It's Official!  NVIDIA's NV30 is Now the GeForce FX!
Will we soon be crowning a new 3D King, again?

By - Marco Chiappetta
November 18, 2002

Virtually all of the companies actively producing 3D chipsets today, have been working toward one goal, to deliver a product that can render interactive, cinematic quality 3D images, in real-time on the desktop, at smooth, playable frame rates.  One company in particular, NVIDIA, has been telling us their eventual goal is to bring "Toy Story" quality graphics to PC games.  Today, NVIDIA is making strides towards that eventual goal with the introduction of their GeForce FX GPU, formerly known as the NV30.  The GeForce FX is NVIDIA's DirectX 9 compliant Graphics Processor targeted at ATi's flagship R300, which powers the RADEON 9700 Pro.  Bringing the GeForce FX to market has been relatively strenuous endeavor for NVIDIA.  It has been almost a year since the company, who "popularized" the 6-Month product cycle, introduced their GeForce 4 line of GPUs.  The NV30's introduction had been delayed due to unforeseen difficulties resulting from the switch to a .13u manufacturing process.  From what we have seen however, the wait will be well worth it.  To put it simply, the GeForce FX is a monster.  We spent every second of the live demo that NVIDIA gave us yesterday, with our eyes glued to the monitor, slack jawed.  On the next few pages we have some screenshots from the demos we saw running on the GeForce FX.  You'll see why we were so impressed when you get there.  But first, let's run down the hardware that is the GeForce FX.

Specifications & Features of the NVIDIA GeForce FX
At Least What We Know So Far...

     
CLICK ANY IMAGE FOR AN ENLARGED VIEW
GPU:
  • .13u Manufacturing Process
  • 125 Million Transistors ( 2X GeForce 4)
  • 256-Bit GPU
  • Flip-Chip BGA Package with copper interconnects
  • 8 Pixel Per Clock (8 Pixel Pipelines)
  • 1 TMU Per Pipe (16 Textures per unit)
  • > 500MHz Core Clock (Will probably vary depending on model)
  • 350 Million Triangles per Second - 3x The Geometry Performance of a GF4 Ti
  • AGP8X (2.1GB/s bandwidth)

Memory:

  • Memory with >1GHz Data Rate (500MHz DDR)
  • 128-Bit "DDRII" Type
  • 128MB & 256MB Memory Capacity
  • 3rd. Generation Lightspeed Memory Architecture

  • 48GB/s Effective bandwidth through the use of compression techniques.  (16GB/s actual @ 500MHz)

Other:
  • Full DX9 Compliance (and more)
  • 64-Bit Floating-Point Color
  • 128-Bit Floating-Point Color
  • 2 x 400MHz Internal RAMDACs
  • Long Program length for Pixel and Vertex Shading
  • Conditional Execution for both Pixel and Vertex
  • True Data-Dependant at Vertex
  • Unified Vertex and Pixel Shading instruction set
  • Unified Driver Architecture
  • nView 2.0 - Multi-Display Technology
  • Digital Vibrance Control 3.0
 


Unfortunately, we do not have a complete list of features and specifications for the GeForce FX, just yet.  Listed above is information we have taken directly from the NVIDIA technical briefs, outlining some of the new GeForce's capabilities.  We also grabbed a few charts that compare the features of NVIDIA's newest GPU to ATi's R300 and to the Microsoft DirectX 8 specification.

CineFX - VERY HIGH COLOR PRECISION

The GeForce FX is capable of rendering images with 128-Bit floating point color precision.  This is the same level of color precision used in movies today.  As you can see in the above chart, the GeForce FX's 128-Bit max precision is much higher than the 96-Bit output of the R300, and far surpasses the requirements for DX8.  It's important to note that having floating point precision throughout the rendering pipeline will yield much smoother color gradients, especially when darker colors are used.

PIXEL SHADER 2.0+

The ATi R300 strictly adheres to Microsoft's DX9 Pixel Shader 2.0 specifications, which are obviously superior to what we had with DirectX 8.  The GeForce FX however, takes things much further, offering a maximum of 1,024 texture and color instructions per shader, versus 32 and 64 instructions on the R300.  The GeForce FX also offers over 5X the temp storage of the R300, and overall offers more pixel shader functionality.

VERTEX SHADER 2.0+

The GeForce FX can also handle much more complex vertex shaders than the R300, and surpass the DX9 specification.  In fact, the GeForce FX can execute 64X the number of vertex shader instructions than ATI's R300, has 4 more temporary registers and maximum of 256 loops versus the R300's 4 loops.  NVIDIA's GPU also offers more vertex shader functionality.  To produce the kind of images used in movies like Toy Story, or Final Fantasy, very complex pixel and vertex shaders have to be used.  With the GeForce FX, NVIDIA gives developers the kind of power they need to replicate that level of realism.  To make it easier for developers to utilize the GeForce FX's capabilities, NVIDIA developed their Cg ("C" for graphics) language and compilers.  Cg is high level programming language, that will reduce the amount a coding developers will have to do considerably.  For a deeper look into the inner workings of Cg, take a look at our interview with David Kirk, right here.

A Peek At The Hardware & Some Screen Shots...