|
Page 2 of 8
AMD's goals and strategy
Before we go anywhere near this new GPU, let's start by discussing what AMD looked to achieve for this generation of graphics boards, and exactly how they went about working on it. First up, let's take a peek at AMD's goals straight from the horse's mouth.
With Windows 7 so close at hand, AMD's primary goal for this generation was an obvious one - To implement DirectX 11's feature set in full, and before it's actually available in retail. Of course, focusing on DirectX 11 support alone would be foolish given the huge ecosystem of DirectX 9, 10 and 10.1 titles both already on the market and continuing to be developed, and so maximising performance for titles based around these revisions of the API was also high on the list.
Having taken a bit of stick for their "Stream" technology both from the press and elsewhere in recent times, and with both OpenCL and DirectX 11's DirectCompute functionality hitting the market, building up a GPU which doubles as a top-notch stream computing platform was also very much on the agenda.
Increasing performance is pretty much a given with any next-generation graphics board, but AMD's specific target for "Cypress" was to double performance over the previous generation, yet without drastically increasing power consumption. Finally, innovation was also on the board, something which AMD feel they've achieved via their Eyefinity technology, which we've already discussed in detail in its own right here.
No arguments about those design goals, but how did AMD look to achieve them? Quite simply, by following the same "sweet spot" strategy that served them so well throughout the life of their Radeon HD 4000 series. In essence, this execution strategy eschews the idea of creating a huge, monolithic GPU for the high-end, and then shrinking, cutting down and reducing that design for lower price points, instead preferring to initially target a certain price point as a sweet spot where most high-end sales are focused (although it's worthy noting that the "sweet spot" for the Radeon HD 5800 series is higher than that chosen for the Radeon HD 4800 launch) for your initial design. From here, you can batch together multiple GPUs to target the outright enthusiast market, while also scaling down your design for lower price points.
Thus, today sees the launch of that "sweet spot" GPU in the form of AMD's Cypress design and the Radeon HD 5800 series - This will then be followed up in the fourth quarter by a part codenamed "Hemlock" which makes use of two Radeon HD 5800 series class graphics cores, while a cut-down version of the same Cypress architecture (named "Juniper") will shoot for lower performance-centric price points before finally dropping towards the mainstream and low-end via "Redwood" and "Cedar" codenamed GPUs. Whether this will give AMD the leadership in every market segment that they hope for remains to be seen of course, but given NVIDIA's rumoured struggles with their monolithic, DirectX 11 GT300 core it's hard to argue with this kind of roadmap given how well AMD have (at this juncture at least) executed it.
Windows 7
Aside from those design goals we've just outlined, there's one other obvious driver for AMD's latest architecture aside from DirectX 11 alone - The launch of Windows 7. While this isn't the huge leap in terms of driver requirements and the like as the move from Windows XP to Windows Vista, it's still important for ATI's graphics division to be ready and raring to go on the day that Microsoft's latest Operating System hits retail shelves.
The most obvious of ATI's Windows 7 requirements is the ability to accelerate the OSes Aero Glass desktop - It probably goes without saying that all of the company's DirectX 9 parts and upwards should be perfectly capable of doing this, although DirectX 10 and 11 boards will have a slight advantage with their ability to make use of the full Windows 7 Aero feature set. Away from the desktop, AMD will also be supporting Windows 7's native video transcoding capabilities via their ATI Stream technology, which should allow users to drag and drop videos to portable devices (aside from Zunes and iPods at present, ironically and due to the way that Windows 7 works with those devices) and see them encoded quickly to that portable device by leveraging the power of the GPU to perform the task.
AMD have already announced that they have WHQL certified, DirectCompute 11 capable drivers ready to go, and thus their Catalyst 9.10 release should see these drivers become available in time for Windows 7's October release.
AMD are also promising performance improvements to some degree by using their new architecture under Windows 7 over Windows Vista, and above you can see today's first set of AMD-generated Radeon HD 5870 performance numbers, comparing performance under these two Operating Systems.
Using two GPUs in CrossFire promises some even more impressive performance improvements, if AMD's numbers are to be believed. So, do AMD believe that they're well and truly ready for Windows 7? You could say that and you'd be putting it mildly, yes.
|
