Yesterday was Qualcomm’s annual Analyst Day, and the company commemorated the occasion with a blitz of marketing material and corporate presentations. Qualcomm (QCOM) is currently a darling of the tech industry with a steadily rising NASDAQ stock price around $60. Intel (INTC), the traditional IT bellwether, is hovering around $20, and down significantly from January 2012.
Company executives are obviously feeling good about the state of things: The first 40 pages of the 100+ page presentation [PDF] are a sustained fanfare of self-congratulation. We’re going to skip those, save to note that Qualcomm sees huge growth opportunities in every market segment and every developing country. Everything is awesome. Everything is Snapdragon.
Small cells are the future of cellular networks
By page 47, things start to get more interesting. A few months ago, Qualcomm acquired DesignArt Networks, a company that specialises in building small cell base stations and remote radio heads. Small cells are a class of device that includes femtocells, which are small-scale cell signal boosters that improve reception and data transmission speeds over an area of a few hundred meters, while conventional cell phone towers (macrocells) typically cover several kilometres.
Macrocells are like giant, super-sensitive microphones; they’re great for monitoring ambient sound in a large space or tracking a handful of speakers. Put them in an auditorium with a full house and multiple cast members, though, and their omnidirectional input becomes a liability. Increasing the sensitivity of the microphone may make distant cast members easier to hear, but it increases the level of background noise as well. Increasing the volume of the actors isn’t an answer either – it increases the chance that the input will be distorted.
The solution is to put a small microphone on each character, with a stage microphone deployed as a secondary solution. That, in a nutshell, is what Qualcomm is advocating for small cell development. The company believes that its technology is well-suited to improving overall cell phone penetration and can make a significant difference in coverage, even if small cell deployment is relatively low.
The other advantage of small cells is that they can be used by macrocells to help manage interference, thus serving a larger number of customers with a higher overall level of performance.
Here be dragons
Modem and wireless improvements are only one anchor of the company’s total strategy. According to Qualcomm, the new “Krait 300” is capable of matching the performance of 32nm HKMG chips that draw too much power to go into anything but a laptop.
The x86 SoC above is obviously Intel’s Medfield, while the 40nm competitor in green is almost certainly Tegra 3. The white line is Samsung’s dual-core Exynos Cortex-A15 CPU. The first yellow line presumably refers to the dual-core Snapdragon S4 designs that launched earlier this year while, the second line is the quad-core Snapdragon hardware that just debuted in the HTC Droid DNA over in the States.
Take these claims with a serious grain of salt, but the quad-core Snapdragon S4 Pro has been well-reviewed at multiple sites. The gap between the 32nm Cortex-A15 and Qualcomm’s Krait has less to do with architecture or superior process nodes, and is more likely the result of binning, SoC design choices, and needing to fit the SoC into a phone as opposed to a netbook-class device. Cortex-A15 is primarily a server/netbook chip, but that doesn’t mean Samsung can’t build smartphones around it.
Adreno 320 performance is also up significantly over what the previous generation delivered. It’s the first Adreno GPU to support OpenCL, and it adds support for OpenGL ES 3.0.
This slide, on the other hand, deserves an award for best self-delusion presented as fact. Clearly adding up the total capital expenditures for all of Qualcomm’s suppliers on all their various customers and manufacturing technologies is equal to the amount Intel spends on its own products. (Alright; moment of snark over).
Qualcomm’s roadmap for the first chunk of next year indicates that the company will focus on pushing the new Snapdragon S4 Pro/Prime products into new high-end devices.
The company is planning a new series of refreshed Krait and Adreno hardware, presumably to debut at the 20nm node.
Of the various ARM vendors, Qualcomm has become the company to beat. With TI’s OMAP out of the race, Nvidia and Samsung have a steep hill to climb.