This past week, the successive drama between Intel and Ultra Micro (AMD) is a inundated, ostensibly just product release and personnel, but it is implied in the current era of AI (AI) priority, even in the past to meet Congon competitors, You must also smile and hug and work together, and Intel should be able to deeply understand the philosophy behind the alliance's minor enemies and fight against the main enemy, after all, this is the real environment that Intel must face today. Intel announced in Monday that it was allied with the Super Micro, which would provide a standalone graphics processor (GPU) to integrate into Intel's future SOC system single chip; Tuesday, Ultra Micro announced the absence of the GPU architecture long Raja Koduji resignation, will leave the company; In Wednesday, Intel immediately announced the extension of Raja Koduji as the Intel GPU architecture leader, and will also dominate Intel's newly established core and visual operations Group (cores and Visual Computing Group), with the post of senior vice president, who will take office since early December. Intel is more expected to import AI and visual operations from the server domain into the Edge computing solution of the actual appliance. Intel announces the extension of Raja Koduji as Intel's GPU architecture, leading Intel's newly established core and visual operations Group (cores and Visual Computing Group), with the post of senior vice president. Any Intel observer, if only to see the development of the Monday or up to Tuesday, I am afraid it will not be able to deduce Intel's Abacus, but by Wednesday, Intel officially announced that the Koduji had previously not stretched in the ultra micro-log to serve as Intel's own GPU architecture long, or just wake up, Finally, it was clear that Intel was looking for him, and after two failed attempts, Intel tried to break into the independent GPU market, especially in the higher-order market. Intel has previously operated a graphics processor, but it is limited to the integrated graphics processor market, and is limited to the lower-order market, as for the independent GPU market, the graphics processor, dual NVIDIA and Ultra micro-oligopoly, is now a high percentage of Nvidia, and Nvidia in recent years in addition to the mapping market , and the AI accelerator does not allow Intel to specialize in the data center market, as a result, Intel's Koduji has come to serve as senior vice president of the newly established core and visual computing business, as well as the task of expanding Intel's GPU operations, in particular Intel's exploration of an independent GPU market. Raja Koduji, aged 49, his seniority in the GPU, a pioneer in GPU architecture, has served two times in the past 25 years of his career, and has been extended by Apple to serve as the director of the graphics architecture of the Apple GPU team, Over the past 4 years, he has focused on the micro-building Vega architecture GPU. At the same time, Intel has unveiled a new GPU strategy from top to bottom while announcing the extension of Koduji. At the bottom, Intel wants to extend the existing integrated GPU market into a brand new edge device, but does not disclose more details. However, the word "edge" from which it is used is sufficient to know that the device that has strongly hinted at the level of the Internet is present, while the word marginal is more about the inference of neural networks. To be sure, Intel has cut from atom processors to a certain extent in this area, and also through the Movidius neural compute engine, but now Intel will extend its integrated GPU into the market. As far as the top view is concerned, it may be that the players, or the technology industry, should be impressed by Intel's lock-in of higher-order independent GPU markets, once again world. Intel has long tried to enter the standalone GPU market with 2 of failed experiences: a i740 attempt in the late 1990, and the second in the late 2000 to announce the Larrabee plan for the break. However, although these plans failed to meet Intel's intended development goals, Intel continues to develop its own GPU architecture and GPU-like devices, which are more of a large parallel computing focused Xeon Phi family, becoming a part of Intel's AI layout today. In contrast to Intel's announced staffing and strategy, perhaps the most variable is the product release time, and in Intel's announcement, the release time of the higher-order independent GPU product is not visible. If Intel intends to develop new products from scratch after a recent Koduji, about 4 years of development time, it is expected that Intel's first higher-order independent GPU products will be released in 2022; However, if Intel had previously had a part of the duck-pull action on the table, perhaps the new product could have been released earlier, given that Nvidia's rapid pace in recent years is expected to be an early move on the table. So, the question goes back to the ultra tiny body, whether the covenants concluded by the two factories on Monday are not historic? I think, is also not. After all, the Ultra Micro Solution for Intel, it's only a transitional one, and ultimately it depends on its own solution, but Intel is not willing to see a big client like Apple turn Shing, so it is bound to find a solution to the frivolous power, and, in the course of all the media interviews, This is Intel's plan, Super Micro is only to provide customers products, there is no long-term covenant what kind of existence, perhaps for the ultra micro, self-reliance and self-improvement all on their own, better than holding Intel came more strong. 10 years Hedong, 10 years hexi, how do you know which day Intel will not join Nvidia to hit itself?
