With nVidia done with its high-end product launches for this generation, it was obviously time for AMD to fire back. Now most of you would remember that AMD’s previous generation of GPUs were less than stellar and about the only product that got anyone excited was the X2. Now since the new RV770 is based on the same R600 roots, we wondered what AMD was able to do to the core to turn it around.

The simple, one-word answer to that is efficiency. AMD did this by reducing the size of the logic blocks on the GPU without reducing any functionality. Now with more space to play with AMD, added more of everything to increase the performance of the GPU. The second thing AMD did was to make sure that all resources were being fully utilized by improving the data flow within the GPU.

The RV770 has ten SIMD cores with each one containing sixteen stream processor units. Each of those SP units is a superscalar processing block made up of five ALUs. Do the math (10 x 16 x 5) and you come up with a total of 800 ALUs onboard, which AMD advertises as 800 stream processors— certainly a lot more than the 320 SPs present in AMD’s previous-gen RV670 GPU. All this power results in a theoretical peak of a whopping 1.2 teraflops on the Radeon HD 4870. The following are the official specification for the HD 4870.

GPU Core Name: RV770
Fabrication Process: 55 nm
Memory Size: 512Mb GDDR5
Core Clock: 750 MHz
Memory Clock: 750MHZ
Stream Processors: 800
Texture Fillrate: 30 Gtexel/s

With great power, comes great responsibility and the HD 4870 certainly has a lot to prove. We got a couple of units from Sapphire and while Crossfire benching is on its way, lets find out how well a single RV 770 compares to nVidia’s latest offerings but lets take a look at the card itself first.