Not sure if this interesting or whether it's relevant content for the Nostalgia forum. If this is not the right place or, if it's just too off-topic for the forum let me know and I'll kill the post.

I grabbed a 3-ring binder from the garbage at the old GMT / MOS building years ago and have had it sitting in storage. Recently, while packing to move I found it and scanned the Manufacturing Report thinking that some of the more knowledgeable folks here might understand some of the process related info; I sure don't.
See the attached PDF.

If anyone does derive interesting tidbits from the tables, I'd love to hear your thoughts and learn about the processes.

Thanks - I can't entirely read it but I did find it interesting. Here's what I think I saw:

They performed a million wafer moves in a month. They put a wafer through 200 process steps in under 30 days. They finished 200 wafers a day. They ran 3 shifts. They had a drop in staffing levels and productivity went up (!). They were making single-metal and double-metal MOS, and also two(?) flavours of bipolar(?) ICs.

They finished the quarter under budget for maintenance costs, but under target for good wafers out.

Thanks for the summary, very intersting.

I wish we knew how the production numbers, rework rates, failures, and overall profitability compared to what was going on during the height of C64 and Amiga production runs.

Yes, some comparative study could be very interesting.

One thing, I think profitability is rather a tricky or dubious measure: your accountant will ask you what number to come up with. Well, almost. A fab's job, day to day, is to get good wafers out - but whether there's a market for the product, or the market sustains a strong price or not, that's out of their control. (Or if the mix of products is right. Or if some of the designs are poor designs which have a low yield.) And every now and then equipment needs servicing, and then replacing, and then the whole line needs an upgrade. So these infrequent but large costs also enter the picture. (Likewise, with losing employees, and hiring and training employees.)

But I think another important aspect is not so much the situation at a given point in time, but any sudden deviations in the wrong direction, compared to any gradual trends, hopefully in the right direction. The idea of a 'yield crash' was very real and prominent for a long time, and pretty much every chip design would have at least one second source, so it would be produced at more than one fab and by more than one company.

Interesting take, not what I would have expected but, it's a world I know nothing about. So again, thanks for the commentary, I appreciate it.

In a later part of the binder the Logistics Report (attached) touches on producing chips for Western Digital and also for IBM and LDO's SCSI 1 interfaces. Made me smile to see the old standard highlighted.

Thanks. Interesting view from the Quality report:


Think of a fab as a large high-volume high-quality photo processor. Everyone's holiday photos are going through, but some are spoiled. How to determine what's wrong? Some raw material? A machine? An operator is mis-trained, or tired, or rushed, or demoralised? Air quality? Water quality? Temperature control? A filter somewhere? Some fix to something which introduced contamination?

From the logistics report, we see gross revenue of 1.7 million per month, and also reported savings of a few thousand a year. That seems a little wacky to me - does it indicate too much pressure to cut costs? (Or do I not have a clue?)

We also see 5200 good wafers out, and 840 wafers scrapped. Is that a good ratio? It sounds like a lot of scrappage to me, but again I don't know. Maybe a machine goes out of calibration and hundreds of wafers go through before corrective action.

I think fabs are a case where you need good management and a good workforce. It's not a good business to have an adversarial approach, or a demotivated or under-resourced workforce. To get high quality, you need everyone to care about their job, and not because they fear punishment. The workforce also need to understand the nature of the work: what kinds of things matter, what kind of things to look out for.

Here's a timely illustration that this is still the way things work - two major fabs have announced contamination problems and production of flash memory chips will be down by many exabytes. (Although this is half a billion dollars, it's also less than half a percent of world production.)

(Edit: note that the article refers to Western Digital as WDC, which is fine and normal because that's the stock ticker abbreviation, but it does not relate to the Western Design Center, who are of course fabless anyway.)

Murphy's law: If you don't control it, it will hit you in the face.

Its slighly amusing for me since my work in R&D is mostly hindered by poor Silicon wafer quality. I've had companies that deliver crp on purpose, companies that deliver it without knowing, companies that introduce it into the wafer and a few that almost succeeded (they had slighly less of it). And its always traces back to lack of control. Managementwise, engineeringwise, timewise or just unwise...