|
|
|
16MB vs 32MB? After feedback from the Press Conference, 3Dfx Interactive hinted
that they might indeed reconsider their 16MB ceiling for Voodoo3. This fact seemed to please
a lot of you but is there any sound technical reason for upping to 32MB,
or would it just be a market driven decision? Not to
invalidate the necessity to respond to consumer's perceptions but there
must
have been a valid reason not to include it in the first place? Tarolli went on to say,
Well, again, the main reason we chose to stick with 16 MBytes was because
we feel the target resolution for Voodoo3 is 1280x1024 at 60 fps -- we
continue to evangelize that games running less than 60 fps are below
"gamer" update rates.
The difference may not turn up in a static screen shot, but it quickly
becomes obvious when playing a game. Gamers
realize that something is "off," things just don't look right and movement
and responsiveness is not what it should be.
Since we don't support 32-bit external rendering (which has been largely
discussed elsewhere) there really is not a need for additional memory
beyond 16 MBytes. And, it allows us to hit extremely aggressive price
points that would not be possible with 32 Mbyte solutions. We've also
increased our bandwidth and can support higher speed memories.
Certainly from what Tarolli says above, it seems as though with the 16-bit rendering ceilind there'd be no real
gain to having 32MB. But we still think that the marketing folks out there would lap it up- even if gamers wouldn't necessarily care one way or the other.
And finally many of you expressed concerns as to the release date and whether or not 3Dfx Interactive will be able
to deliver their Voodoo3 on time? In case any of you didn't know,
3Dfx Interactive wasn't the first chip manufacturer to nail the .25Micron FAB
process firmly on the head, S3 had already done so first with their Savage 3D. Even so NEC/VideoLogic
also claimed that they would be able to make the shift to .25Micron by August. However, NEC/VideoLogic has since
admitted that their reason for the delay of the PowerVR Series 2 PC chip was down to
trouble getting all their features crammed on to .25 micron. With the
die for Voodoo Banshee already pretty full, some of you expressed concerns on whether or not the
transformation from .35 micron to .25 micron with added features may effect the
production schedule for Voodoo3.
The move from .35 micron to .25 micron really didn't affect the
development schedule at all. The development schedule was more a
functional of the complexity of the Voodoo3 and the verification of it.
Devices with 8.2 million transistors have a lot of logic that needs to be
tested and verified, and that's what turned out to be the most time
consuming task of the Voodoo3 schedule. As you know, we're far more
complex than a Pentium II. We've planned for 0.25 microns all along, and
we knew that the TSMC process would be quite solid by this point in time.
So far, if things remain on the right track, OEMs will probably be getting their Voodoo3s in
early Q1 1999. Games developers will most likely be getting their boards the same time if not
earlier (the privileged few such as id Software and co.). This seems likely so long as 3Dfx can get all
of the additional features working. The only two parts of the chip we saw working were the 3D and 2D respectively.
Last time around with the Voodoo Banshee, the TV-Out was never implemented due to technical difficulties.
Either way 3Dfx Interactive doesn't make boards, they just make chips and reference designs. At this early stage,
we're not even sure just who will be adding the Voodoo3 to their product line-ups. Of course, the usual crew have
expressed interest and said that they are impressed so the Voodoo3 should start hitting the shelves in late April and early May- we think.
Alex "Sharky" Ross
Editor-in-Chief and General Psycho
|
 |
|