Smartwatches will be one of the bright spots in an otherwise slowing mobile market, according to analyst Linley Gwennap. Over the next five years, both smartphones and tablets will slump to single-digit growth levels but smartwatches could rise to annual sales of 300 to 400 million units, he said at the Linley Mobile Conference here on Wednesday.

Forecasts that Apple could sell as many as 40 million of its watches in their first year are optimistic, nevertheless, “we expect a big jump,” Gwennap said. “In a year or so there will be a lot of knock-off products that look like the Apple Watch but sell for more like 50 bucks,” he said, estimating by 2019 the 17% of smartphone buyers purchasing high-end products will also get smartwatches.

The devices need more integrated silicon to get down to sub-$100 prices consumers like, probably driving adoption of Cortex-M processors, he said. Gwennap speculated the Apple Watch uses an ARM Cortex-A CPU and an OpenGL GPU co-developed with a third party.

Meanwhile smartphones will slow to compound annual growth through 2019 of 9.6% when a whopping two billion units will ship annually. Tablets are hitting the brakes even harder growing at just 4.4% during the period, he predicted (see below).

“The big driver was converting feature phones users to smartphones, but that’s pretty much over now, so vendors are trying to convert basic phone users and that’s going well,” he said.

Meanwhile LTE-based smartphones are on a fast ramp, representing 28% of units and 55% of revenues in 2015. LTE will be in half of all smartphones by 2018, hitting the crossover five years faster than 3G took over from 2G.

Overall smartphone growth will still be a respectable 16% this year, down from 27% last year, Gwennap predicted. Tablets are already slumping to 7% growth this year, expected to nearly reach a peak in 2019 at about 320 million units, down from estimates a few years ago of as many as 600 million tablets a year.

“The growth is at the low end with white-box tablets as the market matures,” he said, noting Apple iPad shipments declined last year and Samsung’s sales were weak.

Windows tablets are one small bright spot in the mix, growing to about 27 million units or 8% of the market in 2019. “They are still pretty expensive, but are being used in the enterprise,” Gwennap said.

Four merchant vendors plus internal ASICs such as the Apple A-series now account for 95% of the smartphone processor market. “LTE is no longer a differentiator here with 11 LTE baseband vendors,” Gwennap said.

Qualcomm continues to lead the market but saw its 40% share reduced to about 35% with the loss of sockets to the Exynos 7 in Samsung’s Galaxy S6 handsets. Meanwhile Mediatek is distancing itself from competitors to hold a strong #2 position.

The tablet processor market has more competitors with leaders changing every year, Gwennap said. Among merchant chip vendors, Mediatek should come out on top in 2015, thanks to demand for its integrated cellular chips that are adding value to even low-end tablets, he said.

Intel “bought its way” into leadership in merchant tablet chips last year, but is expected to end subsidies this year that had it making sales at a loss, Gwennap said. “Once you take the subsidies away, I think they will lose a lot of that business,” he added.

It’s been a long, rough road for Intel trying to get into mobile systems and the horizon does not look good, Gwennap said.

Merrifield got very little traction in the market. Now they are resting hopes on an integrated Sofia, but it is 3G only and made at TSMC, so its performance is not that exciting and they will have to sell it at a very low price. An LTE version made in Intel’s own fabs is coming sometime next year, but at that point who knows where the rest of the market will be.

Overall the standalone applications processor is disappearing from smartphones. However, as much as 30% of tablets will use such chips by 2018, Gwennap predicted.

ARM’s Big.little approach to mixing high- and low-end cores has largely been a flop with OEMs. Marketing is calling for eight cores in high-end products, adopting eight small cores because eight high-end cores bust thermal limits.

“Users won’t see big performance benefits because there’s not a lot of software that uses four let alone eight cores and the lower-end cores have lower single-thread performance, but the numbers look good on benchmarks,” Gwennap said.

At the low end, single-core chips have nearly disappeared. Dual-core versions are fading as the industry moves to shipping processors with four cores on average in 2015, he estimated.

Heterogeneous SoCs using a mix of different cores are gaining traction. An emerging class of vision processing units likely will be the next big thing, targeting apps such as augmented reality. They will provide 10- to 100-fold performance/watt improvements over CPUs and GPUs on such jobs for less than $5 per device, Gwennap projected.

The mixed-core SoCs require a distributed set of cache coherency engines, opening a market for new chip-design tools, Gwennap said. To that end, Arteris announced at the event it is shipping FlexNOC Physical. It is an add on to its existing version 3.0 tool that gives designers information about physical placement of on-chip network elements, allowing them to be isolated and independently optimized.

Also at the event, startup NetSpeed formally announced Gemini, a competing tool it claims provides designers data on physical placement of cache coherent elements. Its approach provides greater flexibility, more opportunities for optimizations and shorter time to market, it said.

Not to be outdone, the other player in on-chip networks, Sonics, is expected to announce related enhancements to its tools next month.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved