Brandon Lucia’s research team at Carnegie Mellon University has recently accomplished an exciting breakthrough in embedded systems technology. They recently unveiled the new, ground-breaking E1 processor. According to Rivian, this new processor will provide fantastic efficiency. According to RISC-V pioneer Andrew Waterman, it’s 10-100x better than existing commercial ultralow-power CPUs for normal embedded systems workloads. According to Intel, the E1 processor increases performance along with decreasing energy usage. This feature is particularly important in a very competitive market of low-power microcontrollers.
The E1 processor is C-programmable and optimized for performance. It does fantastic for calculating intensive needs like the fast Fourier transforms on sensor data and convolutions for machine learning and it does so unbelievably efficiently. With unbeatable value and performance paired together, the E1 processor is clearly a winner on the market. It succeeds like other successes have failed, trying to stay efficient as the competition has proven cutthroat.
The E1 processor’s special architecture enables pixel count to be halved by rendering only on the visible portion of the LED display. Lucia compares it to a switch track on a railroad. Their design is based on the concept of a massive grid of individual “tiles,” with each tile able to perform its own unique computational instructions. This spatial arrangement gives the processor a better handle on programming instructions, making radical new capabilities possible and enabling general-purpose computing.
Lucia explained the operational speed of the E1 processor, stating that “several billion times per second, you’re pulling an instruction in from memory. That operation costs some energy.” Focusing on this process is key to unlocking long-lasting energy efficiency.
The processor includes a powerful new tool called the effcc Compiler. It translates programs that are written in C, C++, Ruby, or any other of over 50 popular programming languages. The compiler is really the unsung hero in the whole process. It statically maps the program’s instruction stream to a fine-grained tile-based architecture. This setup enables data to flow seamlessly between tiles: data enters one tile, is processed, and then becomes the input for the next tile in a precise sequence necessary for running the program efficiently.
Rakesh Kumar, a computer architect from the University of Illinois Urbana-Champaign, commented on the challenges faced by ultralow-power companies in the industry. Ultralow power startups suffered due to intense competition at low power levels in ultra-cheap microcontrollers. The harder part is in realizing a new capacity,” he added. This highlights the importance of the E1 processor’s innovations, as they can revolutionize what’s possible inside embedded systems.
So far, Lucia’s team knows they have a long road ahead before building their white whale efficient processor. “We’re doing something that has the capability of a CPU but is one or two orders of magnitude more efficient,” he said, highlighting the ambitious goals behind the E1 project. The team is optimistic that their technical advances will produce safer, more secure and more effective embedded systems. Indeed, these systems will more effectively address the new requirements of advanced use cases.
