Latest Product: AI Chip to Speed Up Deep Neural Network Calculations for IoT-Edge
Interuniversity Microelectronics Centre (Imec) has achieved a unique breakthrough in the science of AI chip manufacturing. The AI chip showcases the years of research and development Imec has put into nano-electronics and digital technologies. Imec partnered with specialty foundry GLOBALFOUNDRIES® (GF®) to create the revolutionary artificial intelligence chip.
What is the AI Chip from Imec all About?
AI accelerators in the chip manufacturing industry are a new branch of hardware development. We are witnessing a massive growth in the AI-driven smart devices where OEMs are pushing these devices into our pocket, and into our brain. Imec’s AI accelerators are based on the advanced Analog in Memory Computing (AiMC) architecture. The new AI Chip is built on this AiMC architecture, leveraging GF’s 22FDX® solution.
At the time of this announcement, Diederik Verkest, Program Director for Machine Learning at Imec said —
“The successful tape-out of AnIA marks an important step forward toward validation of Analog in Memory Computing (AiMC). The reference implementation not only shows that analog in-memory calculations are possible in practice, but also that they achieve an energy efficiency ten to hundred times better than digital accelerators.”
Diederik added, “In imec’s machine learning program, we tune existing and emerging memory devices to optimize them for analog in-memory computation. These promising results encourage us to further develop this technology, with the ambition to evolve towards 10,000 TOPS/W”.
Recommended: Dell Technologies Explores VMware Spin-Off
Why is this chip different?
AI Chip is in-memory computing hardware in the analog domain. AI chip from Imec can optimize the entire networking infrastructure that goes into the compute engine of the deep neural network calculations. The AiMC architecture helps to achieve a record-high energy efficiency up to 2,900 TOPS/W, the accelerator is a key enabler for inference-on-the-edge for low-power devices.
Imec’s modern chip making science delivers on expectations related to the privacy, security and latency. Overall, these would have an impact on AI applications in a wide range of edge devices, from smart speakers to self-driving vehicles.
Deep Learning Calculations for the IoT-Edge Neural Networking
The processors and memory are two separate domains in the whole hardware infra. The capacity to process a large volume of data (Big Data analytics) has to be supported by equally powerful and larger memory storage units. For a very long period, Big Data processors stayed separated from the memory storage.
Imec has identified this limitation as part of the von Neumann architecture – also known as the bottleneck due to the von Neumann (VM) model or Princeton architecture. In simple words, Imec AI chips remove the VM bottleneck that have been overshadowing the actual computing time, especially in neural networks. For AI accelerators, neural computations are performed with the precision of a digital computer and require a significant amount of energy. Analog technology adds accuracy to the whole vector-matrix multiplications.
The Role of GF in Imec AI Chip
GF is imec’s industrial partner that is addressing the VM model bottleneck as part of the imec’s industrial affiliation machine learning program. GF has developed a new architecture that eliminates the von Neumann bottleneck by performing analog computation in SRAM cells. Imec Analog Inference Accelerator (AnIA) is built on GF’s 22FDX semiconductor platform.
AnIA has exceptional energy efficiency that can efficiently power the pattern recognition operations in tiny sensors and low-power edge devices. These devices are powered by machine learning in data centers, that can now be performed locally on these power-efficient AI accelerators.
Future Collaborations on AI Chip between Imec and GF
In the coming months, GF will include AiMC as a feature able to be implemented on the 22FDX platform for a differentiated solution in the AI market space.
“GlobalFoundries collaborated closely with imec to implement the new AnIA chip using our low-power, high-performance 22FDX platform,” said Hiren Majmudar, vice president of product management for computing and wired infrastructure at GF.
Hiren added, “This test chip is a critical step forward in demonstrating to the industry how 22FDX can significantly reduce the power consumption of energy-intensive AI and machine learning applications.”
GF’s 22FDX employs 22nm FD-SOI technology to deliver outstanding performance at extremely low power, with the ability to operate at 0.5 Volt ultralow power and at 1 pico amp per micron for ultralow standby leakage. 22FDX with the new AiMC feature is in development at GF’s state-of-the-art 300mm production line at Fab 1 in Dresden, Germany.
Currently, by leveraging the world-class infrastructure and local and global ecosystem of partners across a multitude of industries, GF creates groundbreaking innovation in application domains such as healthcare, smart cities and mobility, logistics and manufacturing, energy and education.