Hardware Podcasts

143: I'm Thinking of Unicorns

Embedded.fm · 17 Mar 2016 · 1h 06m

What this episode covers

Dan Luu, a hardware and software engineer at Microsoft with prior experience at Centaur and Google, discusses the evolution of CPU features, the trade-offs between working at startups versus large tech companies, and the efficacy of computer science research. He highlights how specialized hardware accelerators can offer massive performance gains over general-purpose CPUs for specific workloads like deep learning, and critiques the software industry's neglect of error handling and rigorous testing compared to hardware design. Luu also shares insights from his blog, danluu.com, on topics ranging from compiler bugs to the persistent myths about startup compensation and the path dependence of career success.

Most relevant to

large enterprises

Topics

Mentioned

Companies
Centaur, Google, Microsoft, Intel, ARM, Transmeta, NVIDIA, Apple, Amazon, Facebook, Twitter
Chips
Alpha 21064
Also
ISCA, MICRO, ICSE, FSE, OSDI, Hacker News

Related episodes from other shows

Other shows on the same ground, closest first.

  • The Circuit · 43m

    Episode 15: Growth Drivers for Semiconductors in the Datacenter

    Ben Bajarin and Jay Goldberg discuss the shifting economics of data center semiconductors, highlighting that while total silicon content is growing, CPU share is declining as GPU, FPGA, and accelerator workloads expand. They analyze how hyperscalers like Google, Amazon, and Microsoft are pursuing custom silicon to control roadmaps and solve specific pain points, such as Google's TPU for YouTube, while noting the high risks for startups entering the accelerator market. The conversation emphasizes that heterogeneous compute is driving value, with high-intensity workloads like LLM training justifying expensive chips like Nvidia's H100, whereas commodity workloads remain on cheaper, parallelizable infrastructure.

    Chips, FPGAs & semiconductorsPricing, margins & unit economics
    7 chapters

  • Oxide and Friends · 1h 54m

    Heterogeneous Computing with Raja Koduri

    Raja Koduri, former chief architect at Intel and AMD, joins Bryan Cantrill and Adam Leventhal to explain the differences between CPUs, GPUs, FPGAs, and ASICs through historical and analogical lenses. Koduri traces the evolution from 3dfx and Direct3D to CUDA, arguing that NVIDIA’s success stems from a tight coupling of programming and execution models, which creates a sticky ecosystem. He uses transportation analogies to categorize these technologies—cars for CPUs, buses for GPUs, and construction vehicles for FPGAs—and discusses how advanced packaging and memory bandwidth are becoming the primary constraints in modern AI hardware.

    Chips, FPGAs & semiconductorsMicrocontrollers & dev platforms
    8 chapters

  • Moore's Lobby · 57m

    Intel VP Aims to Deliver AI Everywhere and to Everyone

    Dr. Wei Li, Intel's Vice President and General Manager for AI and Analytics, discusses the evolution of Intel from a PC hardware company to a software-plus-hardware leader in AI. He details the importance of hardware-software co-design, citing a 200x performance improvement through software optimization and the introduction of DL Boost and VNNI instructions in CPUs. Li also explains Intel's strategy of providing open-source AI reference kits to simplify development for complex workloads like chatbots and quality inspection.

    AI & machine learningEmbedded firmware
    8 chapters

  • The Circuit · 45m

    Episode 32: Nvidia and the AI Raging Bull

    Ben Bajarin and Jay Goldberg analyze Nvidia's recent earnings, noting that the company crushed expectations and raised guidance due to unprecedented demand for AI infrastructure. They debate whether Nvidia's dominance in training and inference will persist, discussing the shift from general-purpose to accelerated compute and the critical role of the CUDA software ecosystem in creating high switching costs. The hosts conclude that while there is hype around peripheral AI stocks, Nvidia remains the central beneficiary of a multi-year hardware buildout cycle, with future value potentially shifting to software applications.

    Chips, FPGAs & semiconductorsConnected products & IoT
    11 chapters

  • The Circuit · 51m

    Episode 70!!! NPUs!!!!!!!!

    Ben Bajarin and Jay Goldberg host Austin Lyons and Paul Karazuba to discuss the architecture, market, and future of Neural Processing Units (NPUs). The group debunks the metric 'TOPS per watt' as meaningless, arguing that real-world performance depends on specific network workloads and latency. They explore how chipmakers like Qualcomm, Intel, and Apple are designing NPUs, often by licensing IP or evolving DSPs, and predict that NPUs will become ubiquitous in devices from smartphones to microcontrollers within three to four years.

    Microcontrollers & dev platformsChips, FPGAs & semiconductors
    8 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Embedded.fm. The episode itself belongs to its publisher.