Semiconductors, Systems and Fabs
In the 1980s, if you were a semiconductor company, you owned one or more fabs, which were expensive at the time, but nothing like they are now. You designed chips that either had huge volumes, high ASPs (Average Selling Prices), or ideally, both.
Volume production provides your fab engineers with process experience and improves the yields of chips per wafer. Volume also amortizes both the fab cost and the design cost. Unfortunately, systems vendors with only one fab to produce chips face a nasty problem. If the fab is not full, you have a lot of capital tied up that is not being optimally used and this will make management upset. To illustrate this point, Bob Colwell, the Chief Architect of the Pentium Pro, said in a lecture:
“At some point, they [process people] are ready to start running production silicon through their new process. At that very day, you had better have a chip for them to start cranking through their fab, because otherwise, Craig Barrett himself will call you on the phone and express his unhappiness with you and his predictions for your career at that point. It’s really really expensive if the process is ready and the chip is not.”
On the other hand, if the fab is full, you are capacity constrained and unable to fully satisfy your potential customers, which will cause the same problems, except that the sales people will be upset too.
When you build a fab, first you run high-value wafers. Once the fab has aged, and is no longer state-of-the-art, you can run parts that do not need the most advanced technology. For example, Intel has many older 200mm plants that once produced cutting edge MPUs, they are now used to create chipsets for MPUs made in newer fabs.
In the 1980s, if you wanted to have access to leading-edge VLSI technology for your own designs, either:
- You were a semiconductor company, i.e., the era of “Only real men have fabs”. Motorola, Intel and AMD all fell into this category.
- You were a systems company big enough to afford the fab treadmill. For example, IBM always had great process technology. DEC usually had 1-2 CMOS fabs (more on that later) and HP had at least 1, but sometimes there was conflict between the priorities for high-speed CPU design and high-volume lower-cost parts. Fujitsu, NEC, etc. were also chip and systems companies. In any case, you must carefully amortize the fab costs.
- You were a fabless systems company that could convince a chip company to partner with you, where your designs were built in their fabs, and either you had enough volume alone, or (better) they could sell the chips to a large audience of others. For example, Sun and TI have had an arrangement like this, which still exists.
- You were a small systems/chip company (MIPS) that was convincing various other systems companies and embedded designers to use your chips. A few chip partners would make long-term deals to produce the chips, sell most of them to other companies, and as desired, have licenses to make their own variations of the designs. The motivation was that a chip salesperson could get in the door with CPUs, and be able to sell many other parts, like SRAMs (this worked for IDT with MIPS and Cypress with SPARC), or be able to do ASIC versions (this worked for LSI Logic).
In MIPS case, the first few years, the accessible partners were small to medium chip vendors, and it was only in 1988 that we were able to (almost) do a deal with Motorola and were able to do ones with NEC and Toshiba, i.e., high-volume vendors with multiple fabs.
Why wouldn’t a company like Sun just go to a foundry with its designs, or in MIPS case, why wouldn’t it just be a normal fabless semiconductor vendor? The answer is that accessible foundries, geared to producing outside designs, with state-of-the-art fabs didn’t really exist. TSMC was founded in 1987, and it took a long time for it to become a truly attractive offering.
To Fab or Not To Fab
If you own the process, you can alter it somewhat to fit what you are building. If your engineers are close with the fab process people, and you have clever circuit designers, you can do all sorts of things to get higher clock rates. Otherwise, you use the factory design rules or maybe you can negotiate a little with them. In any case, there is a tradeoff between committing the capital to own a fab and achieving higher clock rates, or preserving your capital, not owning the fab, and losing the ability to produce highly tuned designs. To give a rough notion of the trade-offs, in the same process size, the speed of a process (measured in FO4 latency) can vary by a factor of 3, between the best processes at foundries like TSMC or UMC, and fabbed manufacturers like Intel or IBM.
Discuss (4 comments)