of the Power4 will help IBM muster enough funds to keep competiting against Itanium for awhile, but as a IBM engineer commented on a chip forum a few months back 'how much longer can IBM stay ahead with Intel sinking so much money into it?'
