PDA

View Full Version : French Trans Req: MacBidouille, Motorola, IBM Production


arn
Jul 24, 2003, 11:29 AM
http://www.macbidouille.com/niouzcontenu.php?date=2003-07-23#6206

Human translation please.

arn

robg
Jul 24, 2003, 02:38 PM
quickie (summarization):

The first paragraph just says that the writer will tell you something important about the difference between Motorola and IBM production. He also goes on to say that it is uncomfirmed as of yet.

Paragraph two goes on to describe wafer production as similar to taking a photo. They shoot light on the wafer in certain patterns that become conduits for electricity. The success rate is very variable. Sometimes the conduits are bridged by dust and other mistakes. After making the cores they test them using equipment to make sure they hold their own (equipment is made by KLA-Tencor, just had to throw that in because I have family that works for KLA).

IBM's Fishkill plant produces 12" wafers (one of the first in the
world btw). Nearly 80% of the chips are clocking at 1.6 to 2 GHz (category N). There are 105 cores on a PPC970 wafer. 80% of the category P wafers (2.5ghz) are also passing, but Apple hasn't started using those yet. Good news for future PowerMac production, though.

Motorola has 98 7457 cores on a wafer. But there is a problem. There are 5 more inferior cores per wafer than IBMs 970 wafers.

And Motorola has officially abandoned G5.

The article describes a little bit about chip production, but I don't want to translate. You can get an official write-up on production at IBMs website, if I remember correctly.

chazmox
Jul 24, 2003, 02:51 PM
Can you compare the numbers for P, N, and non-Passing parts ( forgot the letter for this )?

I read another translation that the MacB article has Mot's yields in the .4% range ( which I don't beleive ). Could you provide these numbers here and compare to IBM?

Thanks!

robg
Jul 24, 2003, 05:13 PM
P and N are both suppposed to be yielding 80%.

I think MB was claiming that the G5 never came to be because their prototype yields were .4-1.2%.

Sun Baked
Jul 24, 2003, 05:33 PM
They finally posted the translation on the us.macbidouille.com section.

Motorola and Apple. 4 years of disappointment, a follow up [update] (http://www.hardmac.com/niouzcontenu.php?date=2003-07-24#256)

[Update] We were advised to rub our figures, which we did, keeping only overall values.

The previous news about Motorola CPU unit's disappointments drew an impressive amount of reactions. Before going further on this chapter, we wish to state we hold no personal grief against that company, which has brought to the general public so many important devices. Therefore, one must realize that their CPU business unit, which produces theG4 chips, hasn't been successful in keeping to the success of the other units.
We received very interesting information on IBM and Motorola chip production results, which we must say no-one can confirm for us. First, some information to remind you how chips are produced.

This production is very much like a photographic development, except the support is a ‘wafer’. This wafer is a circular board of different sizes depending on production site. Over this board, light is cast through a mask. This mask holds the drawing of the CPU and all its circuits, and during this insulation process, some places will receive light while some others won't. By a process similar the one used to develop photo, the places that received no light are wiped off so that a CPU emerges. Many CPUs simultaneously produced from a single wafer.

Like in every production process, success rate and quality are very diverse, and know how is crucial. The thinner the engraving, the faster the CPU will run. This means a tenth of a thousandth millimetre error margin! Chips from a wafer are then tested to a certain frequency, for which they'll be certified for, or else they'll be dumped.

Therefore here are some very recent results observed by IBM & Motorola:
Both IBM & Motorola divide their chips in 3 categories according to the frequency they can use. The best are P, then come L and N categories.

On a 12" wafer, IBM simultaneously produces 105 PPC 970. The certified CPU rate is high.
Almost 80% of certified CPUs reach 1.6 to 2 GHz, those ones will be found in our computers.
A small number in the P category even reaches or goes over 2.5 GHz.
As there are so few of them, P category isn't used by Apple though, which is also the case for N.

On a wafer, Motorola simultaneously produce 98 G4 PPC 7457. There's a problem though. Out of these 98, only a very little quantity is usable, around one fifth of IBM results.
According to our sources this would be the explanation of the new Powerbook delay. To end up with it, we'll let you know that when Motorola eventually dropped its G5 project, their engraving results were from 1.2% to 0.8% for the N category, and 0.4% for L (there was no P category).

[Transl. Kalomir]Click on the link and the timeline that follows is interesting, sort of slamming poor Moto. about their innability to perform.

ftaok
Jul 25, 2003, 06:23 AM
Originally posted by Sun Baked
They finally posted the translation on the us.macbidouille.com section.

Click on the link and the timeline that follows is interesting, sort of slamming poor Moto. about their innability to perform. Personally, I don't think the slam on Moto's 7457 lateness is fair. How can the chip be delayed when Moto stated that they'd be out in Q4 2003? They've since moved the date UP to Q3 2003. That means that they should be available at any moment (up to Sept 30). If Q4 rolls around and they're not out, then yeah, bash 'em. But not until then.

BTW, the 7457 is being used by some other company and is almost ready for shipping. I forget who that company was, but Moto had a press release for it a few weeks back.