Could Intel's Atom end up the last chip standing?
When we've discussed Atom and Intel's long-term plans regarding MIDs and handheld devices, we've typically focused on the consumer experience. Intel has made no secret that it envisions a future in which next-generation Atom processors and SoCs power a diverse range of products in both consumer and business markets. What we haven't discussed, save for an occasional nod towards the competition Intel might face from ARM, is the impact Intel's push into the embedded market might have on the integrated device manufacturers (IDMs). Tech-On recently sat down with James Hogan, who helped found Tela Innovations and works in electronic design automation (EDA). The interview gives a different perspective of how Intel's plans for the embedded market could impact the existing structure.
Hogan's predictions paint a bleak picture for the future of IDMs and the companies that work with them. Asked why he felt Atom would could pose a threat to ARM, he responded: "With Intel processors, Microsoft Corp's business applications software run smoothly. In terms of processing power, it is impossible with ARM's processors, which feature small sizes and low power consumptions."
ARM doesn't seem willing to cede that point without a fight. There are a dizzying number of ARM processors available for almost any type of device in existence, but the company's Cortex A9 processors are aimed at devices that require significantly more horsepower. The A9 is available in single-core or multi-core configurations, and sports a superscalar architectures with an integrated media processor block ARM refers to as NEON. At present, the A9 is capable of executing up to two instructions per clockcycle, per core.
The current Cortex A-9 is capable of approximately 2000 DMIPS.
While the Cortex A9 isn't used in any shipping netbook I'm aware of, ARM has clearly targeted the handheld space with its latest family. NVIDIA's Tegra will power the upcoming Zune HD, which could give us a look at how effectively NVIDIA can leverage its own IP to accelerate video and audio processing on the older ARM11 MpCore. Hogan, however, believes the problem is more systemic. Asked why he thinks the number of SoC manufacturers and logic LSI chip makers decrease, Hogan responded:
"SoCs are developed basically by combining IP cores. Therefore, IDMs have to have powerful IP cores, design techniques to integrate IP cores and advanced processes. But it is not easy to have all of them. Many of the processes for SoCs are developed by silicon foundries. General-purpose IP cores are supplied by IP providers. And IP cores that differentiate products are developed by system/device manufacturers. It is difficult to run a business as an IDM."
The problem Hogan is referring to is the fact that IDM's combine IP cores designed by different companies and rely on advanced manufacturing technologies to create any SoC they may design. If a step in that process is flawed, the quality and capability of the final product is jeopardized.
While it may be true that Intel and its long-term plans for the embedded market threaten the existing ARM-based infrastructure, the battle for this space has scarcely been joined. A recent investigation by Ars Technica found that despite the plethora of MID devices at every Intel trade show, the only Atom-based tablet MID currently shipping in the US is the Toughbook U1, with a price between $2500-$3000. (Netbooks that can transform into tablets don't count). None of this is particularly surprising; Intel's plans for the MID market have always been long-term and multi-generational, but it underlines the fact that the MID market is nascent to the point of being embryonic.
Hogan's predictions paint a bleak picture for the future of IDMs and the companies that work with them. Asked why he felt Atom would could pose a threat to ARM, he responded: "With Intel processors, Microsoft Corp's business applications software run smoothly. In terms of processing power, it is impossible with ARM's processors, which feature small sizes and low power consumptions."
ARM doesn't seem willing to cede that point without a fight. There are a dizzying number of ARM processors available for almost any type of device in existence, but the company's Cortex A9 processors are aimed at devices that require significantly more horsepower. The A9 is available in single-core or multi-core configurations, and sports a superscalar architectures with an integrated media processor block ARM refers to as NEON. At present, the A9 is capable of executing up to two instructions per clockcycle, per core.
The current Cortex A-9 is capable of approximately 2000 DMIPS.
"SoCs are developed basically by combining IP cores. Therefore, IDMs have to have powerful IP cores, design techniques to integrate IP cores and advanced processes. But it is not easy to have all of them. Many of the processes for SoCs are developed by silicon foundries. General-purpose IP cores are supplied by IP providers. And IP cores that differentiate products are developed by system/device manufacturers. It is difficult to run a business as an IDM."
The problem Hogan is referring to is the fact that IDM's combine IP cores designed by different companies and rely on advanced manufacturing technologies to create any SoC they may design. If a step in that process is flawed, the quality and capability of the final product is jeopardized.
While it may be true that Intel and its long-term plans for the embedded market threaten the existing ARM-based infrastructure, the battle for this space has scarcely been joined. A recent investigation by Ars Technica found that despite the plethora of MID devices at every Intel trade show, the only Atom-based tablet MID currently shipping in the US is the Toughbook U1, with a price between $2500-$3000. (Netbooks that can transform into tablets don't count). None of this is particularly surprising; Intel's plans for the MID market have always been long-term and multi-generational, but it underlines the fact that the MID market is nascent to the point of being embryonic.
