12/31/2023 0 Comments Making a cpu transistorIf we look at a high-level block diagram of the MLSoC, we see that-in addition to the machine learning accelerator (MLA)-it contains a quad-core Arm Cortex-A65 application processor (AP), a Synopsys EV74 Embedded Vision Processor (EVP) with deep neural network (DNN) accelerator, and… a whole bunch of other IP blocks. Another key differentiator is that they do all of this at very low power. By comparison, one of the MLSoC’s key differentiators is that it’s a complete system that combines host processor and ML accelerator capabilities in one device. In many ML applications, most of the heavy lifting is performed by a host processor, and only the ML-centric portions of the algorithms are offloaded to an ML accelerator. As Srivi says, this means that anything to do with the hardware side of the MLSoC-from architecture to logical design to logical verification to physical design to physical verification to bring-up-is his responsibility. Well, I was recently chatting with Srivi Dhruvanarayan, who is VP of Hardware Engineering at SiMa.ai. More specifically, they wanted to provide users with the ability to solve any computer vision application challenge with a minimum of 10x better performance in terms of frames per second per watt than any competitive offerings while delivering an overall push-button experience in minutes. In a crunchy nutshell, they wanted to design an MLSoC that could run any computer vision application, any network, any model, any framework, any sensor, and any resolution. Their goal when they founded SiMa.ai was to simplify the process of adding ML to products by delivering “The world’s first software-centric, purpose-built MLSoC platform with push-button performance for effortless ML deployment and scaling at the embedded edge so you can get your products to market faster” (pause to take a deep breath). The folks at SiMa.ai say that existing ML solutions aren’t purpose-built for the embedded edge, but rather are adapted to it, similar in concept to forcing a square peg into a round hole. Let’s start with SiMa.ai, which was founded in 2018, just four short years ago as I pen these words. Two related examples spring to mind: the guys and gals at SiMa.ai, whose goal it is to create a machine learning (ML) system-on-chip (SoC) called an MLSoC that can provide effortless machine learning at the embedded edge, and the chaps and chapesses at Arteris IP, whose mission it is to provide state-of-the-art Semiconductor IP and IP Deployment Technology. As I’ve been known to note, one of the great things about being me-in addition to being outrageously handsome, a trendsetter, and a leader of fashion-is that I get to talk to all sorts of interesting companies and people about their uber-cool technologies.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |