News

Memory Roundup: Ultra-low-power SRAM, ULTRARAM, & 3D Flash Hit the Scene

April 24, 2023 by Arjun Nijhawan

New memory technologies have emerged to push the boundaries of conventional computer storage.

At a high level, a computer architecture has several key components: a CPU (composed of computational cores and interconnects), memory (made up of caches, main memory, and the hard drive), and hardware for input and output. While all parts of the architecture are important to overall efficiency, with modern workloads, the main performance bottleneck is memory. As a result, significant research has emerged in recent years attempting to improve the speed of memory while retaining as much power efficiency as possible. 

 

Intel Core i7

Intel Core i7, which is a common high-performance PC computer architecture today. Image courtesy of Intel

 

In this article, we'll discuss three recent memory announcements from both industry and academia to assess how the field is progressing. We'll start with a brief review of memory technologies to underscore the importance of the new developments from SureCore, the University of Lancaster, and Western Digital and Kioxia. 

 

A Brief Review of Memory Technologies

Memory can broadly be categorized into two types: non-volatile and volatile. Non-volatile memory can store its contents through millions of power cycles. The contents of memory remain until they are intentionally erased or overwritten. Two examples of non-volatile memory are flash memory and HDDs. By contrast, volatile memory loses its contents after a power cycle. Two common types of volatile memory are SRAM and DRAM. 

While SRAM and DRAM are both volatile, DRAM is generally slower because of the way it's implemented. In DRAM, each bit is stored using a capacitor. Since capacitors lose their charge unless a potential difference is continuously applied across the plates, DRAM needs to be refreshed periodically to prevent data loss. This periodic refreshing results in high latency and, consequently, a slower memory. SRAM, on the other hand, does not use capacitors for data storage. Instead, it uses several transistors—referred to as an SRAM cell—to store a bit. An SRAM can be written to and read from using a grid of bitlines and wordlines. 


An SRAM cell can comprise six MOSFETs

An SRAM cell can comprise six MOSFETs. Bitlines and worldlines are used to write a value to the cell. Image courtesy of TU Wien

 

To write to a typical SRAM cell, the associated bitlines are each driven low or high depending on the desired value, and then the worldline is driven high. To read from a typical SRAM cell, both bitlines are driven high followed by the worldline being driven high. In this way, an SRAM can distinguish between read and write operations. 

 

Bitlines and wordlines in an SRAM

Bitlines and wordlines in an SRAM control access to individual bits. Image courtesy of ChipEstimate

 

For embedded and mobile computing applications such as smartphones, designers use SRAM, which is as power-efficient as possible to extend the battery life of the device. Power dissipation can be active or passive due to leakage current. In SRAM, parasitic capacitance causes the movement of charge on and off the memory circuitry, resulting in active power dissipation.

 

SureCore Incorporates Power-saving Techniques Into SRAM 

A British company called SureCore specializing in ultra-low power embedded IP has developed a patented technology called the Cascode Precharge Sense Amplifier (CPSA). The company claims this technology can significantly reduce both active and passive power dissipation on SRAM.

CPSA works by controlling the bitline voltage swing on the SRAM, which can vary significantly due to the manufacturing process and consequently reduce power dissipation. While the individual bitline voltage swing is generally small on an SRAM, there are so many bitlines that the total bitline swing actually accounts for the majority of active power dissipation on the SRAM. 

 

Lancaster Researchers Create "ULTRARAM"

Recent research into improved memory has also focused on non-volatile storage. Researchers at the University of Lancaster have announced a spinoff company to create ULTRARAM, a memory technology that combines the non-volatility of flash memory with the performance and power benefits of DRAM.

ULTRARAM leverages quantum resonant tunneling in the semiconductor material comprising the RAM. The semiconductor compounds used in ULTRARAM are of the 6.1 Angstrom family, such as GaSb, InAs, and AlSb, which are particularly well-suited to high-speed designs. While the physics behind quantum resonant tunneling is incredibly complex, it has essentially allowed the researchers to create a memory that is non-volatile, does not consume much power, and offers speeds faster than DRAM

 

ULTRARAM

ULTRARAM is stable due to the unique quantum properties of the semiconductor compounds used. Image courtesy of IEEE Transactions on Electron Devices

 

In ULTRARAM, each logic state is stored in a floating gate. Due to the properties of 6.1A semiconductors, the floating gate can be switched from a highly resistive to highly conducting state with low voltage. These properties are what make it fast and power-efficient yet also non-volatile. 

 

Kioxia and Western Digital Team Up for 3D Flash

Cell density is another very important property of non-volatile memories because it allows more cells to be packed into the same area and increases the capacity of the storage. In March, Western Digital and Kioxia announced details of a new 3D flash memory technology. In this technology, each cell wafer is manufactured separately and then bonded together to maximize bit density. The memory is both vertically and laterally scaled to increase the bit density and, consequently, the capacity of the flash memory. 

 

Innovating Modern Computers Starting With Memory

Modern computers are complex machines. With several cores, multiple levels of cache memory, main memory, and sophisticated multithreading and prediction mechanisms, modern designs are vast and intricate. However, the fundamental goal of a modern computer architect has remained largely the same for several decades: to create a machine with enough computing power while still being power efficient and affordable. These researchers and companies hope to strike that balance by innovating different memory technologies—be it ultra-low-power SRAM, ULTRARAM, or 3D Flash. 

 


 

Featured image (Scanning electron microscope image of two ULTRARAM memory arrays) courtesy of IEEE Transactions on Electron Devices