Ubicom Submits to Qualcomm
By Joseph Byrne
Remember Ubicom? That’s OK, most people don’t. The company had an innovative 32-bit architecture but never achieved the popularity its designers sought. Without fanfare, Qualcomm has acquired the company for an undisclosed (and likely paltry) sum, closing the book on a processor company that struggled for 16 years. Ubicom sought advantages in cost and flexibility by employing soft I/O and multithreading.
Instead of using hardware to implement controllers for PCI, Ethernet, and other I/O, software emulating the controllers twiddled I/O pins. This bit-banging technique eliminated the die area of dedicated I/O blocks and enabled customers to choose the best I/O combination for their design.
The company, founded in 1996 as Scenix Semiconductor, first applied bit banging to an 8051-compatible microcontroller running at a blazing 100MHz. After changing its name to Ubicom, the company then applied soft I/O to a new chip based on a 32-bit CPU of its own design. For area efficiency and I/O flexibility, the Ubicom32 architecture juggled eight or more simultaneous threads (see MPR 4/21/03, “Ubicom’s New NPU Stays Small”).
Ubicom also developed its own RTOS for its first 32-bit processors and provided a library of networking and multimedia functions. This software made good use of the processor’s capabilities and limited memory, but its development placed an additional burden on the small company. To alleviate this load, the company ported uCLinux and subsequently added an MMU so its processor could run a full version of Linux.
Wi-Fi gadgets were a critical application market for Ubicom. The company developed a symbiotic relationship with Atheros, which until recently lacked an access-point processor of its own. Other opportunities for standalone processors started to dry up as process scaling enabled broadband and Wi-Fi ASSPs to absorb computing and interfacing functions. Ubicom’s attempts to diversify into digital picture frames and media adapters faltered as these markets never grew beyond early adopters.
Ultimately, Ubicom could not sustain operations. With modest revenue and a large software burden, the company could invest little in its microarchitecture. Performance gains came only from process scaling, minor tweaks, and new multimedia instructions. The benefits of soft I/O diminished with the advent of interfaces requiring hardware support, such as high-speed serial I/O (e.g., Gigabit Ethernet, PCI Express, and SATA), broadband (e.g., DSL and PON), and Wi-Fi. With an aging microarchitecture and no complementary technology, such as Wi-Fi or broadband, to combine with the processor, growth opportunities were limited.
Despite its lack of commercial success, Ubicom enticed Qualcomm to spend a small sum to acquire its proven engineering team and technology. The team will join the Qualcomm Atheros operation. Ubicom’s patents on multithreading and soft I/O will become part of Qualcomm’s vast portfolio. The fate of Ubicom’s products is unclear, but it’s likely that we’ve seen the last standalone Ubicom32 processor.
Microprocessor Report subscribers can access the full article:
list.linleygroup.com  |
