The Radeon X1950 XTX: DirectX 9ís last hurrah
It has been roughly four years since ATI first ushered in the era of DirectX 9 graphics with their Radeon 9700 Pro GPU. Most everyone would agree that the Radeon 9700 was one of the most, if not the most successful graphics product in ATIís history. Its R300 graphics core was notable for not only being the first DX9 GPU to hit the market, it also delivered performance that was unprecedented at that time, often offering double the performance of anything else on the market. The Radeon 9700 also laid the groundwork for a slew of other ATI products, including the Radeon 9500 and 9600 series, as well as its successor, the Radeon 9800 Pro and 9800 XT. The fact that there are hundreds of thousands, if not millions of gamers still happily gaming away on their Radeon 9xxx series cards to this day is a testament to the foundation laid by the Radeon 9700 Pro.
With this in mind, itís perhaps somewhat fitting that ATI has chosen to close the chapter on DirectX 9 with the imminent arrival of their last high-end DirectX 9 part, the Radeon X1950 XTX, just over four years to the day that the Radeon 9700 Pro was first introduced to the world.
So does the Radeon X1950 XTX live up to the legacy of previous DX9 cards from ATI? Letís find out!
R580+ graphics core
The Radeon X1950 XTX is built on ATIís new R580+ graphics core. R580+ is largely the same as R580, the chip is still based on TSMCís 90-nm manufacturing process and sports ATIís 48-pixel shader architecture, with 8 vertex shaders, 16 texture address units, 16 ROPs and contains 384 million transistors. The one key difference lies in R580+ís memory controller, which has been reworked to support GDDR2, GDDR3, and now GDDR4 memory.
GDDR4 is a new type of graphics memory thatís been designed to consume less power than GDDR3 when running at the same clock frequency. As a result, GDDR4 can be pushed to much higher clock speeds than GDDR3. The 1.1GHz GDDR4 modules used on the Radeon X1950 XTX can push up to 2.2Gbps per pin. In comparison, the fastest GDDR3 modules Samsung lists on their website top out at 900MHz (1.8Gbps/pin), while the Radeon X1900 XTX utilized 800MHz GDDR3 modules offering a max data rate of 1.6Gbps/pin.
The memory configuration itself of R580+ carries over unchanged, the chip boasts the same 256-bit (external) memory controller, with eight 32-bit memory controllers for greater efficiency.
The Radeon X1950 XTXís R580+ graphics core runs at the same speed as the X1900 XTX, 650MHz, while the boardís memory subsystem now runs at 1GHz (2GHz effective), versus 775MHz in the Radeon X1900 XTX. This change nets the Radeon X1950 XTX an additional 22% in peak memory bandwidth, up from 49.6GB/sec in Radeon X1900 XTX to 64GB/sec in the Radeon X1950 XTX.
In other words, other than the memory subsystem, the rest of the GPU carries over unchanged. But thatís not the only area ATIís improved on the X1900 XTX, as the Radeon X1950 XTX also sports a brand new cooling unit as well.