Myths and realities
thatwendigo said:
However, given that the 970 is only likely to get cheaper in volume as IBM refines the process, and perhaps even cheaper when the technology is applied to the 975/980 successor chips, I think that we might see some even better options in the consumer line soon.
If the successor to the 970 is based on the Power5, then it's likely to be a larger and more power hungry chip (since the Power5 is about 25% larger than the Power4 and uses about 25% more watts) and therefore more costly to make than the 970.
As I keep telling peopel, the PowerPC is a more expensive platform just at the basic levels of boards and chips.
The PowerPC is more expensive for Apple due to the higher costs involved with having to develop so much of their own hardware and selling it at a much lower volume than Intel does.
If enough others adopt the standard, this will almost certainly come down, and the AMD and Sony deals are likely just the tip of the iceberg.
I'm not aware of any statement from IBM, AMD or Sony that says the 970 will be used by AMD or Sony. Only rumor sites have stated such.
No, Apple, IBM, and Motorola worked together on the SIMD units in the PowerPC architecture, and each used a different name. The only 'AltiVec' invented by Motorola is their brand name for a joint technology design.
Well, where was IBM's Altivec or VMX implementation for Apple when Motorola came out with Altivec in 1999? Why didn't the great and powerful IBM wait until 2003 to make a Altivec processor for Apple's use if the company was so much better than Motorola? Is that somehow making better PowerPC advancements than Motorola in that time period?
How about IBM's failure to match or beat Motorola with SMP or L3 cache capabilities on a PowerPC processor for Apple's use.
Or why is it that IBM never showed their superior capabilities by matching or beating the frequency advances that Motorola was making with the G4? The G4 is at 1.5 GHz and the G3 is at 1.1 GHz.
The IBM's G3 has never advanced beyond 4 pipeline stages and yet Motorola invested the money to move the G4 to 7 pipeline stages.
Higher pipeline stages only make for faster staging of the clock in revisions.
More pipeline stages makes for greater frequencies and as a result higher performance. There is however a point of diminishing returns where it's not worth it. An Intel research paper stated that the point of diminishing returns was at about 50 pipeline stages, an IBM research paper stated it was at about 20 pipeline stages.
In terms of the design of the PowerPC, the Deep and Wide approach is more efficient and less error-prone.
Your just pulling this stuff out of a hat aren't you. The Power4 and 970 are very error prone.
SOI is not a Motorola invention, and they really ought to have done more with the frequency.
Motorola more than tripled the G4 frequency in 4 1/2 years. Compare that to Intel increasing the Pentium 4 frequency by about 2.25 times in 3 1/2 years. Raising the frequency is only one of many ways to get performance. Adding more cache is another way and putting two processors on a chip yet another method.
Also, the fastest processor that Intel makes runs at 1.5 GHz and it's called Itanium, not Pentium. So clock frequency is not necessarily a good way of judging a processors ability when comparing two different designs.
a timeline of Intel and AMD jumps in the same timeframe.
Motorola: 400mhz-1.5ghz
Intel: 1.5ghz-3.4ghz
AMD: 1.0ghz-2.4ghz
Unlike Motorola and Intel, AMD actively improved their designs while also scaling the clock. We had to go to IBM to get the same effect.
Hogwash. Motorola, Intel and AMD all actively improved their designs while scaling the clock. Motorola moved the bus speed up, added SMP, L3 cache, onboard L2 cache, temperature diode, power management features and of course Altivec. Motorola also added SOI and a low-k dielectric to their chip making process.
Meanwhile IBM basically split the existing dual-core Power4 in two to come up with the 970. IBM also added a less capable VMX or Altivec to the 970 than what Motorola uses on the G4. Seems IBM was in a hurry to get the project done.
IBM did not create a whole new chip architecture for Apple's exclusive use. Nor would any chip manufacture make a design solely for Apple's use in a shrinking marketshare. Yet people expected Motorola to have come up with a high watt usage processor design that they couldn't hope to sell anywhere but to Apple.
Realisticly Motorola's plans have always been to create processors for Apple's use that could also be sold in the embedded market (in fact Motorola's website explains that host processors are introduced in computers first and work on down through the embedded market). Which means keeping the power use down. There were obvious setbacks for Motorola's 2001 roadmap timeframe (still on Motorola's website). Details of the upcoming dual-core G4 design were mentioned for introduction at the .13-micron level and not at the .090-micron process size (perhaps even more pipeline stages). This is basically what the so-called Motorola G5 was going to be. Yet, it would never intended to be a Pentium 4 trump card. Apple realized that and started working with IBM close to 4 years ago to bring out a PowerPC version of the Power4 architecture. The 970 simply brings Apple up to a whole new level of performance. It is not meant to be a complete replacement for a low power use chip like the G4. Lower power use is not it's foretay, much like the Pentium 4 does not make an ideal processor for portable use.
You believe Motorola's roadmap? After all this time, you trust a press release from them?
That pdf file is from a Motorola presentation at a developers conference last year. It is not in any of the companies press releases.
To jump to the conclusion that Motorola is planning on or has stopped ongoing development of the G4 flies in the face of statements to the contrary from both Motorola and Apple. Yes, I stated Apple. An Apple spokesperson recently stated that Apple has no intentions of dropping the G4 in the near future and they were glad that Motorola is continuing to develop it. The 7457A that is found in the PowerBook and it's use in the iBook are recent examples of that. Using your sense of jumping to unfounded conclusions...has IBM stopped development on the 970 since there has been no more speed advancements since August of last year.