Alpha TNT2 is better than Alpha Voodoo3.
Excuse me if I don't get very excited by that news and decide to sell my TDFX shares. I don't base investment decisions on comparisons of alpha level technology. The problem with alpha level boards are that they only provide a snapshot of a company's technological progress. Technology is only half the equation, however. These alpha level boards tell little about the important business details:
- When will the product ship?
- How much will it cost?
- What's the profit margin on the product?
- What OEM's will use it?
- Will consumers buy it?
- Will the product that ships resemble the performance
of the alpha level board?
TDFX still has many of these factors in its favor.
- STB is already producing Voodoo3 cards and will start
shipping them by the end of the month. NVidia claims
that TNT2 cards will start shipping in mid-April.
That's very ambitious given the state of the technology
as presented in Tom and Sharky's review. TNT cards
didn't start shipping until more than two months after
Tom got to review a reference TNT card. Don't forget
the Rage128 which just started shipping as a card more
than four months after reviewer's first got Alpha
boards. I don't think NVidia will repeat ATI's mistake.
However, unless Diamond or Creative announce a
pre-order program next week for TNT2 cards, I won't
expect to see them until sometime in May. TDFX will
have at least a month head-start; more when you include
TDFX's pre-order program.
- The Voodoo3 is better positioned to win the cost war.
It requires only 16MB of RAM for optimal performance.
The TNT requires 32MB. It's also conceivable that the
Voodoo3 chip has higher yields, costing less than the
TNT2. [With all the features that it does not support,
you'd think the Voodoo3 would be a smaller die than
the TNT2.] Moreover, TDFX can unilaterally decide how
to price Voodoo3 cards. Two parties, the board maker
and NVidia collaborate on product cost and will, thus,
be slower to react to sudden changes in the 3D card
market.
- Voodoo3 and TNT2 will both get Tier 1 OEM wins. That's
pretty much guaranteed since there isn't a single board
maker that can produce boards for all Tier 1 OEM's. Even
if the final TNT2 product is better than the final
Voodoo3 product. The Voodoo3 is still in a good position
because its clearly a better performer than the Savage4
or Rage128. I believe that STB will sell all of the
Voodoo3 boards that it can make. The only thing a
superior TNT2 board might do is force STB to sell the
cards at lower margins.
- Voodoo 3 is still the best card for systems with AMD
processors and AMD captured nearly 50% of the consumer
computer market last quarter.
- Consumers, as usual, are value conscience. They flocked
to the lower priced, TDFX branded Banshee rather than
spend the extra $30-$50 dollars for a better performing
TNT card. Consumers considered the Banshee a good deal
on a respected brand. Consumer sales may not be as
lopsided this product cycle; but Voodoo3 will still be
a best seller.
- Performance of the actual shipping product is important.
Diamond has a reputation for supplying reviewers with
cards that run at a higher clock rate than shipping
products or that are accompanied by drivers tweaked
to boost benchmark scores. (See
zdnet.com
for proof.) Thus, I won't trust a review done with
a Diamond card. It's very difficult to say how close
these alpha level boards will come to the configuration
and performance of the final products. Already there
is a rumor that TDFX has new drivers that make a V3
2000 perform better than V2 SLI. (If true, I hope
these aren't just SIMD enhanced drivers that only
boost performance on the Pentium III.) All indications
are that TDFX is getting very good yields on its
143 and 166 MHz parts (including memory). Will the TNT2
be as successful or will the first batch of TNT2 cards
only contain mainly 125MHz chips with the 150MHz chips
in short supply (if not delayed for a month or two).
Who knows?
I only trust comparisons of actual shipping products to
actual shipping products. |
