News Coverage

(Mar 14 2023) SemiEngineering - AI inference benchmarks, such as MLPerf’s, do not represent all of the facets of AI compute relevant to application developers. Further, they distract SoC designers from optimizing performance for end-to-end AI applications that will enable them to capture the fast-growing AI hardware market. For the entire AI compute, there are at present no standardized benchmarks.

(Feb 22 2023)SemiEngineering - Dollars going into R&D, and eventually deployment, have been increasing rapidly over the past few years. In automotive, there are billions of lines of code that need to be updated every few months to keep up with new software new models. For that you have to secure the entire system.

(Feb 15 2023)SemiEngineering - Automotive is almost like a data center in that the software requirements, the models, that you run today will be completely different from what you might want to run 10 years from now. So the hardware being deployed must be powerful and flexible enough to handle different models and pre- and post-processing tasks and stochastics.

(Feb 9 2023)SemiEngineering - Generative ML models – such as Stable Diffusion 2.0 and DALL-E – are poised to radically shakeup the established, boring ubiquitous platforms powering desktops, laptops, and most tablets.

(Feb 2 2023)SemiEngineering - Full self-driving has always been something that’s coming in the future... When we say full self-driving, with no steering wheel, it’s going to be at least 10 to 15 years from today, based on the companies we talk with and the kinds of chipsets they want to drive those algorithms.

(Jan 16, 2023) SemiEngineereing - Building hardware prototypes is more than 10 to 50 times more time-consuming than creating a SystemC model of a subsystem or entire SoC. The SystemC virtual prototype generally runs fast enough to answer performance questions.

(Jan 13, 2023) SemiEngineering - Six or eight years ago when models began to explode in size (parameter count), the sheer volume of shuffling weight data into and out of training compute (either CPU or GPU) became the performance limiting bottleneck in large training runs. Faced with a choice of ever more expensive memory interfaces, such as HBM, or cutting bit precision in training, a number of companies experimented successfully with lower-precision floats. Now that networks have continued to grow exponentially in size, the exploration of FP8 is the next logical step in reducing training bandwidth demands.

(Jan 12, 2023) SemiEngineering - If headcount can’t grow, the use of IP must grow to speed up the design process. That, in turn, leads companies to focus on their real sources of value-add and differentiation, not their hoped-for sources of differentiation.

(Jan 11, 2023) Medium - Quadric caught our attention because they took a software-first approach to its edge AI chips, creating an architecture that controls data flow and enables all software and AI processing to run on a single programmable core. This eliminates the need for other ancillary processing and software elements and blends the best of current processing methods to create a single, optimized general purpose neural processing unit (GPNPU).

(Jan 11, 2023) SemiEngineering - Running a new ML operator on a slow DSP or CPU is the equivalent of charging your EV with an extension cord plugged into a 110V wall socket. What you want for your EV is an 800 Volt fast charger that tops off the battery in 20 minutes, not a low voltage, low amperage wall socket that needs 18 hours to recharge your car.