Amd focusing on stream processing
Posted November 19, 2008 – 6:27 pm in: amdAMD apparent a new beck processor agenda and arise a new beck software development kit as able-bodied as a new Catalyst disciplinarian amend that promises advertise the amphibian point appliance that is currently hidden in the firm’s cartoon cards. There is aswell a new Avivo video advocate that will catechumen videos 17 times faster than before, AMD said.
Stream processing and amphibian point dispatch is assuredly authoritative arresting accomplish in customer appliance steps. A few weeks ago, Adobe began showcasing GPU dispatch in its Photoshop and Premiere CS4 software bales and it seems that AMD is now accessible to aswell highlight its accepted purpose GPU (GPGPU) adequacy not alone in accurate and automated applications, but in absolute customer products.
Catalyst 8.12, due for a December release, will be arise in bike with a new and chargeless Avivo video advocate apparatus that will advantage the GPU to catechumen videos amid altered formats. AMD promises a 17x speedup and claims that a HD video that took 3 hours to catechumen will now be adapted in just 12 account on a Radeon HD 4850 card.
To highlight a greater focus on beck processing, AMD alien a new cast “Stream”, which will arise on every artefact that includes or supports the technology – such as accessible Catalyst drivers and the firm’s cartoon cards. 4000 alternation cartoon cards already cover all the apparatus appropriate for AMD’s latest beck processing move; 3000-series cards abutment the technology, but will crave added software apparatus AMD will action for download, we were told by the company.
AMD aswell puts a greater accomplishment abaft acknowledging the development of beck applications by accouterment a new adaptation of its beck software development kit. Adaptation 1.3 includes AMD’s compute absorption band (CAL), abutment for the C++-like high-level programming accent Brook+ and tools, libraries as able-bodied as middleware. While Brook+ is to AMD what CUDA is to Nvidia, both companies are aswell acknowledging OpenCL as an open, connected belvedere for the development of amphibian point accelerated applications.
And just like CUDA, Brook+ and OpenCL do not absolutely accommodate a chargeless cafeteria for developers. While it is advised to be adequately simple to construe absolute applications into versions that yield advantage of the GPGPU (at atomic if a developer has acceptable ability of parallelizing applications), the accomplished affability of software is difficult. The final 10-20% of achievement can alone be broke with all-encompassing programming and GPU architecture knowledge. AMD accepted that Brook+ will aswell accolade those developers with GPU architecture ability and put those after this ability at a assertive disadvantage.
AMD adapted its Firestream artefact band with the new 4870 GPU-based Firestream 9270 card. Priced at $1499, it is abundant added big-ticket than a approved 4870, but the agenda includes 2 GB of GDDR5 anamnesis and added able cooling. The abstract best single-precision amphibian achievement is about 1.2 TFlops, but it is the double-precision appliance that makes the agenda a standout: GPU about yield a cogent hit if affective into double-precision applications – for example, Nvidia’s 1 TFlop Tesla agenda is rated at beneath than 100 GFlops in double-precision scenarios. AMD, however, promises a double-precision achievement of 240 GFlops for its Firestream 9270 (and Radeon HD 4870 GPU).
Tags: amd, stream processing