On January 2, 2020,Imagination Technologies
Under the new agreement, Apple can use Imagination's broader intellectual property (IP) and pay licensing fees.
In April 2017, Apple announced that it would give up using all technologies of imaging within two years, including patents, intellectual property, confidential information, etc.
This means that the GPU or technology of imaging will not be used in Apple's processors after 2019, and the GPU developed by itself will not be used. Affected by this, the stock price of imaging plummeted nearly 71%.
In fact, as early as 2013, apple began to develop its own GPU.At that time, apple gathered many GPU senior talents to form a team to design GPU independently, among which Raja koduri, former chief technology officer of graphic design of AMD, was also involved.
However, at that time, Apple was not successful in the development of its own GPU, as can be seen from Raja koduri's subsequent return to AMD. As a result, apple renewed its license agreement with imaging for many years in 2014, covering all the powervr graphics and video technology patents of imaging at that time and in the next few years.
However, Apple has not given up the research and development of its own GPU.
In the first half of 2016, Apple further expanded its GPU R & D division to recruit GPU R & D talent everywhere, while continuing to poach from partner Imagination.
Perhaps because Apple is its biggest customer, Imagination seems to be having a bit of a dig at Apple
Although the a11 launched by apple in September 2017 was claimed by the media that its integrated GPU was developed by Apple itself, in fact, it still seems to be customized based on the IP of image, because its kernel design is very similar to the previous image rogue.
It is worth mentioning that Apple has also integrated its own NPU kernel on the a11 processor for the first time.
Photo source: techinsights Apple A12 die shot, chiprebel Apple a11 die shot
The A12 GPU launched by apple in 2018 still supports pvrtc (powervr texture compression), which is a proprietary format, meaning that its GPU is still likely to be associated with the IP of imaging.
Combined with Apple's announcement in 2017,
So is the GPU of Apple's A13 processor self-developed in 2019? According to apple, the GPU performance of the A13 is 20% higher than that of the previous generation A12. However, according to AnandTech tests, the GPU sustained performance score of the iPhone 11 pro with A13 is 50-60% higher than that of the iPhone XS with A12, better than Apple's own market claims.
From this point of view, Apple seems to have made a breakthrough in self-developed GPU.
But...From today's new licensing agreement between image and apple, Apple's A13 GPU is still useful to image's IP, and there may be some technical patents that are difficult to avoid in the short term.
It is worth mentioning that in 2017, after Apple announced that it would give up using all technologies of image within two years, image officials responded at that time:
In addition, it was announced by Apple in 2017
After being successfully acquired by Chinese capital, imaging began to focus on the Chinese market and actively cooperate with Chinese application processors (such as torch core) and mobile phone chip manufacturers (such as MediaTek, zhanrui, etc.). In recent years, with the popularity of AI chip market, imaging is also constantly strengthening its GPU IP's ability in AI. At the same time, it has successively launched the neural network accelerator powervr series2nx / 3nx NNA.
On December 3, 2019, image held a conference in Shanghai, announcing the launch of its 10th generation powervr graphics processing architecture img a series, including img axe, img AXM and img AXT.
Imaging said that its img a series has provided significant improvement in every dimension. In the same clock and semiconductor technology, compared with the powervr equipment currently being shipped, its performance has been improved by 2.5 times, the machine learning processing speed has been improved by 8 times, the maximum computing power can reach 8 tops, and the power consumption has been reduced by 60%.
On the same day, Ziguang zhanrui announced that it had obtained the license of the latest generation of neural network accelerator (NNA) powervr series3nx Semiconductor Intellectual Property (IP) for the future system level chip (SOC) for the middle and high-end mobile device market.