Neoseeker : Articles : Video : Game Accelerators : OCZ Titan3 GeForce 3 Review
Hardware Newsletter:
Email:

News Headlines
New Articles

Compare Prices

Motherboards
Abit
ASUS
Gigabyte
MSI
eVGA
Intel
Tyan
More...

Processors
AMD
Intel
More...

Memory
DDR
DDR2
DDR3
More...

Video Cards
ATI
eVGA
XFX
BFG
Sapphire
More...

search for lowest prices

send article   hardware newsletter   article comments (5)
OCZ Titan3 GeForce 3 Review - PAGE 2
Daryl Grant - Friday, August 3rd, 2001

The 64MB of 3.8ns DDR SDRAM runs on a vastly different memory controller than previous Geforce chips. The term “memory controller” is actually a misnomer because unlike the Geforce2 Ultra which uses a single 128-bit (DDR) controller, the Geforce3 makes use of four, interleaved, and load-balanced 32-bit (DDR) memory controllers culminating in a system that NVIDIA likes to call a “Crossbar Memory Architecture”. Both chips have 256-bits of bandwidth (Geforce2 Ultra: 128-bit DDR / Geforce3: 32-bit DDR times 4), but the Geforce3 is much more efficient. For example, when dealing with small sets of data – 128-bit chunks for example – the Geforce2 Ultra would using the entire 256-bits to transfer only 128-bits worth of data, whereas the Geforce3 can access the data in 32-bit chunks, in parallel. This advancement is part of the Lightspeed Memory Architecture which also includes the Visibility Subsystem.

The Visibility Subsystem is very similar to ATI’s HyperZ technology in that they both remove non-visible values from the z-buffer before they are sent to the frame buffer. This conserves memory bandwidth for pixels that are actually going to be seen on the screen. (The z-buffer is a part of the video memory that holds the depth values of onscreen elements – by comparing these values, hidden pixels, as in, pixels that are blocked from view by other pixels, can be removed. The frame buffer is the part of the video memory that contains the rendered screens before they are displayed to the monitor). Like HyperZ, NVIDIA’s solution also includes 4:1 lossless z-buffer compression which also conserves memory bandwidth.

But the major advancement that everyone should be (and is) talking about is the nFiniteFX Engine – a key milestone in consumer-level, computer graphics. Made up of the Hardware Transform and Lighting Engine that we know all know (and love?) as well an nFiniteFX Pixel processor and an nFiniteFX Vertex processor. The Hardware T&L is no big news, but the chief innovation of the latter two is that they are programmable – that is, game developers now have ability to program their own vertex and pixel algorithms.

If you remember Environment Mapped Bump Mapping (EMBM) which was introduced my Matrox, the nFiniteFX Engine has it beat. EMBM used bump textures to create the illusion of complex surfaces (such as water, rock, and human faces). This method has obvious limitations, not the least of which is the fact that this “complex” surface is fixed in nature – i.e.: it is impossible for the object’s texture to interact with the environment. With the nFiniteFX Engine, these surface complexities can be created and rendered in real-time, all at the developer’s whim, allowing a texture to interact with its environment. This opens up a whole new world of exploration for game developers to creates and game enthusiasts to enjoy.


Article Index

1.Introduction & Specs
2.Features
3.Box Contents, Impression & Test System
4.Benchmarks
5.Benchmark Results Con't & Conclusion

Submit our article to: diggDigg this! de.le.ciousdel.icio.us

Get updates when we publish new articles
Email Address:
(0.0466/d/nova)