Has anyone told Qualcomm, Intel and the gang?
Apple has its own processors, and given its current legal battle with Qualcomm, is likely to accelerate the process of switching to Intel for some modems. Samsung splits its favours between its own SoCs and Qualcomm’s Snapdragon, but the more its silicon drives its profits, the more its handset arm is likely to feel the corporate pressure to stay in-house. Huawei, which now claims to be the second largest smartphone vendor, is increasing its reliance on its HiSilicon subsidiary.
And now Google is planning to build a team to create its own SoC for smartphones and tablets.
Google is not a Samsung, Huawei or Apple in terms of device volumes, of course. But it is very influential over the whole mobile device market. Apparently its desire to go it alone in chip design is geared to driving advanced experiences, including virtual reality and augmented reality technologies, and that could impact on future devices in which Qualcomm might hope to take the lead.
After a string of Google smartphones which were primarily developer showcases – particularly the Nexus branded devices – it unveiled a fully commercial, and heavily marketed, product, the Pixel, in October. This was based on the Qualcomm Snapdragon SoC.
However, it was already showing greater ambitions in the device market, especially with the setting up of its dedicated division for this purpose, and the hiring of David Foster, formerly head of Amazon’s Lab126 group, which created its Kindle and Fire ranges and other products. Since Foster joined, Google has moved into the home hardware space with Google Home, and is likely to want to develop more advanced objects – anything which will drive usage of the internet, and particularly of Google services, and help the search giant to retain its control of the direction of the web user experience.
So Google would want chips optimized for the technologies which it believes will be most important in shaping our interactions with the web, and with other humans and things. The company recently posted almost 200 job adverts within its hardware group, including six specifically related to mobile SoC architecture and design. It is looking for engineers with at least a decade of experience in this field, but it additionally wants to recruit mobile SoC physical, packaging and memory design engineers.
While many other hardware positions are based in many locations round the globe, all the mobile SoC recruits would be based at Google headquarters in Mountain View, California.
Google already has a lead SoC architect in place, having hired Manu Gulati from Apple last month. Gulati worked on Apple custom chips for the iPad, iPhone and Apple TV, and was named on about 15 patents during his time with the company.
May go in-house, or work with a chip partner
Reports from The Information suggest that, while Google may be willing to bear the $500m cost of developing an SoC from scratch, if it helps drive a new generation of devices and user experiences, it will not put all its eggs in one basket. It has also sent a lengthy list of requests to chip vendors for future SoC designs, in parallel with its in-house efforts.
This suggests that Google will pursue a similar strategy to the one it has adopted for its cloud servers, commissioning custom designs from its chip suppliers. Qualcomm, Intel and others are said to be demonstrating heavily customized server processors in the hope of winning a Google deal.
Currently, a similar arrangement in the device market would be less attractive to a chipmaker because of Google’s relatively low volumes – about one million Pixels are thought to have been sold – but its business would be welcome for the potential growth in future, and for the influence over the development of future form factors.
The report says that, during discussions late last year, “Google representatives put forward designs of chips it was interested in co-developing, including a phone’s main processor”.
In particular, Google is said to be interested in chips which will optimize the performance of virtual and augmented reality use cases, as it builds up its VR team, transferring some key executives from elsewhere, including Jon Wiley, the lead designer of Google Search, and Alex Faaborg, the former lead designer for Firefox, Google Now, and Android Wear.
Chips to drive the VR/AR experience
There have been persistent rumors that Google is looking to build on its existing VR systems to create a fully fledged VT operating system around Android. Together with the kind of devices which were emerging from its Project Tango, and with the planned chips, the company could build an entire platform to enable VR/AR experiences if those move to the heart of the web, as Google and Facebook seem to believe – as do others, notably Softbank’s Masayoshi Son, which recently took a stake in Nvidia for its VR and AI processors.
There is a hole in Project Tango since Intel acquired Movidius, which supplied the machine vision chips. Google and Intel are close partners on many projects, but Intel wants to create markets like VR and autonomous cars in its own image, and so conflicts of interest may arise if Google has to remain dependent on a third party for key technology.
As well as computer vision, Google is also said to be interested – again like Intel – in advanced image processors, which could “scan the environment and push images to Google’s cloud-based systems for analysis” and then enable Google to “give you context back” about the image. This would move on significantly from the current Google Goggles, which can identify images, and be able to work with video and real time.
If Google does develop its own VR-oriented mobile SoC, it will presumably work with ARM CPU cores, though it would have a choice of GPUs – perhaps proving a white knight for Imagination Technology, suffering under the impending loss of Apple’s business; assuming Google did not adopt the same reckless route as Apple and try to create its own GPU core.
ARM may be valuable VR-oriented ally
In November, ARM, now part of Softbank, unveiled two new products geared heavily to VR/AR and advanced graphics, which might catch Google’s eye. These were a multi-codec video processor focused on the new standard codec, HEVC, and a GPU to drive mobile VR/AR use cases. ARM is clearly confident that the popularity and prevalence of VR and AR applications in the mobile market will continue to soar. The Mali-G51 GPU is essentially a downgrade from the Mali-G71 GPU – but a downgrade targeted at a significantly larger market.
The Pokemon Go smartphone AR app had such a widespread uptake that it effectively altered the way societies functioned across the world this year, but the primary drawback of the game was its intensive drainage on battery life. ARM’s mainstream mobile GPU aims to change that; claiming 60 per cent more energy efficiency and 60 per cent more area efficiency compared to its T830 – noting a reduced SoC cost for manufacturers.
The G51 is based on the Bifrost architecture, which ARM claims passes the performance capabilities of today’s midrange laptops. The Bifrost architecture has been optimized to run the Vulkan API, developed by the Khronos Group and based on AMD’s Mantle API – which AMD has donated in the hope of pushing Vulkan as the next cross-industry standard to replace OpenGL.
So while the likes of HiSilicon, MediaTek and Samsung Electronics have licensed the G71, pitched as ideally suited to immersive VR and AR in premium devices, the G51 aims to provide better mobile AR and VR experiences to the masses, potentially for the next breed of Pokemon Go, but not necessarily limited to gaming.
Copyright © 2017, Wireless Watch
Wireless Watch is published by Rethink Research, a London-based IT publishing and consulting firm. This weekly newsletter delivers in-depth analysis and market research of mobile and wireless for business. Subscription details are here.