Mobile World Congress kicks off today in Barcelona, and companies like Intel are taking to the forum to discuss their product roadmaps and long-term product efforts. For Intel, 2015 is about building its lower-cost Android platforms and winning more market share in budget segments, while simultaneously ramping up production of its LTE-Advanced XMM 7360 modem.
Last week, we discussed the company’s new Atom branding system (x3/x5/x7). But today we can share more details on how the stack will look. Intel’s goal for Atom is to offer solutions that stretch from the bottom of the tablet and smartphone market into Windows hardware, all on the same platform.
The Atom x3 family is powered by the low-cost SoFIA SoCs that Intel is building at TSMC and selling to companies like Rockchip. Most of these are mass-market devices intended for markets where 2G or 3G are the primary wireless network options.
The x3 family’s hardware specs certainly reflect their modest position. The bottom-end SoFIA device is a dual-core product at just 1GHz, with a bottom-end Mali-400MP2 GPU (the Galaxy S II from Samsung used a Mali-400MP4 nearly four years ago). In fact, the Galaxy S II seems a reasonable point of comparison in multiple respects, since it also offered a 1.2GHz Cortex-A9 — likely slower than the Bay Trail CPU used in the Atom X3-C3130.
The middleweight Atom x3, the x3-C3230RK, bumps up to four cores at 1.2GHz and a Mali-450MP4. This chip will be the better budget option, if companies support it — both the CPU and GPU should be significantly faster, while the cost should still be low. Finally, there’s the x3-C3440, which increases CPU performance and adds LTE, but cuts the GPU block down again. It’s not clear if a Mali T720 MP2 outperforms a 4-way M450.
Intel is claiming that the new x3 cores will be faster than their competition, which tends to be based on the Cortex-A7, but the implications of the results above are that this test isn’t very well multi-threaded. The quad-core Atom clocks up to 20 percent faster than the dual-core flavor and doubles the core count, yet is just 20 percent faster, while the quad-core Qualcomm MSM8226 Cortex-A7 is just 37 percent faster than the Spreadtrum SC7715, despite having 4x the CPU cores.
The other major feature of the 28nm chips is that they’ll offer Intel’s first integrated modem. That capability is the whole reason Intel chose to build chips at TSMC — the company swapped using its own process technology on 14nm with an integrated modem for the ability to bring something to market more quickly on 28nm.
Atom x5, x7, and the XMM 7360 modem
There’s not much surprising in the Atom x5 and x7 products — at least, not yet. These chips will be built on 14nm and include Broadwell-class graphics. While companies like Samsung might be leaping for LPDDR4, Intel is sticking with DDR3 for now, as shown below:
These chips will support Intel’s LTE-capable XMM 7260 modem, but will connect to it via a discrete bus rather than on-die integration. Later this year, we’ll see the XMM 7360 modem shipping commercially — this is Intel’s first LTE-A modem with support for up to 450Mbps connections.
The lead time between Intel’s commercial ship dates for modems and their actual usage in the wild is typically quite long. If the company doesn’t ship the device until the back half of the year, it may not show up in any products until 2016. Presumably this modem is still built at TSMC on 28nm — Intel’s first 14nm modems will be the ones it integrates into its next-generation smartphone and tablet platforms in 2016 to 2017.
Intel’s long-term mobile strategy
This year is important to Intel for two critical reasons. First, SoFIA and Cherry Trail are expected to substantially reduce the company’s need for contra-revenue. Intel doesn’t expect to completely stop using contra-revenue to buoy its mobile market share, but it does expect to reduce the loss to less than a billion dollars, as compared with well over four billion in 2014.
It’s SoFIA partnership with TSMC, meanwhile, is a key plank of the company’s overall mobile strategy. Intel now needs to demonstrate a market for these devices — ideally a robust market. If its partnership with companies like Rockchip pay off, we can expect to see more of these arrangements in the future. If they don’t — if the mobile division continues to struggle with market share issues and limited OEM interest — then Intel will have to wait until its own homegrown silicon efforts and integrated modem are finally ready for primetime in the 2016 to 2017 timeframe.
It’s hard to predict which way the company’s fortunes will turn. I’ve been convinced since Medfield that Intel was fully capable of building a good mobile phone — maybe not a jaw-dropping-amazing phone, but certainly a competitive device. The problem isn’t with the capabilities of Intel’s hardware, but with the cost structures and software supportsurrounding it.
No comments:
Post a Comment