NVIDIA GeForce FX Derivatives

The GeForce FX 5200 & 5600 Debut!
The NV34 and NV31 Officially Unveiled

By - Marco Chiappetta
March 6, 2003

In November '02, during COMDEX in Las Vegas, NVIDIA introduced their eagerly anticipated GeForce FX GPU, formerly codenamed NV30.  The high-end GeForce FX part on display at COMDEX, was NVIDIA's answer to ATi's DirectX9 compliant R300, the core that powers the very popular Radeon 9700 and 9500 product lines.  At the time, NVIDIA did not disclose final clock speeds, or specifications for the inevitable GeForce FX derivatives, but we all knew they would eventually come.  Then, in late January we learned that the high-end part in the GeForce FX lineup would be branded the 5800 Ultra.  The GeForce FX 5800 Ultra ships with a core clock speed of 500MHz, with DDR-II type memory clocked at an effective 1GHz data rate.  The "non-Ultra" GeForce FX 5800 ships with lower core and memory speeds, but it is essentially the exact same part.

Today, NVIDIA is unveiling three new products in the GeForce FX product line, the 5200, the 5200 Ultra and the 5600 Ultra.  These three new products are based on GPUs derived from the "original" NV30 core, but with some architectural differences that reduce costs, and ultimately performance.  The GeForce FX 5200 parts are based on the NV34 core, while the GeForce FX 5600 Ultra is based on the NV31.  With these new parts, NVIDIA is trying to entice the mainstream consumers that comprise the vast majority of the buying public.  The excitement and buzz generated by top-of-the-line products like the GeForce FX 5800 Ultra is intriguing, but the bottom line is, sales of these high-end products account for only a small percentage of overall revenue potential.  The big money is made by selling millions of "mainstream" parts to OEMs and PC users who don't care much about having the latest and greatest.  The new GeForce FX 5200, 5200 Ultra and 5600 Ultra give NVIDIA the ability to offer products with full DirectX9 capabilities to all market segments.  Unlike last year's GeForce 4 MX release, which garnered some negative press because they lacked any hardware pixel or vertex shaders, NVIDIA's new mainstream products don't skimp on features.  The NV31 and NV34 cores have full GeForce FX feature support, including DirectX9 pixel and vertex shaders 2.0+.  What do you say we take a closer look?

Specifications & Features of the NVIDIA GeForce FX 5200 GPUs DX9 For All!

     
CLICK ANY IMAGE FOR AN ENLARGED VIEW

CHART TAKEN FROM NVIDIA TECHNICAL BRIEF

The first cards we'll talk about are the GeForce FX 5200 and 5200 Ultra, the products based on the GPU formerly codenamed NV34.  As you can see by the above feature list and comparison chart, the NV34 employs NVIDIA's "CineFX" engine, offers support for DirectX 9, AGP8X and has dual integrated 350MHz RAMDACs, which means they are capable of powering dual independent displays.  (For a more complete look at the full capabilities of the GeForce FX GPU, click here.)  Unlike the other GeForce FX products though, the 5200s will be built using TSMC's .15μ manufacturing process.  Even though the GeForce FX 5200s aren't being built using the more advanced .13μ process, they will still be able to hit fairly high clock speeds without generating excessive amounts of heat, because far fewer transistors are needed to build this GPU.  In fact, the 5200 is comprised of "only" 45 million transistors, making it roughly 64% smaller than the 125 million transistor GeForce FX 5800!  This relatively low transistor count, coupled with the fact that the more mature .15μ manufacturing process is being used, should mean NVIDIA will be able to quickly ramp up production of the GeForce FX 5200s.  The low transistor count consequently results in lower over all power consumption and heat radiation as well.

Lower heat output means less sophisticated cooling is needed to keep the GPU operating within its temperature limits.  According to the latest information given to us by NVIDIA, the GeForce FX 5200 Ultra should ship with a core clock speed of 325MHz, with BGA DDR-I memory also clocked at 325MHz (650MHz DDR).  The 5200 Ultra GPU is cooled using a single-slot, active solution similar to what is found on the GeForce 4 Ti line of products.  Final clock speeds for the GeForce FX 5200 "non-Ultra" have not been disclosed just yet, but considering it is passively cooled, and does not require a secondary power connection, we'd guess the core clock speed will fall within the 225-275MHz range.  The GeForce FX 5200 also uses standard TSOP memory packaging versus BGAs, so the memory will probably be clocked much lower as well.  Expect the non-Ultra 5200s to ship with memory speeds hovering around 400MHz DDR.

Clock speeds are not the only things that differentiate the GeForce FX 5200s from the other products in the FX line-up.  Obviously, with a transistor count so much lower than the GeForce FX 5800, something had to be removed from the die.  In the case of the GeForce FX 5200s, the color and z-compression logic used in the Intellisample Antialiasing engine has been removed, along with four of the pixel pipelines.  These measures mean less efficient use of memory bandwidth, and lower fillrates, but the performance of these cards should still be quite good, especially considering their projected price points.  As you'll see later, NVIDIA is claiming performance that is twice as high as the GeForce 4 MX and Radeon 9000.

A Peek At The 5600 Ultra...