High-res gaming - 8800 Ultra in action
Author: Luka Rakamaric
Date: 14 Jun 2007

A few weeks ago NVIDIA introduced a new and improved 8800GTX, labeled 8800Ultra. It is based on a new, A3 revision of the G80 GPU, and has increased clocks. NVIDIA has perhaps jumped the gun, because the R600 threat turned out not to be so great, as a stock 8800GTX beats it in almost all benchmarks by a fair margin. This way, NVIDIA introduced a card with a price that has little basis in performance increase over the GTX. A lot of reviews criticized the new Ultra as expensive, which is not very great when your competitor is introducing a card and emphasizes its pretty good price performance ratio. However, flagship models seldom have a good price/performance ratio. In this article we will present benchmarking results of an SLI setup of these monsters compared to a single card, to see if even the highest resolutions and AA AF settings can make use of so much processing power.

The 8800Ultra is very similar to the GTX, it has the same format, PCB layout, but the GPU is different. The new GPU is called G80-450-A3, and is clocked at 612 MHz, an increase of 37 MHz over the GTX. The memory received a bigger boost, from 900MHz to 1080 MHz, which equals exactly 20% increase, not a small amount. All this is accompanied by an improved cooler, which has the fan displaced to the top of the card for better airflow. Since NVIDIA managed to tweak the new revision to require a little less power, the temperature remained the same as with the GTX, despite higher clocks. The card still requires "only" two six pin PCI-E power connectors. Unlike the R600, which has one 8 pin and one 6 pin, meaning that 3x75W specification is not enough, the Ultra fits into these limitations. This is also an important marketing fact, because NVIDIA can boast that its flagship actually consumes less power than the performance wise inferior competitive model. You can read more about Ultra's architecture and performance comparisons in our launch day review here.