The cut-throat market for the chips that power smartphones and handhelds just got more competitive. In the past few days, no fewer than five semiconductors companies, including Samsung, ST Microelectronics, Texas Instruments, Qualcomm, and, yes, even Nvidia, have all announced new application processors:

• Nvidia APX 2500 press release
• Qualcomm QST-series press release
• Samsung S3C6410 press release
• ST Microelectronics Nomadik STn8820 press release
• TI OMAP3440 press release

I won’t spend a lot of time detailing the specs of each of these. (I’ve included some links to more detailed analyses of several at the end of this post.) Instead, I want to highlight some of the trends in these mobile chips, which don’t get quite as much attention as all the new eye-candy from Sony Ericsson, Samsung, LG and others announced at GSMA Mobile World Congress in Barcelona this week, but are growing in significance.

Traditionally, cell phone design has been dictated by the telecommunications companies. They have long design cycles (in other words, it takes a long time to get a handset off the drawing board and into production), use low-cost components and little memory, and are geared primarily for making and receiving calls. Data services have been slow to catch on, especially in the U.S., because of a poor user experience and high prices.

Apple’s iPhone upset this industry dynamic. Even in comparison to users of other smartphones, it’s clear that iPhone users spend significantly more time browsing the Web, watching YouTube videos, checking stock prices, and so on. Over the holidays, the iPhone accounted for more traffic to Google servers than any other mobile device, despite the fact that it has a fraction of the worldwide market share held by smartphones based on the Symbian, Windows Mobile or BlackBerry operating systems. Why? It’s a better browser.

The software gets most of the credit. But the features that make the iPhone a superior smartphone–a sophisticated user interface; a large, high-resolution display; high-quality video playback–all require more powerful hardware too. Just as Windows Vista’s Aero interface needs a decent processor (and lots of memory), a better smartphone requires a powerful application processor. In fact, the mystery application processor in the iPhone is based on the same ARM 11 series core found in the new Samsung S3C6410, and is probably very similar, even though Apple uses different model numbers.

Read the rest of this entry »

At the International Solid-State Circuits Conference in San Francisco this week, Intel revealed details of its new low-power processor, which promises to deliver performance similar to a low-end laptop in a package that uses only a bit more power than a smartphone. The design is innovative–and significant since it marks the start of a new effort by Intel (remember XScale?) to conquer the handheld world. But does the world really want Silverthorne-based Ultra-Mobile PCs (UMPCs) and Mobile Internet Devices (MIDs)?

Silverthorne is hardly a new name. Intel first talked up the chip at the Intel Developer Forum in Beijing last spring. And at CES earlier this year, it showed prototypes of UMPCs and MIDs from 15 different companies all based on Menlow, the new platform that includes both the Silverthorne processor and a chipset named Poulsbo. But this is the first time they’ve discussed the new microarchitecture in detail.

Intel manufactures Silverthorne using 45nm process technology with a high-k dielectric (a high-resistance insulator based on hafnium rather than the usual silicon dioxide) and metal gates. This is the same technology used in the new Penryn desktop and laptop chips. (Since introducing it last November, Intel has already shipped 32 Penryns for desktops, laptops, servers and workstations.) Silverthorne consists of 47 million transistors on a die that measures only 24.2 square millimeters. To put that in perspective, the dual-core Penryn measures 107 square millimeters. You need a small chip to build an ultra-mobile gadget, of course. But it also makes Silverthorne cheaper to manufacture since Intel can fit thousands of chips on a single silicon wafer.

Though it can reach speeds of 2.5GHz, Silverthorne will likely be available at clock speeds between 1.5GHz and 2.0GHz with a TDP ranging from 0.6 to 2 watts. This is peak power, and Intel was quick to point out that the power consumption under more typical usage and during idle time will actually be much lower–measured in “hundreds of milliwatts.” The ultra low voltage versions of the Core 2 Duo, such as the U7700, have a TDP of 10W, but a more accurate comparison is McCaslin, the current platform for UMPCs that uses Intel A100 and A110 chips with a TDP of 3W.

Read the rest of this entry »