 |
|
|
|
|
|
| Barcelona and GPGPU architectures |
|
|
Hardware Upgrade: Can you give us some clues about the release dates of AMD’s new generation of processors for servers and desktops?
Giuseppe Amato: As for Opteron solutions, we have already made public the release date which is in September. At that time we will launch the new processors of the Barcelona family. As for the Phenom processors for desktops, the first samples of Phenom FX CPUs will be available by the end of the year, while the remaining versions of the family will arrive by March 2008.
Hardware Upgrade: So the Barcelona processor will be launched in around two months. Can you tell us some more information about the features of the samples that are currently available?
Giuseppe Amato: At the present moment our main customers have Opteron Barcelona CPU samples, which are begin used for the validation process and internal research for compatibility with existing systems. These processors are not available for the press at this moment, but they will be by the end of August. We prefer not to spread information on operating clock speeds and performance as of right now, but I will say that the early versions of Barcelona processors will reach 2 GHz.
Hardware Upgrade: A very interesting topic in the past few months has been the employment of AMD’s latest generation of GPUs for parallel processing. NVIDIA is particularly active in this market sector thanks to its new TESLA family of products (we have already taken a brief look at GPGPU processing here), but we still fail to clearly see AMD’s approach to this new market.
Giuseppe Amato: Last fall we introduced a product family specifically designed for GPGPU processing. We are calling GPGPU processing on AMD products "stream computing.” In “stream computing" we use the R580 family of GPUs and adjust the boards for a GPGPU environment, making them different from traditional graphics cards typically used for 3D applications.
Hardware Upgrade: Can you tell us something about how AMD intends to make its moves in this sector with the new R600 architecture? Actually, so far no recent announcements have been made concerning GPGPU solutions that are specifically based on the R600 GPU.
Giuseppe Amato: We intend to keep offering GPU solutions that are specifically designed for parallel processing to be placed side by side with our processors. GPGPU products based on the R600 GPU, which uses a unified shader model, are obviously under production, but I cannot tell you more about them at the moment. The only thing that I can tell in advance is that AMD will soon announce more GPGPU news.
Hardware Upgrade: What do you think the future of the GPGPU processing field will be? If correctly used GPUs may add powerful processing power to existing platforms, but so far their use is limited by the intrinsic nature of the GPU which creates large programming hurdles.
Giuseppe Amato: At the moment we are really only at the dawn of GPGPU programming. Because of this, developers are facing extremely hard and complex work programming for the GPU. On the other hand, when correctly and efficiently used this new technology allows significant increases in performance. An example that we have shown that uses our previous video architectures is Tarari. By recompiling its antivirus scanner and using an AMD GPU, Tarari was able to reach significantly higher performance compared to what would have been obtainable using only a CPU.
In the future, GPUs will become more easily usable for general purpose processing, thanks to the increasing availability of compilers designed for this environment. After 2010 we can assume that there will be a real time compiler available that will choose what code to execute by checking which hardware resource is available, choosing from a CPU, GPU or specific accelerators. A dispatcher will check which resources are available and will compile the code in real time to be used with the processor that at that moment is able to provide the most efficiency.
ATI GPU programming for general purpose computing relies on Close To Metal, which is AMD’s first step in the direction of GPGPU programming. Eventually we will witness evolutions in our approach to GPGPU which will allow software developers to more easily use the full capacity of the GPU. Thanks to the availablilty of a higher number of registries General purpose GPU computing will also be made much easier in 2009 when Microsoft releases the DirectX 11 API.
|
|
|
|
|
|
|
|
|
Add to del.icio.us