Over the years NVIDIA has used a wide range of letter combinations and words to designate their products. For instance, many NVIDIA enthusiasts on a budget have flocked to NVIDIA’s “GT” line, starting with the GeForce 6800 GT and a few months later, the GeForce 6600 GT. Both of these cards dominated their respective segments of the graphics market thanks to their extraordinary combination of price/performance. The GeForce 6800 GT delivered roughly 90% of the performance of NVIDIA’s flagship Ultra card, but at a substantially lower price, while the 6600 GT offered performance superior to the previous generation of high-end DX9 cards like the Radeon 9700/9800 Pro with the addition of shader model 3.0 and a $200 price tag.
While the GT has represented price/performance, the Ultra line has historically been NVIDIA’s flagship graphics card. It’s the card for the gamer who craves the fastest graphics board money can buy. NVIDIA has a history of Ultra cards dating all the way back to the TNT2 Ultra days eight years ago. These cards are the exact opposite of being good values though, with price tags that have always been high, but that hasn’t stopped NVIDIA from producing them, and people from buying them: if you asked the average enthusiast, they’d probably tell you the last thing NVIDIA needs right now is a higher-end GeForce 8800 card, but here we are anyway.
Born out of the need to compete with the perceived threat of ATI’s upcoming R600 GPU, the GeForce 8800 Ultra isn’t an all-new graphics chip, it’s actually based largely on the same G80 GPU that’s already been inside the GeForce 8800 GTX and GeForce 8800 GTS lines for the past 6 months now. Let’s take a look at the specs:
|GeForce 8800 Specification Comparison|
|Specification||GeForce 8800 Ultra||GeForce 8800 GTX||GeForce 8800 GTS 640MB|
|# of Transistors||681 million||681 million||681 million|
|Shader Clock (Stream Processors)||1500MHz||1350MHz||1200MHz|
|# of Stream Processors||128||128||96|
|Memory Clock Speed||1080MHz actual (2.16GHz effective)||900MHz actual (1.8GHz effective)||800MHz actual (1.6GHz effective)|
|Memory Bandwidth||103.7 GB/sec||86.4 GB/sec||64 GB/sec|
|# of ROPs||24||24||20|
|Texture Fill-Rate||39.2 Gigatexels/sec||36.8 Gigatexels/sec||24 Gigatexels/sec|
|Max Board Power||175||177||147|
|Video Processor||Video Processor 1||Video Processor 1||Video Processor 1|
As you can see, the GeForce 8800 Ultra shares the same basic architectural features as the GeForce 8800 GTX – its got the same number of transistors, the same 384-bit memory interface with 768MB of memory, and the same number of stream processors, the only difference lies in the clock speeds and max power consumption. Thanks to its tweaked 90-nm manufacturing process, the GPU actually consumes slightly less power than the GTX according to NVIDIA. NVIDIA isn’t saying much on what’s changed with their revised 90-nm process, only that they’ve developed a new silicon revision specifically for the GeForce 8800 Ultra that has been designed to improve the clocks while keeping power to a minimum. On the performance side, the stream processors are clocked 10% faster, while memory bandwidth jumps 20% to over 100GB/sec.
Unfortunately the video portion in GeForce 8800 Ultra carries over unchanged: the GeForce 8800 Ultra uses the same PureVideo HD video processing engine currently used in today’s GeForce 8800 cards. GeForce 8600/8500 GPUs feature a newer video processor that offloads more video functions, including performing full H.264 decoding.
NVIDIA has come up with a revised cooling unit for the new card however. On the next page we’ll take a look at the thermals of the new GeForce 8800 Ultra card…