ICERA

Late last year Intel announced the intention to acquire Infineon’s Wireless Solutions business, highlighting an important trend in the application processor space: those determined to lead in the smartphone/tablet SoC market have an intense desire to control as much IP as possible. It’s not enough to simply bundle licensed IP blocks, acquiring the IP is the first step towards consolidating the market.
It started years ago with the acquisition of companies that made audio and video decoders. Qualcomm even purchased ATI’s Imageon GPU business, again with the mindset that if you don’t own your IP, you’re no different than just another competitor who licenses it.

Qualcomm itself has done very well as it has a rare combination of owning a considerable amount of IP within its application processor but also having a successful cellular baseband business. Qualcomm designs its own CPU cores (based on ARM instruction sets), its own GPU cores and likely controls much of the other IP going into its SoCs. On top of all of that, Qualcomm designs its own modems which can be both integrated into its application processors or used externally as a separate chip.

Qualcomm modems have been used in many of the smartphones we’ve played with lately including the HTC Thunderbolt and the Droid Charge.

Having an extensive baseband portfolio is no different for a player in the app processor space as owning your own chipset business was for Intel in the early 2000s. If you can provide your customers (smartphone/tablet vendors) with a platform rather than individual chips there’s the potential to market synergies between your offerings and then of course you can sell more chips to the same customer rather than sending that business elsewhere. Eventually this approach drives the individual vendors of various IP out of business as power is consolidated in the hands of the strongest SoC players.

Today NVIDIA announced that it too is going to play in this game with the acquisition of Icera. The acquisition is expected to close in 30 days and is worth $367 million in cash.

Icera’s modems aren’t too common in high end smartphones. A quick look through its product offering shows a strong focus on mobile broadband devices (e.g. USB modems), however it’s possible that Icera simply could never break into the handset market - something NVIDIA can help address with its recent success on the back of Tegra 2. It’s clear that eventually integrating Icera’s technology into Tegra SoCs now becomes an option which, assuming there are no fundamental issues with Icera’s baseband designs, would obviously increase its presence in the handset space.

What Icera offers that’s unique is what it calls a software baseband. All cellular baseband processors require some sort of on-board processing, which is usually handled by a licensed ARM core, fixed function logic or combination of the two. Icera’s baseband chips feature its own custom designed microprocessor (something Icera calls its DXP - Deep Execution Processor) that can be upgraded, through software, to support a wide range of wireless standards.

The Icera Livanto ICE8060 baseband for example supports HSPA category 8 (7.2Mbps) all the way up through category 24 HSPA+ (42Mbps) and 50Mbps multimode LTE. Presumably Icera charges more for the faster software packages, but it does allow its customers to build a single design that’s quickly adaptable to other networks.

Icera argues that its software baseband architecture results in a smaller overall die size and lower power consumption than the competition. If true, this would mean better levels of integration both in an SoC and in a handset as well as longer battery life. That being said, we all know better than to trust a vendor’s claims without hard evidence.

Icera did publish this diagram which shows the size of its baseband IC compared to solutions from Qualcomm and ST-E:

It’s unclear how similar Icera’s approach is to what companies like Qualcomm already do. Given that there’s only a single baseband listed on Icera’s product pages I suspect it’s at least a bit more configurable than the competition. NVIDIA’s press release makes mention of Icera’s patent portfolio, which is likely a big part of what NVIDIA is after in this acquisition.

Today’s announcement is mostly about NVIDIA continuing to take the SoC market seriously. If you believe that those who will ultimately succeed here will own as much of their own IP as possible, then NVIDIA is on the right path.

Nvidia's CEO Jen-Hsun Huang has given his opinion on the Tegra 2 products currently for sale, and he's not particularly pleased. Glancing at the market, it's hard to blame him. Motorola's Xoom—easily the highest-profile Tegra 2 tablet launch thus far—launched without Flash support, a functional microSD slot, or support for Verizon's 4G LTE network; early adopters had to ship the unit back to Motorola for appropriate hardware to be installed.


Initial reviews noted that both the apps and operating system were unstable, a problem Google has since blamed on the Xoom's rushed launch. The unit's price has also been problematic. Huang referred to all these points, noting to CNET that ""It's a point of sales problem. It's an expertise at retail problem. It's a marketing problem to consumers. It is a price point problem...The baseline configuration included 3G when it shouldn't have. "Tablets should have a Wi-Fi configuration and be more affordable."

Our own review of the Xoom, published here, was generally positive, but noted that microSD slot still isn't enabled as of early May. The Xoom is evolving, but still isn't finished.

It's been a hectic few months for Nvidia—Samsung's Galaxy Tab 10.1 originally used the company's Tegra 2, right up to the point when the iPad 2 sent company execs sprinting for the drawing board. While the original GT 10.1 (now known as the 10.1V) still features Tegra 2, the 'new' Galaxy Tab 10.1 only refers to a dual-core 1GHz CPU. We're betting that Samsung's own Exynos 4210 SoC powers the graphics of the updated Tab 10.1, just as it powers the Galaxy S II in certain markets.

One point in Nvidia's favor is that none of the issues surrounding Honeycomb and the latest run of Tablets Made By Companies Who Aren't Apple point to a problem with Tegra 2. Unfortunately, the product's strongest points have been blurred by a rogue's gallery of problems. If the Xoom is an example of a poorly-positioned product rushed out the door, the Atrix 4G is a solid phone with an insanely priced feature.

When the Atrix debuted, all eyes were on the phone's ability to meld seamlessly with a dock that gave it an 11.6" screen at 1366x768, a nearly full-size keyboard and trackpad, and a slick 'webtop' interface. Motorola, realizing they had a real winner, apparently felt it had no choice but to kill the product in favor of more Razr knockoffs. The dock—2.4lbs of molded plastic—sells for $300 when bundled with the phone, $500 if purchased separately. Worse, the label of "expensive" inevitably bleeds back over to the phone itself—the same linkage that made the Atrix seem unique at debut also makes the phone appear expensive, even if the phone sans dock is reasonably priced.

All things considered, we're disappointed with the first iteration of Tegra 2 tablets as well. Virtually all of the problems, from Honeycomb's instability, to a dearth of software, are best categorized as growing pains—issues we're confident will be resolved as manufacturers gain a better sense of the product features and price points consumers prefer.