A Multimedia PCI card released in 1995 and
most commonly sold to retail as the Diamond
Edge 3D under license from EDGE Games. It
featured a complete 2D/3D graphics core
based upon quadratic texture mapping, VRAM
Memory (DRAM for the STG2000), an
integrated 32-channel 350 MIPS
playback-only sound card, and two Sega
Saturn compatible joypad ports. As such, it
was intended to replace the 2D graphics
card, Sound Blaster-compatible audio
solutions, and 15-pin joystick ports, then
prevalent on IBM PC compatibles.
May 1995, two-year-old startup NVIDIA made
a long awaited entrance in the graphics
market with their two chip solution to
accelerate wavetable audio, video playback
and 3-D graphics. Although the NV1 was
NVIDIA's first product, the chip was very
advanced for its time.
To give you an idea of the state of the
industry at the time, "MPC Level 2" had
just been declared to promote audio and
video technologies on the computer. CD-ROMs
and 16-bit sound cards just started growing
in popularity. Graphics accelerators at the
time supported full motion video as well as
photo-realistic, high-color 15-bit and
higher color depths. It was in this
environment that NVIDIA developed its first
graphics chip with the help of SGS-Thomson
Microelectronics (ST Micro). Here was
fledgling NVIDIA integrating all these
multimedia functions and more into one
expansion board.
Why did NVIDIA choose such an integrated
solution?
Alongside the original announcement,
Michael Hara (NVIDIA’s director of
developer marketing) said the following :
"What the consumer is used to is high
quality for a low cost. But for the PC, the
software developers don't know what the
consumer is going to buy, so he assumes no
video, GUI acceleration and an 8-bit
SoundBlaster. We believe that this is the
next level of integration for multimedia in
the PC. Compressing high performance
multimedia functionality onto a single
multimedia accelerator will ensure software
developers a common performance baseline to
work from, and in turn assures consumers
will get the most from the multimedia
experience.”
When the NV1 was produced, many of the 3D
standards we now take for granted had not
yet been decided. Since polygons had not
yet become the standard for 3D gaming,
NVIDIA chose to implement Quadratic Texture
Maps, not polygons, as its graphics
primitive. NVIDIA's Chief Technical
Officer, Curtis Priem, was enamoured by
quadratic texture maps, and as CTO, he made
quadratic texture maps the standard at
NVIDIA. Whereas today's 3D cards must use
many small polygons to replicate
smoothness, the NV1 used the curved sides
of polygons. This allowed the NV1 to
display smoother looking 3D models with
fewer calculations. Textures were stored in
system RAM and pulled over the PCI or VESA
Local Bus as needed - exactly what AGP
promised a few years later.
The NV1 also integrated a playback-only
sound card. With 32 concurrent audio
channels of 16-bit CD-quality audio and
hardware phase shifting for simplistic 3D
sound, the NV1 was actually more impressive
than many first generation PCI sound cards.
The MIDI playback used a 6MB patch set
stored in system RAM and was even Fat Labs
certified.
The NV1 featured direct support for Sega
Saturn gamepads and joysticks instead of
using the traditional 15-pin game port. The
Saturn controllers could be used in any
DirectInput compliant games, including the
hardware accelerated ports of Virtua
Fighter, Virtua Cop, and Panzer
Dragoon.
The Audio, video and graphics were hewed
together using an internal 600MByte/sec
packet bus. The device also interfaced to a
64-bit PCI for maximum transfer
rates.
Instead of making the chips themselves,
NVIDIA took a different avenue by
establishing an early relationship with
SGS-Thomson Micro. By gaining manufacturing
stability, NVIDIA gave a high-volume
consumer segment to SGS-Thomson, which
introduced a DRAM-based version called the
STG2000 while NVIDIA would market the NV1
video RAM version for higher-end PC
applications.
The two companies also co-developed the
digital-to-analog (DAC) converter found on
the board.
The DRAM version by SGS Thomson's STG2000,
was available in 1MB upgradable to 2MB (the
Diamond Edge 3D 2120XL) or 2MB standard
(the Diamond Edge 3D 2200XL.) The VRAM
configuration, based on NVIDIA's NV1, was
available for PCI in 2MB upgradable to 4MB
(the Diamond Edge 3D 3240XL) or 4MB
standard (the Diamond Edge 3D 3400XL.) Both
chips were based on NVIDIA's NV1
technology.
The Diamond Edge was the most well known
product, and by far the most popular,
although there were a number of other
manufacturers (seen below) that announced
products based on the NV1/STG2000 with
small differences.
NV1 :
Genoa Stratos 3D
Jazz Multimedia 3D Magic
STG2000 :
Leadtek WinFast Proview 3D
GD400 & GD500
Yuan JRS-3DS100
NVIDIA and STS-Thomson weren't competitive
since they served different price points.
NVIDIA would focus on performance and
enabling more features and SGS would
focus on serving the commodity
market.
Late 1995, many software companies
announced products, such as Electronic
Arts, Activision, Sega, Crystal Dynamics,
Papyrus, Domark, Alexandria, Martin Hash,
Inc., and Sonic Foundry.
There was one platform which shared the
NV1’s use of Quadratic rendering technology
- the Sega Saturn - which created an
alliance between Sega and NVIDIA resulting
in Saturn games like Nascar Racing, Virtua
Fighter and Virtua Cop being ported to PC
to take advantage of the NV1 and its unique
method of 3D processing, being bundled with
the card at the time of sale.
Even the well known Mechwarrior II was
supposed to be ported to the NV1. NVIDIA
assigned one engineer to them, but the
whole project was dropped since it was too
big of a task to be done by one engineer.
The NV1 version was actually the first
3D-accelerated Mech II version to start
production, but was never completed. All of
the builds NVIDIA provided them with were
archived, documented and put into storage
somewhere at Activision. (think big
warehouse like at the end of Raiders of the
Lost Ark.)
What went wrong with the NV1 and how did it
almost kill NVIDIA?
The proprietary quadratic
texture mapping of the NV1 was its death
sentence. When Microsoft finalized Direct3D
not too long after the NV1 had reached
store shelves, polygons had been chosen as
the standard primitive, and despite
NVIDIA's and Diamond's best efforts,
developers were no longer willing to
develop for the NV1.
Nvidia did manage to put together limited
Direct3D support, but it was slow and buggy
(software-based), and no match for the
native polygon hardware on the
market.
Other factors like questionable audio,
overall cost to NVIDIA and the consumer,
slow 2D speed and the failure of the Sega
Saturn vs the Playstation all contributed
towards its demise. But nothing doomed the
NV1 more than the release of
Direct3D.
With Direct3D, Microsoft nearly killed
NVIDIA. PC OEMs refused to produce boards
with a non-Direct3D compliant chip,
software development was dropped, and
NVIDIA's engineers knew they could not come
up with a completely new polygonal 3D
accelerator and bring it to market in time.
The company retreated from the public
interest and was forced to lay off several
employees.
Subsequent NV1 quadratic-related
development continued internally as the
NV2, which was all ill-fated console chip,
partially because of the same quadratic
problems that arose, which was regarded as
a nightmare to work with.
|