VLSI 2020: Empowering next-gen apps through flash innovation

Posted on June 22, 2020 Updated on June 22, 2020

Shigeo (Jeff) Ohshima, Technology Executive, SSD Application Engineering, Kioxia Corp., presented a paper at the concluding plenary session of the 2020 Symposia on VLSI Technology and Circuits, titled: Empowering next-gen apps through flash innovation.

He said that for the long-term future with ‘flash native solutions’, here is one real disruptive approach to introduce as an innovative example. By skipping the regular flash memory and SSD manufacturing processes such as assembling, packaging, and drive building, there will be a huge possibility to drastically reduce the cost and lead time for manufacturing.

This will enable flash native solutions to expand the market opportunities such as big data science, co-existing with extremely high data density and high performance with low flash storage cost. We tentatively call it wafer-level SSD as one of the disruptive ‘flash native solutions.

Wafer-level SSD
With wafer-level SSD concept, all of the flash dies in each wafer are probed and operated simultaneously, managed by SSD controller devices. This ‘super multi-probing’ enables parallel operation of hundreds of chips on a whole single wafer, which will enable huge performance, millions of IOPS.

Regarding the cost advantage of this wafer level SSD configuration, a ball park estimation of the storage cost in comparison with conventional configured SSDs is approximately as small as 20 percent, which is almost equivalent to that of HDDs. Not to mention, the overall performance is far beyond HDDs. It will be in line with acceleration of super multi-level cell flash adoptions, QLC, PLC, and HLC.

One future example of large-scale data analysis that could follow on the heels of AI will be life analysis. A ‘flash native solution’ to explore ways to contribute to this field, will be invaluable. For instance, in order to capture highresolution brain images of tens of nanometers scale using a regular optical microscope, a unique technique called expansion microscopy has been invented, that embeds the observed object in a water-absorbing polymer and physically expands it.

One of the major challenges of this technology was to efficiently process and display large amounts of image data output from the microscope. A ‘Flash native solution’ can contribute to build high-speed signal processing systems, effectively combining parallel computing devices such as CPUs and GPUs, and SSDs to accelerate reading and writing data, creating a 3D real-time visualization system that displays image data acquired from microscopes.

Such data is genuinely big data, which exceeds the size of DRAM. High-capacity and high-speed SSDs are indispensable. In the life science field, data volume is expected to further increase in the future, and data processing speed of more than 1TB (terabyte) per second will be required. We will continue to pursue new and exciting “flash native solutions” that will contribute to the meaningful use of the explosive amount of information being produced in our society.

Similarly, 3D also started with a paper by Toshiba back in 2007, when the first prototype of BiCS FLASH was introduced. 3D flash has enhanced technical wonders, such as module-based SSD for laptop PCs, enterprise SSD for data center and on premise storage and high performance UFS for smart phones. Innovations continue to roll forward like an everlasting technology showcase by various memory vendors.

Among all the ‘flash native solutions’, SSDs will play a cutting-edge role for disruptive innovations independently or together with enhanced software solutions. As a driving force in memory, we will create uplifting experiences and change the world.