Erwin's Pages Central GeForce 4 Ti4200 Face-Off

Table of Contents

1. Introduction
2. The cards
3. Test Rig/Setup
4. The Benchmarks: 3D Mark 2001SE
5. The Benchmarks: JKII
6. The Benchmarks: DronezMark
7. Overclockability
8. Omega Drivers
9. Conclusions - Final Thoughts

1. Finally



We've whined for a long time. We hoped for a long time. We dreamed for god knows how long. But now it's time. Myself and Spetsnaz finally crossed the line and gotten ourselves a GeForce 4 based graphics card. I left the GF2MX stage, but poor Spetsnaz even had to cope with a TNT2! Well, at least it was an Ultra.

The GeForce 4 Ti chipset needs little introduction, but for the sake of completeness, let's briefly go over its most important features.
The GeForce 4 Ti series, codenamed NV25, is the inheritor of the GF3 series, codenamed NV20. While inherently they are of the same technological generation (much like TNT/TNT2 and GF/GF2), the GeForce 4 Ti series (the MX series or NV17 is a GF2 with GF4 architecture but lacks the latter's core technologies) has some improvements which add up to substantially greater efficiency and raw speed.

The GF4 Ti is built around the same nFiniteFX engine, but now ships with 2 vertex shader units, instead of one with the GF3. Also, the parallel units have been finetuned to decrease build-up latencies. "Fine-tune" btw is perhaps a keyword in describing the GF4.
Another improvement can be found in the update of LMA - LMA II - which we also can find on the GF4 MX series. Without going into the matter too deeply (too much technical garble) we can soundly say that all the breakthroughs made in the GF3 memory architecture have been refined: Z-Occlusion Culling for example, which does not render masked pixels, has been tweaked to include more pixels in the occlusion. Also another feature, the Memory Crossbar (which split the memory in smaller banks to make processing more efficient and to make maximum use of the available bandwith), has been improved with patches to the loading and prioritization schemes.
The GF4 also comes with nView (nVidia's answer to Matrox' Dualhead) but this is nothing new. Again, dual-display with Analog Monitor, Tv and DVI are possible. The GF4 ships with 2 350Mhz RAMDACs for improved display quality and refresh rates. Another improvement is the addition of Accuview, or simply put a tweak to the GeForce 3 set of anti-aliasing. The latter introduced HRAA, a new sampling procedure to anti-aliasing (as opposed to the "old" supersampling method found on GF2). In the GF4 the pixel offset in the sampling procedure has been shifted, and a new filtering technique everytime an AA sample is created, dramatically improve AA performance and quality, especially with GF3 introduced Quincunx AA. This AA mode would supposedly be as fast as 2xAA. Also, Accuview allows anisotropic filtering.

The GeForce 4 Ti 4200 is the "little big sister" of the GF4Ti series, and was eagerly anticipated because of very attractive performance (better than GF3 Ti500) and pricing (64MB version under 200 eur). After the disappointing release of the GF3 Ti200 (too expensive and substantially slower than GF3 and GF3 Ti500) the GeForce 4 Ti4200 is certainly the best card you can buy right now. The card ships in both 64MB and 128MB version. The latter has slower memory clock speeds, but in light of the next generation of graphical applications (Unreal Warfare engine, Doom III, Age of Mythology) a larger head-room of memory is preferred.

>> The Cards