Significance: Just as the device processor race really seemed to be narrowing down to two viable contenders, ARM and Intel, mobile GPU core specialist Imagination Technologies has shifted the goalposts once again. It has acquired MIPS, the fading star of RISC processor architectures, with a view to adding a strong CPU platform to its graphics base. That could enable it to strengthen MIPS’ failing hand in the mobile space, leveraging the strength of Imagination’s presence via its PowerVR graphics base. As well as trying to keep MIPS clients like Sony PlayStation loyal with integrated graphics, MIPS also aims to provide an alternative to ARM ‘ with Apple an off-beam but interesting potential target. It could use MIPS’ architecture to gain a place in sectors where the CPU has not yet been sewn up, such as the internet of things, and where a GPU-only play will not be relevant. And it can take advantage of MIPS’ progress among Chinese chipmakers and in the set-top and broadband segment. It faces a major transition in its technology and its business model, as it may have to work against ARM and Intel rather than alongside them ‘ but it needs a new source of growth, especially as ARM is clocking up wins for its own GPU offering, Mali. And those worried at the prospect of a processor duopoly will welcome a renewed chance to keep the venerable MIPS platform alive.
Things are getting even more interesting in the mobile processor market. Just as ARM stepped up its challenge to Intel with its move to 64-bit architecture and an alliance with AMD, so the UK-based processor IP provider faces a rejuvenated rival of its own, with Imagination Technologies’ purchase of another processor platform, MIPS.
Imagination provides graphics processor (GPU) cores to many of ARM’s customers, including Apple, and also to Intel, and is the chief target for ARM’s own GPU architecture, Mali. However, while Imagination ‘ which also boasts investments from Apple and Intel, which own about 25% of the firm between them ‘ may trump its fellow UK processor IP firm in graphics, it only has fledgling activities in the central processor, where ARM dominates. Now it is indicating, with this purchase, that it may get serious about offering a combined GPU/CPU platform. Eating into ARM’s massive mobile share would be tough, but an obvious first target could be Apple, which increasingly wants complete control of its hardware design ‘ to the extent that, in another possible twist in the device processor saga, it is reported to be considering ways to migrate future Mac products from Intel to its inhouse chips.
Before considering the future options for Apple, what are the immediate terms of the deal? In fact, MIPS ‘ which has been up for sale for about a year ‘ has been split into two. ARM itself leads a group of firms calling themselves AST (Allied Security Trust), whose acquisition vehicle Bridge Crossing will pay $350m for the US company’s valuable patents (see inset). They gain access to the IPR but also prevent it falling into the hands of a firm which might have used it to launch widespread lawsuits. MIPS itself had never followed that route, despite the recent vogue for extracting more revenues from patents, and its possession of highly valued intellectual property, as a pioneer in 32-bit and 64-bit microprocessors. According to analyst Scott Gardner of the Linley Group, ‘the MIPS patents could have been used to extract licence fees from most of the companies building CPUs’.
The operational business of MIPS goes to Imagination, for a far smaller sum – $60m ‘ reflecting the pressures the company has come under in recent years as its architecture was increasingly squeezed by ARM and Intel x86. Although about 700m chips based on MIPS technology were shipped in its last fiscal year (to June), and that generated about $60m in revenue, it made a $9m loss. Although it retains a strong place in the set-top box and home gateway market, it has failed to make a major impact with its mobile designs, nor to move up convincingly to servers and network equipment as ARM is doing (though Cavium uses its cores for network chips). It also faces the risk that key customers such as Broadcom and Sony will defect ‘ Broadcom uses MIPS for its home broadband ranges, but is seeing rising growth from mobile sectors, where it has an ARM strategy and could seek to adopt a uniform platform across all its chips; while Sony uses the platform for the PlayStation Portable, but is now harmonizing its various gadgets after taking full control of its handset venture (which uses ARM-based chips).
Imagination gains some of the IPR portfolio as part of its purchase ‘ 82 patents, plus licensing rights to the 498 which go to AST. It also acquires a team of 160 engineers, to boost its own fledgling efforts to create a central processor platform to complement its flagship graphics processor family. This is the main area of interest. A complete demise of the MIPS architecture would have forced some customers to rethink but would hardly have been a shock outcome. But Imagination is committed to keeping the platform alive as part of its own bid to broaden its reach into the heart of the device. The chipmakers themselves are increasingly spanning both GPU and CPU, and integrating the two in mix-and-match multicore designs (as seen at AMD, Qualcomm and Nvidia. The latter moved from its graphics speciality into CPUs via its mobile range, Tegra, and now plans to add processor capabilities to its desktop/server GPU platform too). So the processor IP suppliers will make similar moves ‘ ARM expanding into graphics with Mali, and Imagination going the opposite way, initially with its inhouse initiative, which will now be boosted considerably by MIPS.
“MIPS is the company that pioneered the RISC CPU architecture,” said Imagination CEO Hossein Yassaie in a statement. “I believe that the combination of our existing CPU technologies…with MIPS’ capabilities will help us to create a new force to be reckoned with in the CPU IP market.’
That is a daunting process, in a world where the received wisdom is that device and even server processors are now a two-horse race between ARM and Intel, and Imagination would be coming from a long way behind. However, as many devices become increasingly centered in graphics and video performance, it will hope to leverage its GPU presence ‘ and the politics of the mobile industry ‘ to establish an alternative power base. And Yassaie believes this is a battle Imagination must enter, saying CPUs are ‘the fundamental business we need to get engaged in’, especially in the high growth area of connected devices. These provide two reasons to push its own CPU ‘ high end products will need integrated graphics and processing, while lower specified gadgets will not justify dedicated GPUs and so will be off-limits to a graphics-only player.
Yassaie confirmed that his company would aim to merge the MIPS architecture with its own efforts, saying ‘there are significant similarities between our inhouse CPU and that of MIPS’. The acquisition should accelerate Imagination’s progress, since its own Meta core would have required massive investment and time to become a serious ARM challenger.
The CEO also outlined three other advantages of the MIPS purchase. One, the architecture is the only one, apart from ARM and Intel, to support Android directly, so is ready for the mobile market in a way that Meta was not. Two, although MIPS had made limited impact on the smartphone sector, it had increased its existing position in China by broadening its reach, and Imagination hopes to build on this. Chinese licensees include fabless semiconductor provider Ingenic. Three, the MIPS platform pushes into higher end sectors than Meta, notably with the Aptiv family of cores. This has three ranges, with the ProAptiv targeting ARM’s recently announced high end, Cortex-A57, and claiming higher performance ratings (4.5 CoreMark/MHz compared to 3.9 for the A57). And unlike Intel, analysts say MIPS has fully convinced the market that it can deliver performance chips that consume as little power as ARM’s, or even less.
Among Imagination’s major customers, the prospects for ousting existing CPU platforms initially look bleak. Intel is impossible of course, but largely stuck with Imagination, even if irritated by its processor moves, because it could not politically adopt ARM Mali. Samsung uses both Mali and Imagination’s PowerVR in different models, but its own Exynos processors are firmly ARM-based. However, Imagination’s hopes rest on future demand for a more unified GPU/CPU approach, as well as new chip entrants, especially in the fledgling internet of things sector.
It also argues that the CPU space will follow the same path as GPUs, where a rising number of firms take a multivendor approach to target different types of devices or to hedge their bets. While the move to a unified platform across all chips was a risk factor for MIPS, which would inevitably be the one squeezed out, Imagination thinks some vendors will actually go the opposite way. Samsung, Qualcomm, ST-Ericsson and others all use a mixture of cores in graphics, for instance. Late last year, Qualcomm licensed the display IP element of PowerVR, even though it has its own graphics chip platform, Adreno, which it acquired from AMD, and many expect it to rely more heavily on Imagination while it builds up the Adreno line.
And then there is Apple. The iPhone maker has become increasingly autonomous in its mobile processors, relying initially on a Samsung chip, then moving to one co-designed with the Korean firm, and now using the fully homegrown A6, which is just manufactured by Samsung. Its strategy rests on iron control of every aspect of its products, including software, hardware and components, and that would make it more likely than many ARM users to consider a different architecture in future, if it thought that would deliver competitive edge and increased differentiation. Although Apple has an architectural licence to the ARM platform, which it inherited as a co-founder of the firm, it also uses PowerVR for graphics and is an investor in Imagination.
Apple may be looking to expand its homegrown architecture ‘ whether or not that introduces new cores in future ‘ to span the iDevices and the Mac family in the same way that its two operating systems and user interfaces are moving more closely together. Bloomberg this week revived recurrent reports that Apple will oust Intel in a future Mac/MacBook generation and base them on its inhouse designs.
As well as increasing control, Apple would reduce costs by using a common inhouse platform instead of Intel’s expensive chips. It might also be able to leapfrog the Intel architecture, especially as ultra-mobility and low power become almost as important in notebooks and netbooks as in smartphones (and are not Intel’s greatest strength). Much will depend on how effectively Intel levers its higher end Core platform into the tablet and mobile notebook form factors, with the advent of next generation processors like Haswell ‘ which promise to slash power consumption in the same way as the lower performance Atom.
Apple will need 64-bit architecture to replace x86 with ARM in the Macbook or Mac lines, which means waiting for another year at least ‘ and even longer, given that Apple now uses its own ARM-based designs, rather than standard ones, to obtain performance advantages and differentiation. But at that stage, it could be well positioned to make a transition (as it did back in 2006 when it moved the Mac from PowerPC to Intel, even though it had been a co-designer of the Power platform with IBM and Motorola. That partnership, and Apple’s founding membership of ARM, remind us that its interest in controlling chip platforms is nothing new.)
Apple has invested considerable effort into using its architecture licence to adapt the current ARMv7 instruction set to boost performance and according to analysis by AnandTech, its A6 processor could already rival a traditional desktop CPU. And the Mac operating system, OS X, now runs on the same kernel software as iOS ‘ as long ago as 2007, Apple said OS X could run on an iPhone. That suggests that much of the work of optimizing the operating system for an ARM-based architecture has already been done, and even the applications base of the two systems is increasingly converging.
The kind of resources that go into such efforts to optimize systems for ARM indicate why Imagination will have a tough fight actually to displace ARM in most of its existing customers, though we still think an Apple deal ‘ or PA Semi-style acquisition ‘ is still not impossible as the firm focuses even more deeply on silicon differentiation (one of the new groups set up under CEO Tim Cook’s recent reorganization was hardware, with a special concentration on chips). However, Imagination’s main targets must be to accelerate the faltering moves MIPS had made into the mobile market, in particular to establish a stronghold among the host of emerging Chinese and Taiwanese processor makers; to keep existing MIPS customers loyal by offering improved graphics capabilities; and to gain a place in areas where the CPU race is not yet decided, notably the internet of things.
The patents deal:
ARM is leading the Bridge Crossing consortium, which will pay $350m in cash for rights to MIPS’ portfolio of almost 500 processor patents. ARM itself will contribute almost half the sum ($167.5m). Bridge Crossing is an acquisition vehicle for Allied Security Trust (AST), a consortium of companies set up to buy, sell and license patents, both commercially and to keep valuable IPR out of the hands of single players which might use them aggressively. Members include Avaya, Hewlett-Packard, IBM, Intel, Motorola, Oracle, Philips, RIM and others. It is not yet clear whether ARM customers like Qualcomm, Broadcom and Apple will have to purchase their own licences for access to the AST portfolio, or whether they may be part of AST themselves.
Once the Imagination-MIPS deal is completed, MIPS royalties will go to Imagination, not to AST, said the UK firm’s CEO Hossein Yassaie. The 498 MIPS patents bought by Bridge Crossing provide the group with general patent protection rather than access to specific aspects of the MIPS architecture, he added, but the consortium will make available licences firms outside the group.
“ARM is a leading participant in this consortium, which presents an opportunity for companies to neutralize any potential infringement risk from these patents in the further development of advanced embedded technology,” ARM CEO Warren East said in a statement. “Litigation is expensive and time consuming and, in this case, a collective approach with other major industry players was the best way to remove that risk.”