Integrating Ethernet, PCIe, And UCIe For Enhanced Bandwidth And Scalability For AI/HPC Chips


By Madhumita Sanyal and Aparna Tarde Multi-die architectures are becoming a pivotal solution for boosting performance, scalability, and adaptability in contemporary data centers. By breaking down traditional monolithic designs into smaller, either heterogeneous or homogeneous dies (also known as chiplets), engineers can fine-tune each component for specific functions, resulting in notable im... » read more

Why PCIe And CXL Are Essential Interconnects For The AI Era


As the demand for AI and machine learning accelerates, the need for faster and more flexible data interconnects has never been more critical. Traditional data center architectures face several challenges in enabling efficient and scalable infrastructure to meet the needs of emerging AI use cases. The wide variety of AI use cases translate into different types of workloads. Some require high ... » read more

Addressing Reset Tree Design Challenges For Complex SoCs With Advanced Structural Checks


As SoC designs continue to evolve, the complexity of reset architectures has grown significantly. Traditionally, clock tree synthesis has been a major focus due to timing challenges, but now reset trees demand equal attention. With multiple reset sources, designers must deal with reset trees that can be more intricate than clock trees. Errors within a reset tree can lead to serious issues, incl... » read more

Automakers And Industry Need Specific, Extremely Robust, Heterogeneously Integrated Chiplet Solutions


Chiplets offer great potential for the automotive and industrial sectors, especially as these applications often have high performance requirements but are needed only in small quantities. The modular principle behind chiplets enables efficient design and production: individual components have to be produced only once and can then be flexibly combined to create tailored solutions. This offers m... » read more

NPU Acceleration For Multimodal LLMs


Transformer-based models have rapidly spread from text to speech, vision, and other modalities. This has created challenges for the development of Neural Processing Units (NPUs). NPUs must now efficiently support the computation of weights and propagation of activations through a series of attention blocks. Increasingly, NPUs must be able to process models with multiple input modalities with ac... » read more

Dual-Grid Interpolation: For Improved Accuracy Of Overset Grid Systems


Computational fluid dynamics (CFD) has become an integral part of engineering decision-making, providing a deeper understanding of how fluids behave in various scenarios, from the high skies in aerospace all the way to the fast-paced realm of automotive engineering. The task of accurately simulating fluid dynamics, particularly when faced with complex shapes or moving parts, demands innovative ... » read more

Ghostbusting With Simulation: Solving Engineering Challenges In Automotive Radar Development


Probably the biggest trick to the adoption of full autonomy in the automotive space is learning how to safely achieve a level of perception that matches that of a human driver. Carmakers are rising to the challenge with a combination of advanced camera, radar, and lidar sensing technologies, machine learning, and artificial intelligence that makes self-driving possible. This includes the adva... » read more

The Future Of AI For Games


Earlier this month, I had the pleasure of attending the inaugural AI and Games Conference at Goldsmiths in London, for which Arm was an associate sponsor. Hosted by Dr. Tommy Thompson, and borrowing its name from his AI and Games YouTube channel, the day really delivered on the promise of bringing experts and enthusiasts (and subscribers) together for interesting talks on the intersecti... » read more

Small Language Models: A Solution To Language Model Deployment At The Edge?


While Large Language Models (LLMs) like GPT-3 and GPT-4 have quickly become synonymous with AI, LLM mass deployments in both training and inference applications have, to date, been predominately cloud-based. This is primarily due to the sheer size of the models; the resulting processing and memory requirements often overwhelm the capabilities of edge-based systems. While the efficiency of Exped... » read more

To (B)atch Or Not To (B)atch?


When evaluating benchmark results for AI/ML processing solutions, it is very helpful to remember Shakespeare’s Hamlet, and the famous line: “To be, or not to be.” Except in this case the “B” stands for Batched. Batch size matters There are two different ways in which a machine learning inference workload can be used in a system. A particular ML graph can be used one time, preced... » read more

← Older posts