Hardware Podcasts

Heterogeneous Computing with Raja Koduri

Oxide and Friends · 18 Jul 2024 · 1h 54m · Episode 18

What this episode covers

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.

Most relevant to

large enterprises

Topics

Mentioned

Companies
3dfx, Intel, AMD, NVIDIA
Products
Direct3D, Glide, CUDA
Chips
HBM
Also
Python

Related episodes from other shows

Other shows on the same ground, closest first.

  • Moore's Lobby · 43m

    Ep. 29 | NVIDIA CTO Michael Kagan on the New Age of AI and Supercomputers

    Michael Kagan, CTO of NVIDIA, discusses the evolution of high-performance computing, tracing his career from Intel and the 860 processor to co-founding Mellanox and its eventual acquisition by NVIDIA. He explains how NVIDIA has shifted from a graphics company to an accelerated computing platform, integrating GPUs, CPUs, and DPUs to support AI workloads. Kagan emphasizes that modern data centers function as unified computing units, delivering processing as a service rather than standalone hardware, driven by the massive computational demands of training neural networks.

    Chips, FPGAs & semiconductorsConnected products & IoT
    8 chapters

  • EE Times Current · 34m

    Whither Semiconductors? The Wheels on the Bus Go ‘Round and ‘Round

    Jim McGregor, co-founder of Tirias Research, discusses the evolving competitive landscape of the semiconductor industry, identifying Intel, AMD, and NVIDIA (following its potential acquisition of Arm) as the new dominant trio. He explains how the convergence of computing, connectivity, and intelligence is blurring traditional boundaries, with companies like Qualcomm, Samsung, and MediaTek competing across smartphones, servers, and automotive sectors. McGregor highlights the shift toward heterogeneous computing, specialized accelerators like DPUs, and the move away from monolithic dies toward multi-chip modules and advanced packaging to overcome Moore's Law limitations.

    Chips, FPGAs & semiconductorsMechanical engineering
    8 chapters

  • The Circuit · 41m

    Episode 105: Sanity Check on Custom ASICS?

    Ben Bajarin and Jay Goldberg discuss the debate between custom ASICs and GPUs in AI compute, arguing that GPUs retain significant value due to software flexibility and installed base. They analyze the uneven maturity of hyperscaler silicon efforts, noting Google's TPU success versus the struggles of Amazon's Trainium and Meta's MTIA. The conversation then shifts to analog semiconductors, highlighting their critical role in humanoid robotics and data center power management, with specific mentions of Infineon's inventory buildup linked to NVIDIA's Blackwell.

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

  • 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

  • Moore's Lobby · 51m

    Ep. 43 | Mark Papermaster: The EE and Exec Behind Apple, IBM, Cisco, and AMD's Success

    Mark Papermaster, AMD's CTO and Executive Vice President of Technology and Engineering, discusses his career at Apple, IBM, and Cisco, and his current role leading AMD's turnaround. He explains how AMD used modularity and the Infinity Architecture to 'punch above its weight' against larger competitors, enabling the development of the Zen architecture and APU. Papermaster details the strategic shift toward chiplets and heterogeneous computing through acquisitions of ATI and Xilinx, allowing AMD to mix CPU, GPU, and FPGA technologies. He emphasizes the importance of multidisciplinary engineering teams, balanced system design, and AI as a fundamental skill for modern engineers.

    Chips, FPGAs & semiconductorsCareers & education
    10 chapters

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