A new technical paper titled “xTern: Energy-Efficient Ternary Neural Network Inference on RISC-V-Based Edge Systems” was published by researchers at ETH Zurich and Universita di Bologna.

Abstract
“Ternary neural networks (TNNs) offer a superior accuracy-energy trade-off compared to binary neural networks. However, until now, they have required specialized accelerators to realize their efficiency potential, which has hindered widespread adoption. To address this, we present xTern, a lightweight extension of the RISC-V instruction set architecture (ISA) targeted at accelerating TNN inference on general-purpose cores. To complement the ISA extension, we developed a set of optimized kernels leveraging xTern, achieving 67% higher throughput than their 2-bit equivalents. Power consumption is only marginally increased by 5.2%, resulting in an energy efficiency improvement by 57.1%. We demonstrate that the proposed xTern extension, integrated into an octa-core compute cluster, incurs a minimal silicon area overhead of 0.9% with no impact on timing. In end-to-end benchmarks, we demonstrate that xTern enables the deployment of TNNs achieving up to 1.6 percentage points higher CIFAR-10 classification accuracy than 2-bit networks at equal inference latency. Our results show that xTern enables RISC-V-based ultra-low-power edge AI platforms to benefit from the efficiency potential of TNNs.”

Find the technical paper here. Published May 2024.

Rutishauser, Georg, Joan Mihali, Moritz Scherer, and Luca Benini. “xTern: Energy-Efficient Ternary Neural Network Inference on RISC-V-Based Edge Systems.” arXiv preprint arXiv:2405.19065 (2024).

The post Efficient TNN Inference on RISC-V Processing Cores With Minimal HW Overhead appeared first on Semiconductor Engineering.

Gone are the days of expensive tapeouts only done by commercial companies. Thanks to Tiny Tapeout, students, hobbyists, and more can design a simple ASIC or PCB design and actually send it to a foundry for a small fraction of the usual cost. Learners from all walks of life can use the resources to learn how to design a chip, without signing an NDA or installing licenses, faster than ever before. Whether you’re a digital, analog, or mixed-signal designer, there’s support for you.

We’re excited to support our academic network in participating in this initiative to gain more hands-on experience that will prepare them for a career in the semiconductor industry. We have professors incorporating it into the classroom, giving students the exciting opportunity to take their ideas from concept to reality.

“It gives people this joy when we design the chip in class. The 50 students that took the class last year, they designed a chip and Google funded it, and every time they got their design on the chip, their eyes got really big. I love being able to help students do that, and I want to do that all over the country,” said Matt Morrison, associate teaching professor in computer science and engineering, University of Notre Dame.

We also advise and encourage extracurricular design teams to challenge themselves outside the classroom. We partner with multiple design teams focused on creating an environment for students to explore the design flow process from RTL-to-GDS, and Tiny Tapeout provides an avenue for them.

“Just last year, SiliconJackets was founded by Zachary Ellis and me as a Georgia Tech club that takes ideas to SoC tapeout. Today, I am excited to share that we submitted the club’s first-ever design to Tiny Tapeout 6. This would not have been possible without the help from our advisors, and industry partners at Apple and Cadence,” said Nealson Li, SiliconJackets vice president and co-founder.

Tiny Tapeout also creates a culture of knowledge sharing, allowing participants to share their designs online, collaborate with one another, and build off an existing design. This creates a unique opportunity to learn from others’ experiences, enabling faster learning and more exposure.

“One of my favorite things about this project is that you’re not only going to get your design, but everybody else’s as well. You’ll be able to look through the chips’ data sheet and try out someone else’s design. In our previous runs, we’ve seen some really interesting designs, including RISC-V CPUs, FPGAs, ring oscillators, synthesizers, USB devices, and loads more,” said Matt Venn, science & technology communicator and electronic engineer.

Tiny Tapeout is on its seventh run, launched on April 22, 2024, and will remain open until June 1, 2024, or until all the slots fill up! With each run, more unique designs are created, more knowledge is shared, and more of the future workforce is developed. Check out the designs that were just submitted for Tiny Tapeout 6.

What can you expect when you participate?

• Access to training materials

• Ability to create your own design using one of the templates

• Support from the FAQs or Tiny Tapeout community

Interested in learning more? Check out their webpage. Want to use Cadence tools for your design? Check out our University Program and what tools students can access for free. We can’t wait to see what you come up with and how it’ll help you launch a career in the electronics industry!

The post How Quickly Can You Take Your Idea To Chip Design? appeared first on Semiconductor Engineering.

Silicon chips are central to today’s sophisticated advanced driver assistance systems, smart safety features, and immersive infotainment systems. Industry sources estimate that now there are over 1,000 integrated circuits (ICs), or chips, in an average ICE car, and twice as many in an average EV. Such a large amount of electronics translates into kilowatts of power being consumed – equivalent to a couple of dishwashers running continuously. For an ICE vehicle, this puts a lot of stress on the vehicle’s electrical and charging system, leading automotive manufacturers to consider moving to 48V systems (vs. today’s mainstream 12V systems). These 48V systems reduce the current levels in the vehicle’s wiring, enabling the use of lower cost smaller-gauge wire, as well as delivering higher reliability. For EVs, higher energy efficiency of on-board electronics translates directly into longer range – the primary consideration of many EV buyers (second only to price). Driver assistance and safety features often employ redundant component techniques to ensure reliability, further increasing vehicle energy consumption. Lack of energy efficiency for an EV also means more frequent charging, further stressing the power grid and producing a detrimental effect on the environment. All these considerations necessitate the need for a comprehensive energy-efficient design methodology for automotive ICs.

What’s driving demand for compute power in cars?

Classification and processing of massive amounts of data from multiple sources in automotive applications – video, audio, radar, lidar – results in a high degree of complexity in automotive ICs as software algorithms require large amounts of compute power. Hardware architectural decisions, and even hardware-software partitioning, must be done with energy efficiency in mind. There is a plethora of tradeoffs at this stage:

• Flexibility of a general-purpose CPU-based architecture vs. efficiency of a dedicated digital signal processor (DSP) vs. a hardware accelerator
• Memory sub-system design: how much is required, how it will be partitioned, how much precision is really needed, just to name a few considerations

In order to enable reliable decisions, architects must have access to a system that models, in a robust manner, power, performance, and area (PPA) characteristics of the hardware, as well as use cases. The idea is to eliminate error-prone estimates and guesswork.

To improve energy efficiency, automotive IC designers also must adopt many of the power reduction techniques traditionally used by architects and engineers in the low-power application space (e.g. mobile or handheld devices), such as power domain shutoff, voltage and frequency scaling, and effective clock and data gating. These techniques can be best evaluated at the hardware design level (register transfer level, or RTL) – but with the realistic system workload. As a system workload – either a boot sequence or an application – is millions of clock cycles long, only an emulation-based solution delivers a practical turnaround time (TAT) for power analysis at this stage. This power analysis can reveal intervals of wasted power – power consumption bugs – whether due to active clocks when the data stream is not active, redundant memory access when the address for the read operation doesn’t change for many clock cycles (and/or when the address and data input don’t change for the write operation over many cycles), or unnecessary data toggles while clocks are gated off.

To cope with the huge amount of data and the requirement to process that data in real time (or near real time), automotive designers employ artificial intelligence (AI) algorithms, both in software and in hardware. Millions of multiply-accumulate (MAC) operations per second and other arithmetic-intensive computations to process these algorithms give rise to a significant amount of wasted power due to glitches – multiple signal transitions per clock cycle. At the RTL stage, with the advanced RTL power analysis tools available today, it is possible to measure the amount of wasted power due to glitches as well as to identify glitch sources. Equipped with this information, an RTL design engineer can modify their RTL source code to lower the glitch activity, reduce the size of the downstream logic, or both, to reduce power.

Working together with the RTL design engineer is another critical persona – the verification engineer. In order to verify the functional behavior of the design, the verification engineer is no longer dealing just with the RTL source: they also have to verify the proper functionality of the global power reduction techniques such as power shutoff and voltage/frequency scaling. Doing so requires a holistic approach that leverages a comprehensive description of power intent, such as the Unified Power Format (UPF). All verification technologies – static, formal, emulation, and simulation – can then correctly interpret this power intent to form an effective verification methodology.

Power intent also carries through to the implementation part of the flow, as well as signoff. During the implementation process, power can be further optimized through physical design techniques while conforming to timing and area constraints. Highly accurate power signoff is then used to check conformance to power specifications before tape-out.

Design and verification flow for more energy-efficient automotive SoCs

Synopsys delivers a complete end-to-end solution that allows IC architects and designers to drive energy efficiency in automotive designs. This solution spans the entire design flow from architecture to RTL design and verification, to emulation-driven power analysis, to implementation and, ultimately, to power signoff. Automotive IC design teams can now put in place a rigorous methodology that enables intelligent architectural decisions, RTL power analysis with consistent accuracy, power-aware physical design, and foundry-certified power signoff.

The post Maximizing Energy Efficiency For Automotive Chips appeared first on Semiconductor Engineering.