News
Google Details Five Generations Of TPU Training Supercomputers
1+ week, 1+ day ago (267+ words) Researchers from Google and University of California, Berkeley published a technical paper titled "Google's Training Supercomputers from TPU v2 to Ironwood: Architectural Stability, Scale, Resilience, Power Efficiency, and Sustainability Across Five Generations." The paper summarizes five generations of Google TPUs, from…...
Modeling Multi-GPU Traffic For Distributed AI Workloads (UW Madison, AMD)
1+ week, 2+ day ago (326+ words) Researchers from University of Wisconsin-Madison and AMD Research and Advanced Development published a technical paper titled "Eidola: Modeling Multi-GPU Network Communication Traffic in Distributed AI Workloads." Abstract: "As distributed AI workloads grow in scale, multi-GPU systems have become essential for…...
Foundation IP: Pushing the Boundaries of Energy-Efficient Chip Design
2+ week, 18+ min ago (182+ words) Semiconductor Engineering Foundation IP: Pushing the Boundaries of Energy-Efficient Chip Design Energy-efficient So C design, optimizing PPA, deep low-voltage operation, and advanced power management techniques. Access "Foundation IP: Pushing the Boundaries of Energy-Efficient Chip Design" to explore six articles that…...
How To Start Building Edge-Native AI
2+ week, 13+ min ago (355+ words) Cloud AI enables features like voice assistants and recommendations via centralized data centers, but it relies on consistent network connectivity, which often fails in real-world conditions. Edge-native AI shifts inference to devices such as phones, cars, and sensors, enabling real-time…...
Beyond PCIe Compliance: Why Stress Testing Is Crucial For Edge AI Deployments
3+ week, 14+ min ago (452+ words) Replicate real-world operating conditions to determine how much margin there is and what happens when operating conditions begin to consume it. Compliance testing validates conformance: does this device interoperate correctly within the ecosystem and meet specification-defined electrical parameters? It is…...
Beating the Edge AI Power Wall with Low Voltage Foundation IP
3+ week, 1+ day ago (195+ words) How to achieve reliable near-threshold operation while accelerating design convergence and meeting aggressive PPA targets. The post Beating the Edge AI Power Wall with Low Voltage Foundation IP appeared first on Semiconductor Engineering. Edge AI is pushing the limits of…...
Improving GPU Energy Efficiency With Component-Level Power Management (AMD)
4+ week, 2+ day ago (252+ words) Researchers from AMD released "Comp Pow: A Case for Component-level GPU Power Management. Abstract "The ever increasing demand for ML-driven intelligence in a wide spectrum of domains has led to ubiquity of GPUs. At the same time, GPUs are notorious…...
Why Vision LLMs Force A Rethink Of Edge AI Hardware
1+ mon, 1+ week ago (990+ words) Peak TOPS is becoming a weaker proxy for delivered edge performance. As vision-centric large language models move on-device, performance measured in raw TOPS is no longer enough. Architectures need to be built around real workloads, memory behavior, and sustained utilization,…...
The AI Server Challenge: Testing Power At Scale
1+ mon, 1+ week ago (687+ words) Why next-gen AI architectures demand purpose-built power test systems. Artificial intelligence is most often framed as a story of compute advancements. Faster GPUs, denser accelerators, and advanced process nodes. But behind every AI workload, the most fundamental constraint is power....
Designing GPUs For Developers: A Conversation With Godot
1+ mon, 2+ week ago (1188+ words) The evolution of game engines, the realities of mobile performance, and what GPU vendors need to deliver to support the next generation of developers. Godot has rapidly established itself as one of the most important graphics engines in today's ecosystem....