News

  • Home > News

        Intel aims x86 at GPU market

        Jonathan Angel | Date: Aug 4, 2008 | Comments: 1



        Intel says it will "target the personal computer graphics market" as early as next year with a new chip employing multiple Pentium-based cores. "Larrabee" offers claimed support for DirectX and OpenGL, and may compete with the discrete graphics processing units (GPUs) offered by AMD/ATI and Nvidia.




        Details of the architecture, code-named Larrabee, are detailed more fully at a paper the company plans to present next week at the Association for Computing Machinery's SIGGRAPH conference in Los Angeles. But -- while it kept a lid on exactly how many x86 cores Larrabee-based chips will ultimately sport -- Intel reportedly provided some highlights for journalists at a briefing last week.

        "It looks like a GPU and acts like a GPU, but actually what it's doing is introducing a large number of x86 cores into your PC," Intel spokesperson Nick Knupffer said, according to Cnet.com. Intel also presented a graph showing how Larrabee's performance ramped up linearly as the number of cores was increased from eight to 48.

        According to a fact sheet on Intel's web site, Larrabee will offer the following features:
        • A pipeline derived from the dual-issue Pentium processor, but with significant modern enhancements such as a wide vector processing unit (VPU), multi-threading, 64-bit extensions, and sophisticated pre-fetching
        • A select few fixed function logic blocks to support graphics and other applications
        • A coherent on-die second-level cache, allowing efficient inter-processor communication and high-bandwidth local data to be accessed by CPU cores, making the writing of software programs simpler
        • A native programming model supporting a variety of highly parallel applications, including those that use irregular data structures
        • Task scheduling that is performed entirely with software, rather than in fixed function logic, so rendering pipelines and other complex software systems can readily adjust their resource scheduling
        • Four execution threads per core with separate register sets per thread, said to allow the use of a simple efficient in-order pipeline, while retaining many of the latency-hiding benefits of more complex out-of-order pipelines when running highly parallel applications
        • A 1024 bits-wide, bi-directional ring network, for "super fast" communication between cores
        • Full support for IEEE standards for single and double precision floating-point arithmetic
        • Support for the industry-standard DirectX and OpenGL APIs (application programming interfaces)
        Further information

        For more details on Larrabee, see coverage on our sister site eWEEK.com, here. The paper, "Larrabee: A Many-core x86 Architecture for Visual Computing," is apparently now available to ACM (Association for Computing Machinery) members at the organization's website, here.



        Related stories: