Weebit Nano – Breaking the Memory Wall

As AI models have ballooned in size and complexity, the storage needed for them to thrive is struggling to keep up. While chips and CPUs are sprinting towards smaller, faster futures, non-volatile memory (NVM) has hit a brick wall at 28nm. Weebit Nano, however, has made a breakthrough.

NVMs are a type of computer storage that retains data even when power is removed. Since they are used in a wide variety of applications, NVM requirements are stringent. They must be extremely low power to support IoT and battery-operated devices, while also being capable of withstanding the high temperatures and harsh conditions where automotive and industrial applications often operate. Finally, to meet the demand for emerging applications, such as AI and advanced IoT, NVMs need to be scalable to advanced process nodes.

Flash memory is the most widely used form of NVM, but embedded flash doesn’t scale well below 40nm. At 28nm, the technology reaches its limitations when it comes to speed, power, cost, and endurance. So, what’s the alternative?

Resistive RAM (ReRAM) is an emerging form of NVM that makes it possible to achieve the vision of a low-cost, low-power, single-chip solution. ReRAM’s scalability allows for the integration of large memory arrays into small form factors, enabling more complex AI models to be deployed on edge devices.

Weebit Nano’s ReRAM technology goes beyond the average ReRAM. It offers 10X-100X better endurance than flash, handling between 100,000 and 1 million write cycles versus the usual 10,000 program/erase cycles that flash can manage. It offers high performance at a lower cost and complexity while also being low on power consumption.

NVMs are a foundational element of current AI and a critical enabler of emerging AI trends. Weebit ReRAM, designed as the successor to flash, has the potential to enable advancements in areas such as edge AI and in-memory computing.

Of course, designing best-in-class IP requires best-in-class tools. Weebit Nano is a long-time user of Cadence tools. Due to its accuracy and performance, the team found Cadence Spectre X Simulator to be crucial for their signoff simulations, while the Spectre FX Simulator’s performance and seamless integration with Cadence’s Virtuoso Studio and Liberate MX Trio Characterization were incredibly helpful.

Additionally, Weebit Nano uses the Cadence Xcelium Logic Simulator for digital design and UVM verification of their ReRAM controller, which implements their proprietary programming algorithms. Learn more about how Weebit Nano is changing the NVM game with ReRAM technology.

“Designed with Cadence” is a series of videos that showcases creative products and technologies that are accelerating industry innovation using Cadence tools and solutions. For more Designed with Cadence videos, check out the Cadence website and YouTube channel.