XFX GeForce GTX260 XXX Edition Video Card Review :: XFX GeForce GTX260 Features

• 65 nanometer process from TSMC
• 1.4 Billion transistors
• 896MB DDR3 memory
• 192 Stream Processors
• DirectX 10.0
• PhysX
• CUDA
• NVIDIA PureVideo technology
• 64 Texture Address units
• 64 Filtering units
• 28 ROPs
• 640MHz Core Clock (576MHz standard)
• 1150MHz Memory clock (999MHz standard)
• 1363MHz Shader Clock (1242MHz Standard)
• 448-bit memory interface
• $329 on Newegg

Brand Name	XFX
Part Number	GX260NADDU
Graphics Chip	GT200
Core clock	640MHz
Shader Clock	1363MHz
SPs	192
Fabrication Process	65 nm
Transistors	1.4 Billion
Memory cloick	1150MHz (2230MHz effective)
Memory Interface	256-bit
Memory bandwidth	128.8 GB/second
Memory Size	896MB
ROPs	28
Texture Filtering Units	56
Texture Filtering Rate	44.8 Gigatexels/second
HDCP Support	Yes
HDMI Support	Yes (via adapter)
Connectors	2x Dual-Link DVI, TV-Out
RAMDACs	400MHz
Bus	PCI Express 2.0
Form Factor	Dual Slot
Power Connectors	2x 6-pin

The GeForce GTX260 is the “little brother” to the GTX280 card that is NVIDIA's flagship at the moment. The card is built upon TSMC's 65 nanometer process, with 1.4 Billion transistors. This is the largest graphics card in terms of transistor count in the world. The GTX260 has 192 Stream Processors which is less than the 240 SPs the GTX280 sports.

NVIDIA was the first graphics chip maker on the market with DirectX 10.0 support with the GeForce 8800GTX and GeForce 8800GTS cards in 2006. The key features of DX10 include Unified Shader support, support for Pixel and Vertex Shader 4.0 and Windows Vista Aero Glass support (the GUI). Games that support DirectX 10 include: Crysis, Hellgate London, Lost Planet and many others that have hit the market in recent months.

CUDA is NVIDIA's computing initiative that allows developers to solve computational problems by using the multi-core parallel nature of the GPU. NVIDIA calls their GT200 chip a graphic card that goes beyond gaming into physics, math calculations and GPGPU. The GTX280 and 260 are the first NVIDIA cards to nearly reach the 1 TeraFLOP of computing power level. ATI HD4870 and 4850 are over 1 TFLOP already.

The GeForce GTX260 has 896MB of GDDR3 memory which seems like an odd number for a video card to have. The card has 7 1 Gigabit chips, bringing a total of 896MB. The memory on the XFX card is clocked at 1150MHz, or about 151MHz faster than the default clocked GTX260. The GTX260 has a memory bus of 448-bit, giving the card a memory bandwidth of 128.8GB/second instead of the 112GB/second that the GTX260 standard card has.

XFX has several versions of every card they sell. There is the standard reference clocked version, the XXX Edition and the Black Edition. The XXX Edition GTX260 has a clock speed of 640MHz in comparison to the 576MHz reference clock speed. The Shader clock speed is set to 1363MHz, compared to the 1242MHz reference clock. It should be an interesting comparison with the ATI HD4870 and GTX280 to see how the GTX260 performs and price/performance across the board with the higher clock speeds.

NVIDIA bought Ageia the makers of PhysX accelerators a few short months ago. Hardware accelerated physics allows such effects as real Fluid motion and other effects such as thousands of objects randomly being strewn apart by an explosion. NVIDIA has released a PhysX driver for their GeForce 9800GTX and GTX280/260 cards. The GPU score nearly stays the same but the CPU score jumps bringing the 3DMark over 1400 points higher. Unreal Tournament 3 also benefits from Hardware Physics by over 10% increase without PhysX when turning on Hardware Physics. Games using the Unreal 3 engine will also benefit from PhysX.