Samsung’s revised view on Computational Storage Devices (CSD) could be another key factor in high performance computing (HPC) environments. With AMD Xilinx Versal FPGAs, Samsung’s new product reduces computing time by 50%, power consumption by 70%, and CPU usage by around 97% compared to the execution cost of traditional SSD devices. I promise to do that. For frequently accessed data. Smart SSDs are even smarter.
Samsung’s second-generation Smart SSDs, made possible by the recent AMD-Xilinx marriage (ending on February 14, like any other mainstream love story), can process the data itself. This may seem like finding a solution to a problem that doesn’t exist, but in reality it’s one of the most pressing bottlenecks in HPC, the flow of data between different elements of the compute path. I will solve it.
Data is processed directly within the SSD, so there is no need to shuffle data between storage banks. Not only does this free up bandwidth for the flow of other data, but it also helps reduce latency because there are no additional trips between the central processing unit and the data storage device itself. It also frees the system’s CPU for more performance-critical tasks, rather than handling reading and writing data to the CSD’s memory bank.
This is a negligible cost to the end user in 99% of the scenarios, but as the amount of data processed increases, its reading and writing puts a heavy load on the system. Since the extreme computational densities of modern systems remain the same, even a small amount of processing can be directed to components that are relatively passive compared to the CPU to better disperse the generated heat.
Samsung’s 2nd generation Smart SSD has a built-in Arm core. In particular, it is compatible with low power and low temperature environments for SSD deployment in supercomputers.
According to Samsung, CSD specifically targets the AI / ML inference market, business intelligence analytics, big data analytics, video transcoding, financial services, genomics, search queries, and transparent compression.
Choi Jin-hyuk, Executive Vice President and Head of Memory Solution Product Development at Samsung Electronics, said in a press release: “The upgraded processing capabilities of second-generation Smart SSDs make it easier for Samsung to meet the growing customer needs of the database and video transcoding sector as it expands the boundaries of the next-generation storage market.”
At the very least, it helps free up CPU cores for these extra percentage point work. The power budget of the supercomputer remains the same, allowing you to do more work within the same CPU or GPU, with the benefit of a 97% reduction in CPU load compared to regular SSDs.
As a reminder, this isn’t the first computational SSD we’ve covered-its honor belongs to the NGD system. ” Chia-AutoPlot CSDThis is intended to provide Chia users with a way to achieve the same macro level results as Samsung’s SmartSSD, but to mine time and space proof Chia cryptocurrencies.
