nVIDIA GeForce 4 Ti and GeForce 4 MX

Today nVIDIA unveils their latest range of videocards, consisting of two chip variants called GeForce4 MX (low end) and GeForce4 Ti (high end). Without much further ado let’s jump in and look at the specs of these…

GeForce4 Ti

    3D Feature set:

  • 'nfiniteFX II' Engine
    • Dual Vertex Shaders
    • Programmable Pixel Shaders
      • Phong Style Lighting
      • Dot 3 Bump mapping
      • EMBM
      • Anisotropic Lighting
      • Procedural Textures
      • Per Pixel Reflections
      • Z Correct Bump Mapping
  • Shadow Buffers
  • DirectX and OpenGL S3TC Texture Compression
  • Cubic Environment Mapping
  • 'Accuview' Antialiasing
  • 'LightSpeed Memory Architecture II'

    2D Feature set:

  • 'nView' Display Technology
    • Dual integrated 350Mhz RAMDAC's
    • Dual Channel TDMS transmitters
  • HDTV/DVD Playback Processor

nVIDIA are launching two models based on this “GeForce4 Ti” processor, namely Ti 4600 and Ti 4400. Note that nVIDIA are now listing 'AA Samples' rather than fillrate in Pixels / Texels per second, I've supplied what I would assume the relevant Pixel/Texel fillrates would be (for reference the specs of the highest GeForce3 are included):

  GeForce4 Ti 4600 GeForce4 Ti 4400 GeForce3 Ti 500
AA Sample Fillrate 
(Pixel Fillrate/Texel Fillrate)
4.8 Billion
(1.2Gpps/2.4Gtps)
4.4 Billion
(1.1Gpps/2.2Gtps)
3.84 Billion
(0.96Gpps/1.92Gtps)
Triangles / Sec 136 Million 125 Million 55 Million
Memory Bandwidth 10.4GB/sec 8.8GB/sec 8.0GB
Max Memory 128MB 128MB 128MB

nVIDIA doesn't mention the configuration of the chip in their documentation, which indicates that it hasn't changed from the GeForce3. This being the case its possible that we can extrapolate a core clock speed of 300Mhz for Ti 4600 and 275Mhz for Ti 4400, with DDR memory clocks of 325Mhz and 275Mhz respectively. If we are to believe any of the information leaked prior to the launch then it sounds as though the board vendors may have some leeway with regards to chip and memory configurations.

GeForce4 MX

    3D Feature set:

  • nVIDIA Shading Rasteriser (NSR)
  • Integrated Hardware Transformation & Lighting Engine
  • 4 texture-mapped, filter, lit texels per clock cycle
  • DirectX and OpenGL S3TC Texture Compression
  • Cubic Environment Mapping
  • 'Accuview' Antialiasing
  • 'LightSpeed Memory Architecture II'

    2D Feature set:

  • 'nView' Display Technology
    • Dual integrated 350Mhz RAMDAC's
    • Dual Channel TDMS transmitters
  • Video Processing Engine (VPE)

The MX cards are split into three different models, as follows (This time GeForce2 MX is included for reference):

 

GeForce4 MX 460

GeForce4 MX 440 GeForce4 MX 420 GeForce2 MX 400
AA Sample Fillrate 
(Pixel Fillrate/Texel Fillrate)

2.4Billion
(0.6Gpps/1.2Gtps)

2.2Billion
(0.55Gpps/1.1Gtps)

2Billion
(0.5Gpps/1Gtps)

0.4Billion*
(0.4Gpps/0.8Gtps)
Triangles / Sec 38 Million 34 Million 31 Million 25 Million
Memory Bandwidth 8.8GB/sec 6.4GB/sec 2.7GB/sec 2.7GB/sec
Max Memory 64MB 64MB 64MB 128MB

*Note: GeForce2 MX used supersampling so this is not a direct comparison bearing in mind the differing methods between the two.

The 4 texels per clock specification indicates that the pipeline setup of GeForce4 MX is similar to the older GeForce2 MX's with two pixel pipes each with two texture units. This being the case the core speeds and memory would be consistent with the following rates (core/memory): MX 460 300Mhz/275Mhz DDR, MX 440 275Mhz/200Mhz DDR, MX 420 250Mhz/166Mhz SDR.

The inclusion of 'NSR' and T&L would indicate that the 3D feature set of the GeForce4 MX remains at the same level as the older MX range, however the massive speed improvements, in most cases, and the 'Light speed Memory Architecture II' should make for considerable performance improvements over its predecessor. The inclusion of nVIDIA's new antialiasing schemes should make game play with FSAA enabled far more of a realistic prospect to the low end gamer than it previously has been.