IBM Preps for Apple Announcements

[Phinius]

Re: Re: Apple isn't going to put all their chips in one basket.

Originally posted by Vlade
If the 970 is cheeper to make and faster why would apple not use it in all there machines?

Chip manufacturers charge more for higher speed on a given chip model. So, a 970 running at the slowest speeds might sell for less than a 1.42 GHz G4 that Motorola now makes. That could also be due to Motorola introducing a low-k dielectric substance on some of the highend G4 chips and bringing the voltage up to achieve 1.42 GHz.

In the third quarter of this year, Motorola will move the G4 to the same process size as the 970, and its very likely that the G4 will then be less expensive to produce than the 970 due to the 7457 G4 being a much smaller chip than the 970 on the same process size.

 

[Phinius]

Re: Re: Re: Apple isn't going to put all their chips in one basket.

Originally posted by JoeRadar
Product differentiation. I would not be surprised to hear about a major upgrade to the G3 line Monday.

See The Register Article for some additional rumors.

I would not be surprised to see Apple eliminating the IBM G3 entirely in Macs. That's due to the G4 being faster on a given process size. The listed top speed for the upcoming 7457 G4 on a .13 micron process size is 1.3 GHz and the 750FX ( G3 now used in the iBooks) seems to top out at less than 1 GHz.

If IBM does add Altivec to the 7XX series G3 PowerPC processors, then Apple may intend to use it in a low powered and inexpensive computer model. But it's extremely doubtful that Apple will eliminate the Motorola processors from the Mac lineup, since Moto intends to move the G4 to a dual-cpu chip in the near future and its's unlikely that another customer would need a dual-cpu processor (running at 25 watts) for an embedded use.

 

[Phinius]

Speed of memory restricting frequency boost of 9XX chips?

The 970 is designed so that the core frequency runs at twice the speed of the main memory. So, using a dual-channel memory motherboard, the 970 will need 533MHz DDR-II memory in each channel in order to keep up with a core frequency of 2.4 Ghz. The trouble is, 533 MHz DDR-II memory is not expected to be in production until late in 2003. So for the 9XX processors to obtain 4GHz speeds in 2004, there needs to be memory running at 1 GHz for a dual-channel memory chipset or else Apple will have to come up with a more expensive quad-channel chipset in very short order.

So, its very likely that the rate of frequency boosts for the 9XX series of processors will depend very heavily on the availability of faster memory. But that does not preclude the ability to increase the performance of the 9XX chips in the near future, due to IBM using other methods to increase the speed of the upcoming Power5 and 9XX chips.

 

[pianojoe]

Re: Optical Out

Originally posted by analogkid
I love this idea, especially combined with core audio... saves me a pci slot and I can probably whip audio into the mac's input, process, and feed the digital output about 3ms later.

That's exactly IT! Like, buy a couple of G5s, use one to run your sequencer, and the others to process your realtime digital effects. Logic 7 will sport distributed realtime audio rendering... drool )

 

[scan300]

Originally posted by solvs
Then why Dual 1.4 and Single 1 GHz G4s? Sorry, but anyone who calls the current G4s overclocked, obviously doesn't know the meaning of the term.

Originally posted by Adobe 75 NO NO NO NO!! for the millionth time, apple is NOT using overclocked G4s. Motorola simply chooses not to market the G4 as a desktop processor, simply quoting the 1GHz max speed for EMBEDDED use.

I agree, but I was using shorthand to get to the point about yields and pricing and so why it would currently be unviable for Apple to implement a Dell-style 'make-your-own' processor configuration.

I too posted in earlier discussions that the post 1GHz chips weren't overclocked, and why, but it seems to me that it still pervades the current vernacular so I thought it OK to use the phrase.

I'll clarify...

Low and high clocked chips come from the same process, are tested for imperfections and given a rating. The higher quality chips are rated with higher clock speeds and generally there is a lower yield of these. Hence the delays lately of availability of the highest performing chip ever since the 500 MHz debarcle... Motorola had fabrication issues.

My reasoning is that Apple will try to control which chips are sold by fixing the type of configuration available, by making the lower rated higher volume chips available in a larger range of products, so we don't have a backorder crisis in one area and a glut in another area.

I also made the point that Apple has an opportunity to make much cheaper PMs by offering a single processor design, now that they have better performing chips. The price of Macs being the other thorn in Apple's side as well as the performance issue.

Delivery is another issue which Apple needs to control well. By having fewer options they can manufacture ahead of orders and keep a buffer in their distribution channels, keeping distribution delays down.

The main reason, IMO, that Apple introduced dual proc PMs across the board was to give the impression that they still had competitively performing machines to the rest of the PC industry, despite only having at best 1.4GHz procs. That they could devise and implement this solution is a credit to Apple, but due to the fact that they used 2 procs meant that they couldn't offer a cheaper PM that still could brandish the moniker PM.

Thanks for not flaming me...

 

[pianojoe]

Originally posted by scan300


The main reason, IMO, that Apple introduced dual proc PMs across the board was to give the impression that they still had competitively performing machines to the rest of the PC industry, despite only having at best 1.4GHz procs. That they could devise and implement this solution is a credit to Apple, but due to the fact that they used 2 procs meant that they couldn't offer a cheaper PM that still could brandish the moniker PM.

But, I wonder, will the new midrange system be significantly faster than the current midrange model??

2x 1.25 G4 <-> 1x 1.8 G5

If you assume that with the G4 the performance gain of the 2nd proc is only 50%, we have a G4 performance equal to a 1.875 G4. Since the G5 is said to be 1.25 faster than the G4 (1.8 G5 -> 2.25 G4), we'll see a performance gain of, well, 20%. That's nice, but not earthshaking.

 

[scan300]

Originally posted by pianojoe
But, I wonder, will the new midrange system be significantly faster than the current midrange model??

2x 1.25 G4 <-> 1x 1.8 G5

If you assume that with the G4 the performance gain of the 2nd proc is only 50%, we have a G4 performance equal to a 1.875 G4. Since the G5 is said to be 1.25 faster than the G4 (1.8 G5 -> 2.25 G4), we'll see a performance gain of, well, 20%. That's nice, but not earthshaking.

You're right, but it's beyond me to speculate the optimum price/performance ratio in the product grid they design. All I can say is that Apple IMO do have an opportunity to make cheaper PMs through a single processor option. Whether they could so this with a single motherboard or whether they need 2 designs, what would be the hypertransport implications, bus controllers etc... I have no idea.

It depends on their market research. Everyone seems to want faster PMs, but they also want cheaper. Generally I find Apple will target a specific sector in the market and try to win customers by offering a product they think can win in that sector.

eg If a new mid range PM is significantly cheaper than the current G4 version, (which isn't selling well) with only a 20% perfomance gain, there might be customers, like those holding off upgrading their G3s until Quark 6 was released, who like the overall price/performance package.

 

[analogkid]

Originally posted by pianojoe
the G5 is said to be 1.25 faster than the G4

You obviously weren't looking at the same benchmarks i was.

 

[solvs]

Originally posted by scan300
Thanks for not flaming me...

See, now if you had said all that, we wouldn't have come in with guns a blazin'.

Saying the 1.4s are OC'ed 1 GHz is like saying the 3.06 P4s are overclocked ( wait... bad example ).

 

[Phinius]

Originally posted by analogkid
You obviously weren't looking at the same benchmarks i was.

Well, according to this Aceshardware.com review, a 1.25GHz G4 has a performance equal to at least a 1.8GHz Pentium 4. If you add in a 14% speed increase moving the G4 to 1.42GHz and then another 30% boost from moving to a smaller process size, then the upcoming topend G4s should be at a minimum equal to a 2.66GHz Pentium 4 with a 533MHz bus.

A 970 at 1.8GHz might have about the performance of a 2.66GHz Pentium 4 on SPECint according to 970 SPECint estimates given by IBM. So, that would put a 1.8 GHz G4 close to the same performance as a 1.8GHz 970 processor, with both of them being made on the same process size. Of course the 970 has more potential for a higher frequency and the next version of these 9XX chips will undoubtedly jump up considerably in performance when they are based on the upcoming Power5 processor.

 

[rhpenguin]

Ive seen the screenshots and i think theyre brutaly ugly. But its new so i wont pass judgement till i actually use it.

Maybe a theme picker could be included (dont give me hell if there is one in 10.2.6 because im new)

New features are always fun so hopefully theres a beta of this available soon!