A Preview of NV20
tnaw_xtennis 2001.01.18
NV20, the new generation 3D graphics processor of NVIDIA, is
coming. With the capability
of executing programmability instructions just like
CPU, NV20 strides into a new stage of 3D
graphics processor development. For
those waits eagerly for NV20, here is a brief preview
interpreting the new
features of NV20 and how the performance improvement can users expecting
at
present. In addition, feature survey
tables of the NVIDIA based video cards from TNT to
NV20 are presented also.
1. NV20 - a DirectX 8.0 compliant and programmable video card
NVIDIA always brings us some marked new technology when
released its new video chip.
Hardware Transforming & Lighting (T&L)
came with its first generation GPU - GeForce 256,
while GeForce2 took on NVIDIA
Shading Rasterizer (NSR). This time, NV20 is a designed as
a DirectX 8.0 compliant and programmable video card. Microsoft's DirectX is a suite of
APIs (Application Programming Interfaces) that be applied to write nearly every game for the PC.
DirectX 7 already provides most of the basic features wanted by game developers
when making
games. While DirectX 8, the latest version of DirectX,
has been improved remarkably in almost
every
aspect and will give game programmers a real leaf.
Here are some descriptions of new features and performance
enhancements of the DirectX 8.0
concerning the graphics aspect from Microsoft's
site:
Programmable vertex processing language
Enables developers to write custom shaders for morphing and
tweening animations, matrix
palette skinning, user-defined lighting models,
general environment mapping, procedural
geometry operations, or any other
developer-defined algorithm.
Programmable pixel processing language
Enables programmers to write custom hardware-accelerated
shaders for general texture
and color blending expressions, per-pixel lighting
(bump mapping), per-pixel environment
mapping for photo-real effects, or any
other developer-defined algorithm.
Multisampling rendering support
Enables full-scene anti-aliasing and multisampling effects,
such as motion blur and depth-
of-field blur.
......
The addition of programmable shaders for vertex and pixel operations provides
the framework
for real-time programmable effects that rival movie quality. The
innovative freedom that this
programmability gives back to game developers by
allowing them to implement whatever effect
they see fit with the programmable
pipeline has the potential to unlock a new round of incredible
games. ...Shaders and the programmable pipeline offer great creative freedom for 3D
programmers and provide a vehicle to stay on top of the ever-increasing feature
set of today's
3D graphics hardware.
......
As you can see, DirectX 8 defines programming languages and introduces vertex and
pixel shaders
that allow developers to write programs to tell video cards how to act. Games
that are wrote specially
for DirectX 8 could easily present
a lot more realistic visual effects, and furthermore performance gain
from 5%
to 50% could be attained. However, only the video cards with DirectX 8 compliant graphics
processor, such as NV20 which equipped with programmable
vertex shader and programmable pixel
shader, have the competence to take these sweeteners.
2.
The performance improvement expecting on NV20 at present
Though
the NV20 3D graphics processor has the capability of executing programmability
instructions
just like CPU, we will probably have to wait half year before
games are released that take advantage
of these new features. This is the status just as the case
of hardware transforming & lighting (T&L) of
GeForce 256
and GeForce2. The performance improvement we can expect on NV20 when running
games at present are mainly
rely on its new architecture and advanced memory-interface.
Table 1. The Specifications and Quake Performances of NVIDIA Based Video Cards
Video Card | Core Speed (MHz) |
Memory Bandwidth (GB/sec) |
Pixel Fill Rate
/ Texel Fill Rate (MegaPixels/sec MegaTexels/sec) |
Triangles Processing Speed (Million Triangles/sec) |
Quake3
Performance Index * (FPS) |
TNT | 90 | 1.76 | 180 / 180 |
6 | 43 (Quake2, 16 bit)** |
TNT2 | 125 | 2.4 | 250 / 250 |
8 | 58 (Quake2, 16 bit)** |
TNT2 Ultra | 150 | 2.928 | 300 / 300 |
9 | 39 |
GeForce SDR | 120 | 2.656 | 480 / 480 |
15 | 50 |
GeForce DDR | 120 | 4.8 | 480 / 480 |
15 | 73 |
GeForce2 | 200 | 5.312 | 800 / 1600 |
25 | 100 |
GeForce2 Pro | 200 | 6.4 | 800 / 1600 |
25 | 115 |
GeForce2 Ultra | 250 | 7.360 | 1000 / 2000 |
31 | 122 |
NV20 | 300 | 8.0 | 1200 / | 40 | 152 |
Table 2. Comparisons of Different NVIDIA Based Video Cards
Memory Bandwidth Improvement |
Quake Performance Improvement |
Important New Technology Introduced | |
TNT2 / TNT | +36% | +34% | |
GeForce SDR / TNT2 Ultra | -9% | +28% | Hardware Transforming & Lighting (T&L) |
GeForce DDR / GeForce SDR | +81% | +46% | Double Data Rate (DDR) SDRAM |
GeForce2 / GeForce DDR | +11% | +37% | NVIDIA Shading Rasterizer (NSR) |
GeForce2 Pro / GeForce2 | +20% | +15% | |
GeForce2 Ultra / GeForce2 Pro | +15% | +6% | |
NV 20 / GeForce2 Ultra | +9% | +25% | DirectX
8.0 Compliant Graphics Processor ( programmable vertex shader, programmable pixel shader, multisampling rendering support ...) |
In Table 1 and Table 2, Quake3
was chosen to evaluate NVIDIA based video cards. Though the
system bottlenecks
that limit the Quake3 performance at 1024 x 768 x 32bit are the memory
bandwidth of those cards, the new
generation of video processor always trade on those limited
memory bandwidth more efficiently. In spite
of -9% less bandwidth, a GeForce SDR has a +28%
performance lead over a TNT2
Ultra. A +15% higher bandwidth of a GeForce2 Ultra over a
GeForce2 Pro brings
out +6% performance gain. While with only a +11% higher bandwidth, a
GeForce2
has a performance lead as high as +37% over a GeForce DDR thanks to the
advanced
architecture of GeForce2. So the saying that a GeForce2 Ultra will has
same performance as a
NV20 if the memory speed of that GeForce2 Ultra was
overclocked as high as that of NV20 is
not true.
According
to the NVIDIA chip released policy of +30% better performance of a new
generation
chip over own top rank chip on the market, and the immature driver
state of NV20, a +25%
performance lead of NV20 over GeForce2 Ultra is quite
acceptable. If NVIDIA could polish up
its hidden surface
removal (HSR) support to a level that really usable, the
NV20 users will have an
additional performance gained of +35%.
Links of Turning your GeForce/GeForce2 into a Quadro/Quadro2
Pentium 4 1.5GHz or Athlon Thunderbird 1.2GHz (DDR) ( 28/11/2000 )
Calculating the Potentiality of Your GeForce2 ( 26/12/2000 )