AMD Radeon HD 7970
Posted on: 12/23/2011 02:31 AM
Today we are checking out the new AMD Radeon HD 7970 graphics card that is targeting the $549 price range. AMD hopes to take down the Nvidia GeForce GTX 580, which is currently the fastest single-GPU graphics card money can buy. In order to do so they have designed a new architecture which uses the 28nm design process...
It has been a year since AMD released the Radeon HD 6970 which has since served as their flagship single-GPU graphics card. With an MSRP of $369 it was only slightly more expensive than the GeForce GTX 570, while at times could be found challenging the much more expensive GTX 580.
Although the Radeon HD 6970 did present gamers with a pretty good bang for their buck and it did support the latest rendering technologies, such as DirectX 11, OpenGL 4.2 and OpenCL 1.2, it was based on an aging architecture. That architecture was of course VLIW, the Radeon HD 6000 series used the VLIW4 revision.
It has been many years since AMD has made the kind of changes that they have with the Radeon HD 7000 series, namely the high-end models. AMD has implemented their GCN (Graphics Core Next) architecture which can handle graphical and computing tasks with a high level of efficiency.
Not only has AMD made significant design changes with the Radeon HD 7970, but it is also the first graphics card to utilize the 28nm design process as well as offer PCI Express 3.0 support. Both are noteworthy features, however the move to the 28nm process means that AMD were able to develop by far their most complex GPU yet, in fact the Radeon HD 7970 architecture, which has been codenamed Tahiti XT, is the most complex we have seen yet.
Comprised of an incredible 4.3 billion transistors, the Tahiti XT boasts over 40% more transistors than the GeForce GTX 580. Perhaps even more impressive than that is the fact that there are 1.7 billion or 65% more transistors than that of the Radeon HD 6970 and yet the die size is 6% smaller.
Whereas the Radeon HD 6970 featured 1536 SPUs, 96 TAUs, and 32 ROPs, the new Radeon HD 7970 boasts 2048 SPUs, 128 TAUs, and the same 32 ROPs. That means it has 33% more SPUs and TAUs, while the amount of ROPs remains the same.
The GPU core is clocked at 925MHz, 5% higher than the Radeon HD 6970. The GDDR5 memory on the other hand operates at the same frequency of 1375MHz (5.5GHz DDR). Still, pairing that frequency with a 384-bit wide memory bus gives the Radeon HD 7970 264GB/s of theoretical bandwidth, a 50% advantage over the HD 6970.
Furthermore, while the Radeon HD 6970 came loaded with a 2GB frame buffer, the 7970 has been upgraded to a larger 3GB buffer. All this sounds great, as it should make for one very fast high-end graphics card. However unlike previous generations AMD will not be as aggressive on pricing and at this point they might not need to be if the Radeon HD 7970 performs as expected.
Therefore the MSRP has been set at $549, making the Radeon HD 7970 almost $200 more expensive than the Radeon HD 6970 when it launched. This is roughly the same price early adopters paid for the GeForce GTX 580, so already that says to us that AMD is very confident in what they have created. With that said, letís move on to check out the performance results...
Printed from Legion Hardware (http://www.legionhardware.com/articles_pages/amd_radeon_hd_7970,1.html)