Skip to main content

Confabbing on the Fabless Fad

Real men have fabs. 

Jerry Sanders – AMD Founder (circa 1993)

This quote made a lot of sense for AMD, being one of Intel’s main competitors at the time. But for other fresh start-up companies in the mid-1980s, this wasn’t quite the case. Owning a fab wasn’t as easy as getting abs. Capital and maintenance costs were big hurdles, not to mention the additional cost of improving the manufacturing equipment and machines (remember Moore’s Law). Also, the transition from fab to fabless for already existing semiconductor companies can’t happen overnight. There was the issue concerning design architecture and compatibility, the conflict with the ASIC model, and the lack of investors due to the outrageousness of the concept. These issues were resolved with innovations in EDA and the publication of “Introduction to VLSI Systems” (that lead to the creation of simplified abstractions to manage chip design complexity), aaaand… the fabless fad was on.

The Pros and Cons of a Fabless Model

Nowadays, a fab can be worth as much as $10 Billion, most of the cost attributed to the spontaneous change in process development. The price is so high that only Intel and TSMC have modern fabs today. No wonder Bernie Vonderschmitt of Xilinx and Gordon Campbell of Chips and Technologies envisioned a semicon. company thriving without one. As Bernie vowed, “If I ever start a semiconductor company, it will be fabless.” And so in 1984, Xilinx became the first fabless company to specialize in FPGAs while Chips and Technologies became the first fabless company to specialize in graphic chips for PCs.

The fabless concept was a big hit; fabless companies no longer had to pay for equipment depreciation, idle production lines, irregular production schedules, etc. resulting in many other semiconductor companies following on their footsteps. Such success even led to the birth of TSMC, the first “pure-play” foundry in 1987. TSMC, being a pure-play foundry back then, only manufactured and sold wafers and had no design capability whatsoever. New relationships rose between fabless companies and foundries/IDM. A long-term relationship between a fabless company and a foundry had to be built upon mutual trust, and must be mutually beneficial to each other. The fabless company benefits by acquiring access to the most modern fabrication process minus the hefty capital investments while the foundry benefits through obtaining a diversified manufacturing capacity over different market segments.

How the Fabless Model inspired growth and change in the Semiconductor Industry

During a semiconductor downturn, an IDM can find itself in trouble with its fab since the cost of a running production line is estimated to be almost the same as an idle one. But a fabless company with an urgent need to bring a product to market can come to its rescue. This is exactly what Chip and Technologies did for Hitachi. C&T had to deliver a chipset when the PC business was booming, so it had to deliver it fast. Hitachi had a lot of available fab resources at the time and was in dire need of business as well. Given the law of supply and demand, things ended up with C&T fabricating its product at exceptionally low prices from Hitachi. It was a better choice than not having any customer consequently accumulating expenses from an unused resource.

ASIC companies once dominated the electronics industry because one can overcharge on uncontested IP. This changed too when know-how on chip design became publicized and when the fabless model emerged. With pure-play foundries, systems companies in the front end had a way to fully skip pure-play ASIC companies. The whole design can be laid out and the only thing needed for tape-out is a foundry like TSMC (which is also known as customer-owned-tooling). Re-designing a chip is a must due to incompatible processes when shifting between ASIC companies, but not necessary when shifting foundries. Given such lofty advantages, pure-play ASICs were driven to extinction. Today, systems companies, with the help of design service companies, now commonly implement application specific functions on FPGAs.

Perhaps the fabless model wouldn’t be as successful as it is now without the support of EDA. The key enablers that EDA provided were software that covered the entire semiconductor design flow, including verification, as well as a plethora of IP libraries. These complemented the needs of fabless companies so much so that some software had licensing prices that surpassed hardware automation cost by hundreds of times over due to high demand (something that wasn’t thought possible in the past because the scenario was topsy-turvy – hardware was more expensive than software). Companies that cater to this kind of EDA, like Mentor Graphics, Synopsys and Cadence Design Systems profited a lot. They also didn’t just rely on profit, but spurred growth through acquisitions of promising EDA start-ups, thus keeping ahead of competition and making it quite difficult for today’s competitors to keep up. As for IP, ARM and Imagination Technologies have dominated this field of EDA, developing architectures of processors and graphics technologies of visual interfaces in products we use and love (i.e. smartphones and tablets) respectively.

What Future does a Fabless Industry Bring?

Perhaps the need for the fabless model is becoming more crucial than ever, as modern trends in technology demand every segment of the semiconductor industry to become better and more efficient at what it does. EDA, IP, and fabless companies as well as foundries and IDMs, need to synergize and work hand in hand to open new frontiers for development and innovation. As foundries implement fabrication processes that approach the atomic level, EDA companies are faced with the challenge of developing their current products to work with these ever changing processes. One example of continuing dynamic improvement in EDA would be the replacement of Cadence’s Diva and Dracula with Mentor’s Calibre physical verification software.

Taking a closer look at the fabless solution, maybe we could clone the gist of the model and apply it to other applicable areas. Segmentation of a process and identifying which particular part would be better off developing as an individual entity is in my opinion a good way to adopt what we have learned. Can you think of any areas where such an approach would be remunerative?


Popular posts from this blog

Calculator Techniques for the Casio FX-991ES and FX-991EX Unraveled

In solving engineering problems, one may not have the luxury of time. Most situations demand immediate results. The price of falling behind schedule is costly and demeaning to one's reputation. Therefore, every bit of precaution must be taken to expedite calculations. The following introduces methods to tackle these problems speedily using a Casio calculator FX-991ES and FX-991EX.

►For algebraic problems where you need to find the exact value of a dependent or independent variable, just use the CALC or [ES] Mode 5 functions or [EX] MENU A functions.

►For definite differentiation and integration problems, simply use the d/dx and integral operators in the COMP mode.

►For models that follow the differential equation: dP/dx=kt and models that follow a geometric function(i.e. A*B^x).

-Simply go to Mode 3 (STAT) (5)      e^x
-For geometric functions Mode 3 (STAT) 6 A*B^x
-(Why? Because the solution to the D.E. dP/dx=kt is an exponential function e^x.
When we know the boundary con…

Yay or Nay? A Closer Look at AnDapt’s PMIC On-Demand Technology

Innovations on making product features customizable are recently gaining popularity. Take Andapt for example, a fabless start-up that unveiled its Multi-Rail Power Platform technology for On-Demand PMIC applications a few months back. (read all about it here: Will PMIC On-Demand Replace Catalog Power Devices?) Their online platform, WebAmp, enables the consumer to configure the PMIC based on desired specifications. Fortunately, I got a hands-on experience during the trial period (without the physical board (AmP8DB1) or adaptor (AmpLink)). In my opinion, their GUI is friendly but it lacks a verification method for tuning (i.e. the entered combination of specs). How would we know if it will perform as expected or if there are contradicting indications that yield queer behavior? Also, there is not just one IP available, but many that cater to a differing number of channels and voltage requirements (each with their own price tag).
Every new emerging technology has the potential to oversh…

Common Difficulties and Mishaps in 6.004 Computation Structures (by MITx)

May 6, 2018
VLSI Project: The Beta Layout [help needed]Current Tasks: ►Complete 32-bit ALU layout [unpipelined] in a 3-metal-layer C5 process. ►Extend Excel VBA macro to generate code for sequential instructions (machine language to actual electrical signals).
Current Obstacles/Unresolved Decisions:
►Use of complementary CMOS or pass transistor logic (do both? time expensive, will depend on sched.
►Adder selection: Brent-Kung; Kogge Stone; Ladner Fischer (brent takes up most space but seems to be fastest, consider fan-out) [do all? time expensive, will depend on sched.)
►layout requirements and DRC errors

Please leave a comment on the post below for advise. Any help is highly appreciated.