|
Ti 4400 Features & Benefits
- 128MB DDR memory
- 8.8GB/sec. memory bandwidth
- Unique maroon-colored card
- Lightspeed Memory Architecture™ (LMA) II
- NVIDIA nfiniteFX™ II engine
- Enhanced Vertex and Pixel shaders, with new Z-Correct Bump Mapping technology
- NVIDIA nView Display Technology
- Accuview™ Antialiasing
- Dual Vertex Shaders
- Advanced Pixel Shaders
- 3D textures
- Shadow buffers
- Z-Correct Bump Mapping
- Lossless Z Compression
- Unified Driver Architecture (UDA)
- High-Definition Video Processor (HDVP)
- TV-Out and Video Modules
- Microsoft® DirectX®, 8.X and OpenGL Optimizations
Compatibility Designed to WHQL Compatibility Standards
- Windows XP, 2000, 98, 95, ME, NT® 4.0display drivers
- Microsoft DirectDraw®, Direct3D®,
- DirectVideo®, and ActiveX® drivers
- OpenGL ICD for Windows XP, 2000, 98, 95, and NT
- Complete Linux display and OpenGL drivers support
- Fully PC00, PC99 and PC99a compliant
Nothing entirely exciting about the specification if you've seen an NV25 board before but some items of note. NV20 pioneered programmable effects hardware for NVIDIA and NV25 continues that with an evolution of the hardware and logic you saw on NV20 and GeForce3 boards. Dual vertex shader units and a beefed up pixel shader form the core of the evolved silicon with the TMDS/nView and the Accuview anti aliasing engine providing the supporting roles.
Accuview is NVIDIA's new anti aliasing logic providing a range of multi sampled modes including the excellent Quincunx mode bringing good performance much like 2X AA but with 4X level samples and a softer edge than true 4X sampling, making it the best of the modes for me.
The hardware also implements bandwidth maximising technology that we first saw on the original ATi RADEON like a lossless Z-Buffer (depth buffer) compression routine, quick Z-Buffer clear and selective Z-Buffer testing to discard pixels that wont be drawn, similar in concept to HSR, deferred rendering and to some small extent tile based rendering, all technologies that seek to only draw what's needed.
This saves bandwidth by discarding pixel data that will never be drawn before it hits the texturing stage of the pipeline and costly texture (colour, usually a 32-bit RGBA pixel format) data has to be pulled from memory, either on card or from system memory.
So while Ti4400 supports around 8.8GB/sec of card memory to GPU bandwidth, the GPU still tries to be as effecient as possible in its usage, maximising performance. Of course these are GPU benefits and implemented on all cards, not just PNY's, but the effect is clear.
Onto the performance!
|
