Intel's previously high-priced takeover of Mobileye, which shows its determination to aggressively grab the global auto chip market, has recently been said by Intel CEO Brian Krzanich that the company, developed by Mobileye EyeQ5, is 2.4 times times more efficient than the NVIDIA product line, This appears to show Intel has the advantage of surpassing Nvidia's self-driving chip technology, although it is expected to start production in 2020 from the Intel EyeQ5 Product Planning blueprint, but Nvidia's Xavier is expected to be put into production in 2018, showing that the production process has been poor for about two years. , so it may not be fair for Intel to compare its EyeQ5 with Nvidia Xavier as a generation, perhaps Nvidia will be able to introduce performance efficiencies and optimize the next generation of self-driving chips in two years. According to Fubis (Forbes), EyeQ5 is Intel's first chip product for Level 4 and Level 5, which is expected to start production in 2020, meaning it will not be in the car until 2022. EyeQ5 was launched before Mobileye was acquired by Intel, with its original specifications designed to fine-tune the EyeQ5 after Intel's acquisition, raising the EyeQ5 performance by 1 time times to 24 megabits per second (TOPS), but increased power consumption by 1 time times to 10 watts, The power consumption is more than 4 times times more than the EYEQ3 2.5 watts and is expected to start producing EyeQ4 3 watts in 2018, and the target is at the time of the technology-leading 7 nm process EyeQ5. Intel EyeQ5 performance data is based on product simulations, and is 1 time times higher than the Mobileye original EyeQ5 performance, which means that Intel needs to double its EyeQ5 performance to meet the previously projected level 5 self-driving technology requirements, and may be in the final version of the system single chip (SoC) Two EyeQ5 chips are used. Although Intel CEO claims that EYEQ5 performance is superior to competitors, the EyeQ5 performance from the above analog data should be good, but EyeQ5 is a chip that was planned to begin production in 2020, more than 3 years ago, with Intel comparing Nvidia Xavier Chip has now been sent to manufacturing (tape out), will be put into production in 2018, which is equal to the EyeQ5 scheduled to put into production time a full two years earlier, meaning that Nvidia Xavier and so expected to produce in 2020 from the previous generation of driving chip product line. In this case, it is problematic to compare Nvidia Xavier with a 16 nm process with EyeQ5, which is expected to be produced in advanced 7 nm process, to say that Intel is too early to argue that EyeQ5 is superior to any type of energy efficiency of Nvidia's self-driving chips. In addition, since Nvidia Xavier and Intel EyeQ5 will be part of a larger self-driving platform design, it may not be meaningful to compare who's chip performance better. such as Xavier for an integrated Soc, with other Xavier Soc and independent graphics chip (GPU), combined with the design of Nvidia drive PX platform, built-in two Xavier Soc and two Volta GPU Drive PX Pegasus that is a case . In addition, Nvidia is also the current global artificial intelligence (AI) solutions leadership vendors, especially Nvidia's new Volta architecture and integration of the tensor core, which also plays an important role for Xavier Soc, under the Volta architecture, It also makes Xavier the highest performance solution ever announced for automotive applications, not to mention the NVIDIA architecture that supports all training architectures. Anti-view is still unclear whether the platform using the Intel EYEQ5 chip will be designed with multiple chips, with other accelerators, or which AI architecture will be supported. Not only NVIDIA, Intel EyeQ5 is still facing many competitors and competing products in the current global self-driving chip market, including the NXP (NXP) Bluebox platform of Qualcomm, and Renesas (Renesas) ' open self-driving platform ' (open Autonomous vehicle Platform, Siemens and mentor DRS360 platform, Samsung Electronics (Samsung Electronics) and Harman's self-driving platform, and the mainland Baidu Apollo (Apollo) platform. It is worth noting that the new self-driving chip and platform to achieve formal shipment to the automotive manufacturing side, it will take at least two years to estimate the assumption of support level 4, Level 5 Nvidia Xavier chip is actually in 2018 into production, It is expected that the first wave of 2020 will be the first car to carry Xavier chips.
