Research Archive

December 20, 2024 / Ben Bajarin

The GPU vs. Custom ASIC Competitive Landscape: A Deeper Cost-Performance Analysis

Recent first-party benchmarking data provides crucial insight into the cost-performance dynamics between custom ASICs and GPUs across both training and inference workloads. The data reveals a nuanced competitive landscape where ASICs demonstrate meaningful cost-performance advantages, particularly in inference scenarios, though the implications for market share are more complex than raw performance metrics might suggest. Cost-Performance…

November 1, 2024 / Max Weinbach

Tesla AI’s New Architecture with FSD v13 Software Stack

Key Points Tesla AI FSD v13 Update Highlights: Tesla has shipped its latest version to 50,000 customers, enhancing highway autonomy significantly. Version 13 (v13) Upgrade: v13 introduces major scaling and architecture improvements, including 3x model scaling, audio inputs for emergency vehicles, and a redesigned control system for smoother driving. Silicon Performance: AI4/HW4 silicon achieves 1.3…

September 24, 2024 / Max Weinbach

After a few days with Intel Lunar Lake, Intel is back!

Key Takeaways: Intel is back(!) with a really good, really efficient core design. Intel’s tiled SoC with TSMC N3B compute tile and N6 I/O tile works really well We’re at a point where performance and efficiency among all major PC silicon providers is more or less equal, moving the differentiation from silicon to actual device…

August 15, 2024 / Max Weinbach

Geekbench AI and the State of the NPU

Key Takeaways: Geekbench AI 1.0 is a new cross-platform benchmark suite designed for machine learning, deep learning, and AI-centric workloads. It provides three main scores: Single Precision, Half Precision, and Quantized, reflecting different precision levels used in AI tasks. The benchmark includes both computer vision and natural language processing workloads. Accuracy measurements are incorporated alongside…

August 7, 2024 / Ben Bajarin

Data Center Cooling Innovation: A Critical Imperative

As data centers continue to proliferate and expand to meet the ever-growing demands of our digital world, the need for innovative cooling solutions has become increasingly urgent. This report examines why new methods of cooling data centers are critical, considering environmental, technological, economic, and operational factors. The rise of artificial intelligence (AI) and high-density computing…

June 13, 2024 / Ben Bajarin

Key Points in the Infiniband vs. Ethernet Debate

We have been digging through a host of resources, speaking with technical executives, and investors, to summarize a few key points in the debate about Infiniband vs. Ethernet for AI networking in the data center.  Below are the key point and pros and cons of each. The comparative analysis of Ethernet versus Infiniband for AI…

Trusted by 80% of the top 10 Fortune 500 technology companies