InsaneTek

Gigabyte 7800 GTX GV-NX78X256V-B Continued

The 7800 GTX features Dual DVI for the graphical gurus out there who need the extra monitor space with dual LCD's.  The 7800 GTX is powered by a single PCI-E connector, which is supplied with most mainstream power supplies, such as OCZ's PowerStream series and SilverStone's Zeus series. 

I decided to take apart the 7800 GTX and a Leadtek 6800 GT to compare the change in die size.  The top bare video card is the 6800 GT core, while the bottom is the 7800 GTX core.  As you can see, the 7800 GTX is a bit larger in comparison to the 6800 GT core, which is the same size as the 6800 Ultra core.  As previously stated in the G70 section, the 7800 GTX core boasts more transistors than even an FX-57.  Looking at the back and front ram IC placements, you can see there are empty solder ball grids to fit extra ICs for more size.  Can you guess the availability of DDR2 512MB 7800 GTX's in the future?  This amount of graphics memory coupled with the power of the 7800 GTX should make for a very powerful combination.

Here, you can see how much larger the Gigabyte 7800 GTX is in comparison (from bottom to top) to the 6600 GT, the X800XL, and the 6800 GT.  These cards were pictured for size purposes, but I will also add the X850 XT PE for benchmarking purposes.  The X850 XT PE is similar in size to the pictured X800XL. 

Testing Setup

Test Setup and Drivers:

• Athlon 64 4000+ (234HT x 12 multi = 2800 Mhz)
• 2 x 512MB PC3200 Centon Advanced
• Abit AN8 Ultra SLI
• 2 x 80GB Hitachi SATA II Hard Drives
• OCZ PowerStream 520W
• Viewsonic V90F

Graphics Cards used for Testing:

• Connect3D Radeon X850XT PE
• 2 x MSI 6600 GT's in SLI mode
• 2 x Leadtek Winfast 6800 GT's in SLI mode
• Leadtek Winfast 6800 GT
• ATI Radeon X800XL

Drivers

• Nvidia Forceware 77.77
• ATI Catalyst 5.8

To eliminate as much bottlenecking from my system as possible, I overclocked the 4000+ to 2.8 GHz and verified its stability with Prime95 and SuperPi.  I utilized the latest official drivers from both ATI's and Nvidia's website and ensured that when I switched drivers, all traces of each driver were eliminated by deleting the appropriate files, folders, and registry settings in Windows.  All graphics cards shown for testing were kept at their specified stock speeds and were never overclocked.  I avoided the use of budget cards, such as the 6600 GT by itself, the X700, and any graphics card of similar performance or below, considering the 7800 GTX is targeted towards the high-performance and high-end market.  The monitor used was a Viewsonic V90F monitor, although I tested only up to 1600x1200, since anything higher was not possible with this monitor.  I kept all testing resolutions at above 1280x1024, since today's high-end video card offerings are meant to run at high resolutions.  For 4x anti-aliasing and 16x anisotropic filtering enabled, I tested only at 1024x768 resolution and above.  We are moving into an era where resolutions of 1024x768 are no longer an option in the minds of a high-end gamer and are only used for testing purposes.  I only tested at settings used in a typical real-world situation, which is why I chose 16x anisotropic filtering always, as the change from 8x to 16x posed a minimal difference in performance. 

Testing:

In order to capture as much real-world gaming performance as possible, I tried to make sure all time demos and custom demos were best indicative of real world performance.  For the games that already included adequate time demos that reflect real world game play, such as the demos of Doom 3 and Unreal Tournament, I utilized those existing demos for testing.  For ones that either did not come with adequate time demos or good enough time demos, such as Half-Life 2 or Battlefield 2, I created custom time demos myself to make sure real-world game play was captured.  These demos will be available shortly for download in case any users would like to use them for their own testing purposes.  I will discuss more in detail each testing procedure in the benchmark descriptions.