Tseng Labs

~ the last words ~

Tseng Labs made the chip, not the card. That distinction is why almost nobody remembers the name and almost everybody who used a PC between 1990 and 1995 was looking at its output. The company sold silicon to board manufacturers, who put their own names on the box, so the ET4000 turned up under a dozen brands nobody kept track of and one chip effectively became the standard VGA of its era.

The ET3000 came first, in september 1987, and it was decent. The ET4000 landed in 1990, the same year Windows 3.0 shipped, and the timing was the whole thing. Windows made ordinary users care about how quickly a screen could redraw for the first time, and the ET4000 redrew faster than anything near its price. The trick was a small least-recently-used cache on the chip and an aggressive ISA host interface, which meant it could get away with plain DRAM where competitors needed video RAM. Tseng claimed up to a tenfold speedup on text and area fills. IBM put the ET4000/AX into its own PS/1 machines in 1992, which is about as strong an endorsement as a component vendor could get.

It also became the benchmark chip in a slightly circular way. The performance yardstick everyone quoted had the ET4000 as its reference point, so for a few years "as fast as a Tseng" was the specification you argued about in a shop before you bought anything.

The ET4000 sits just before my own machines. By the time I was picking parts myself the argument had moved to Voodoo cards and 3D, and Tseng was already an asset sale. Where the name survived was in the driver list. Anyone who installed DOS software or set up Windows 3.1 in that period scrolled past "Tseng ET4000" in a menu of display options without knowing it was a company in Pennsylvania. That is the strangest sort of fame: a corporate name that millions of people read hundreds of times and nobody ever pronounced.

~ the record ~

~ how it landed ~

The business model worked because graphics in 1990 was a solved problem with one variable. Everything drew the same rectangles and the only question was how fast. In that world a fabless design house with one very good chip could take a large share of the market without a factory, a brand or a retail presence, because the people making the buying decision were board vendors who cared about cost per unit and nothing else.

That is also the entire vulnerability, written out in advance. Tseng had no relationship with the person who ended up with the card, no software ecosystem, no reason for anyone to wait for its next part. When the basis of competition changed, there was nothing holding the customers in place.

~ the fall ~

The change came in two pieces. First the RAMDAC, the converter that turns digital pixels into an analog signal, moved onto the chip everywhere else while Tseng kept it as a separate part. S3 and others got a cost and speed advantage out of that for years before the ET6000 finally integrated one in 1996, by which point the argument had moved on again. Then came 3D acceleration, and Tseng had nothing. The ET6300 was planned and never shipped.

Development stopped in november 1997. On 22 december ATI announced it was buying the graphics design assets and roughly forty engineers for about three million dollars, which was a rounding error against the company's cash. What was left was a public shell with money in it, and in 1998 management merged that shell with a pharmaceutical company called Cell Pathways. So the corporate entity did not die so much as get repurposed, and the part that mattered went into ATI, then into AMD, where some of the people who built the fastest 2D chip of the early nineties spent the following decade building the 3D chips that had made theirs obsolete.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »