The Graphics Chip

The G70 chip represents NVIDIA's most ambitious to date. Here are the main details of it:

Chip Name G70
Silicon Process 110nm (TSMC)
Transistors 300M
Die Size 334mm²
[18.65mm (w) x 17.9mm (h)]
Packaging Flipchip
Pipeline Configuration 24 / 16 / 32
(Textures / Pixels / Z Samples per clock)
Memory Interface 256-bit (64x4 Crossbar)
DDR, GDDR-2, GDDR-3
DirectX Capability DX9.0 - VS3.0, PS3.0
Display Dual 400MHz RAMDAC's
Host Interface PCIe X16

When we look at the configuration of the chip and the process its based on, this does appear to be very similar to what the rumours were suggesting for NV47. With the increased number of pipelines, adding an extra 50% of fragment pipelines in relation to NV40, let alone the per pipeline improvements and other increases, the number of transistors has grown to over 300M, but with only a move to 110nm the die size is even larger than that of the NV40 chip, reducing the number of die per wafer. It will be interesting to see what other board configurations appear on the G70 chip and what numbers of pipelines are enabled. As is likely to be the case for all high end graphics chips released from here on, the host interface used on this board is PCI Express; NVIDIA aren't yet announcing any AGP configurations, however it is possible to achieve by the use of their "HSI" bridge chip, as was the case with NV45 (being an NV40 chip bridged to PCI Express).

Here is an overview of the features made available by the G70 chip:

  • CineFX 4.0 Architecture
    • Full DirectX9 Support
    • DirectX9 Shader Model 3.0 Support
    • Vertex Shader 3.0
      • Vertex Shader 3.0
      • Pixel Shader 3.0
      • Internal 128-bit Floating Point (FP32) Precisions
    • Unlimited Shader Lengths
    • Up to 16 textures per pass
    • Support for FP16 Texture Formats with Filtering, FP32 without
    • Non-Power of two texture support
    • Multiple Render Targets
  • NVIDIA High Precision Dynamic Range Technology
    • Full FP16 Floating Point Support throughout the entire pipeline
    • FP16 Floating Point Frame Buffer Support
  • Intellisample 4.0
    • Up to 4X, Gamma Adjusted, Native Multi-sampling FSAA with rotated grid sampling
    • Transparent MultiSampling and SuperSampling
    • Lossless color, texture, z-data compression
    • Fast Z Clear
    • Up to 16x Anisotropic Filtering
  • UltraShadow Technology
  • NVIDIA SLI Support
  • NVIDIA Pure Video Technology
    • Adaptable Programmable video processor
    • High Definition MPEG2 and WMV9 acceleration
    • Spatial Temporal de-interlacing
    • Inverse 2:2 and 3:2 pull-down (Inverse Telecine)
    • 4-tap horizontal, 5-tap vertical scaling
    • Overlay color temperature correction
    • Microsoft® Video Mixing Renderer (VMR) supports multiple video windows with full video quality and features in each window
    • Integrated HDTV output
  • Advanced Display options
    • Dedicated on-chip video processor
    • nView Multi Display technology
    • Digital Vibrance Control 3.0