The Intel Larrabee project has lit up countless people's infinite expectations for the third force in the field of discrete graphics. However, it turns out that the graphics architecture based on the x86 multi-core architecture is not so simple. It is difficult for Intel, such as Intel, to bring it out of drawings. San Francisco autumn IDF 2010, Intel is even more frankly admitted that Larrabee discrete graphics is basically impossible.
In a panel discussion, someone questioned Intel’s academician and architectural director of architecture, Thomas Piazza, wondering if he could see a Larrabee-based graphics processor in the future. Throughout the group’s stare, Thomas Piazza sat in his chair and said with a quips: “I thought no one would ask me this question for two days.†Then gave a brief answer: “ I don't think it's a game."
Over the years, Intel has been emphasizing the powerful performance and flexible programmability of the Larrabee architecture. Even at the IDF 2009 a year ago, even a practical demonstration was performed, but the continuous performance of the bouncing, too low performance and excessive performance Consumption eventually led to the Larrabee project being put on the shelf and changed to become a software development platform. Although it was not completely terminated, Intel could eventually release hardware products. The performance would be several times behind NVIDIA and AMD.
When asked why he failed, Thomas Piazza was unexpectedly frank. He explained: "As far as software complexity is concerned, I think that all functions are unrealistic through software implementation, and we are also trying to do so when it comes to performance and power consumption. It is natural that rasterization requires fixed functionality. There is no reason (to allow everything) to be programmable. The space for doing so is too narrow."
According to him, the fundamental reason for Larrabee's failure was that Intel tried to make everything programmable, but failed to find a "reasonable balance between programmability and fixed functionality."
In a panel discussion, someone questioned Intel’s academician and architectural director of architecture, Thomas Piazza, wondering if he could see a Larrabee-based graphics processor in the future. Throughout the group’s stare, Thomas Piazza sat in his chair and said with a quips: “I thought no one would ask me this question for two days.†Then gave a brief answer: “ I don't think it's a game."
Over the years, Intel has been emphasizing the powerful performance and flexible programmability of the Larrabee architecture. Even at the IDF 2009 a year ago, even a practical demonstration was performed, but the continuous performance of the bouncing, too low performance and excessive performance Consumption eventually led to the Larrabee project being put on the shelf and changed to become a software development platform. Although it was not completely terminated, Intel could eventually release hardware products. The performance would be several times behind NVIDIA and AMD.
When asked why he failed, Thomas Piazza was unexpectedly frank. He explained: "As far as software complexity is concerned, I think that all functions are unrealistic through software implementation, and we are also trying to do so when it comes to performance and power consumption. It is natural that rasterization requires fixed functionality. There is no reason (to allow everything) to be programmable. The space for doing so is too narrow."
According to him, the fundamental reason for Larrabee's failure was that Intel tried to make everything programmable, but failed to find a "reasonable balance between programmability and fixed functionality."
Zhejiang Synmot Electrical Technology Co., Ltd , https://www.synmot-electrical.com