GeForce 4 Ti4200 64mb 3D Card
Manufacturer: NVidia 
Purchasing: nvidia website (for a list of distributors only)
Reviewed: 30th June 2002

Introduction

For a long time now NVidia have been the top-dogs of the consumer 3D-card market. Their current line of GeForce cards (going 2 years back to the GeForce 1) have consistently beaten the competition, and won rave reviews the world over.

For the last 18 months it was looking like the 3D Card market was closing down to being a war of two sides - ATI and NVidia. However, in recent months announcements from Creative and Matrox have indicated that the war is going to be heating up once more. Also, with the Radeon being a fully featured DirectX8.1 graphics card NVidia's GeForce series may be seeing it's first big challenge...

It's all about programming

As mentioned, Creative and Matrox have signaled their intent to move back into the consumer market, but what they have to offer has yet to be seen. ATI and NVidia have both been pioneering and developing their latest graphics cards around a programmable core - shaders. The Radeon 8500 and the GeForce 4 are both companies second attempt with this approach (the Radeon 7500 and GeForce 3 both had programmable units built in), and it'll be interesting to see who comes out on top.

Whilst all the cards are getting more and more powerful with each revision, it's starting to become a battle of features - with programmable shading units providing graphics programmers endless possibilities for special effects. This probably won't bother end-users/consumers too much, but for developers - choosing the correct video card (and getting the most features) is paramount, for NVidia/ATI/etc.. the one who's cards are supported the most by developers will eventually get more sales from end-users wanting to see top-quality graphics.

Direct3D8.1 is the current graphics API of choice for many, and there are currently only two families of graphics cards that are designed to be Direct3D8.1 compatible - the ATI Radeon 8500 and the GeForce 4 Ti series (Ti = chemical symbol for Titanium). The ATI Radeon 8500 was reviewed last month, you can read it here.

Chipset Overview

The GeForce 4's core chipset is quite an impressively powerful piece of equipment on paper, with large numbers in no short supply. The majority of the core components are based on the GeForce 3's architecture - which was the first of the programmable consumer 3D cards (think of it as the GeForce 4's parent). The following list is a run-down of the main features and technologies:

• Latest generation Graphics Processing Unit (GPU), incorporating the latest Transform & Lighting (T&L) engine. The hardware T&L engine was introduced in Direct3D7's API specification, thus will dramatically boost almost all 3D games from D3D7 onwards.
• nFiniteFX II - NVidia's implementation of shaders for advanced pixel-level and vertex-level effects.
     • dual vertex shader pipelines make vertex shaders much more efficient / faster.
     • Z-Correct bump mapping clears up anomalies where two bump mapped surfaces intersect.
• Accuview - Anti-Aliasing implemented in hardware, this newer method is designed not to 'hurt' frame rates as much as before
• nView - for those with 2 or more monitors this software/hardware can display windows on multiple screens; if you're an artist or a programmer it is absolutely amazing how useful this can be. For example, you can run your game on one monitor, and output debugging information on a second monitor (and see both at the same time).
• upto 128mb of video memory - an almost criminal amount of space to store textures/geometry.
• Extremely high performance memory architecture 'Lightspeed Memory Architecture II' (LMA II).
    • Uses a crossbar system to improve efficiency of memory controllers/managers
    • 4:1 lossless Z compression improves Z-Buffer read/write speeds. Early Z-Rejection/occlusion culling helps to reduce overdraw (improving fill rate and reducing bandwidth usage).

The GeForce 4 comes in two main versions - the 'MX' budget range and the fully-featured 'Ti' range. The board on review here is from the Ti family (but is the slowest one). You cannot directly compare the two types - whilst there are obvious similarities (beyond the name!), the 'MX' family misses out on some of the great features that make the GeForce 4 Ti series great; other areas of the press have criticized NVidia (including John Carmack of Doom/Quake fame) for giving the 'MX' family the GeForce 4 branding.

There are three variations of the GeForce 4 Ti cards, the 4200 (reviewed here), the 4400 and the 4600. Essentially they are the same chip with the same features, but different clock multipliers/memory bandwidths - ie, the 4600 is faster than the 4200.

The above values are quoted from the NVidia website, and whilst it is almost certain that performance will follow the same trends, they shouldn't be taken as benchmark ratings (this comes later). For example, the value of 136 million vertices/sec is an impressive number, but what it doesn't say is what type of vertices and what supporting system it is quoted from (if it is even from a real-world test). The bottom line is this: if you do buy a Ti4600 don't necessarily expect to render 136 million vertices every second! 

Click here to go straight to the next page...

Or select a page from the list:
• Introduction
• Installation, Benchmarks and Programming
• NVidia's Developer Relations, Conclusion