The Future of Compute: Ampere Computing‘s CPO, Jeff Wittich, On The Data Center Chip Industry

1:05:56
 
مشاركة
 

Manage episode 306137709 series 2543942
بواسطة Shortman Studios، اكتشفه Player FM ومجتمعنا ـ حقوق الطبع والنشر مملوكة للناشر وليس لـPlayer FM، والصوت يبث مباشرة من خوادمه. اضغط زر الاشتراك لمتابعة التحديثات في Player FM، أو ألصق رابط التغذية الراجعة في أي تطبيق بودكاست آخر.

The semiconductor industry is in a period of transition. Supply chain problems and questions over whether we are now in a secular growth environment; changing leadership as Intel loses ground and Taiwan Semiconductor, Nvidia, and even a new generation of start-ups stake out a claim; and the new demands posed by Artificial Intelligence and its burgeoning compute needs.

We're rolling out a little Future of Compute series to cover this. We speak with several executives and experts in the field to hear what the state of semiconductors, technology usage, and artificial intelligence from the hardware and software side looks like.

We kick off with Jeff Wittich, Chief Product Officer at Ampere Computing. Wittich, like several of his Ampere colleagues including CEO/founder Renee James, is an Intel veteran. Ampere’s aim is to develop server chips designed explicitly for cloud usage, using an ARM chip framework, with the target of delivering much greater power efficiency. They seem to be gaining traction, with the most recent evidence being reports SoftBank is considering an investment in Ampere at an $8B valuation.

We speak with Jeff about Ampere’s journey, about why now is the time for Arm-based chips in servers, about how hyperscalers shape the industry’s demands, the state of semiconductors, and of course a bit on Intel and its challenges.

Topics Covered
  • 4:00 – Ampere’s story
  • 6:00 – What does a cloud focus mean for a chip maker?
  • 11:30 – ARM’s experience in the data center world
  • 16:45 – Why now for ARM-based server chips?
  • 19:30 – TSM’s passing Intel and Intel losing its data center advantage
  • 24:30 – The role of the hyperscalers as pace setters for cloud hardware
  • 28:30 – Can Intel hold onto a shrinking datacenter TAM?
  • 30:30 – The inflection point in the competitive landscape
  • 35:00 – The in-house vs. outsourcing question for AI companies
  • 41:00 – The inference vs. training distinction and the role of the CPU
  • 47:00 - Optimizing for AI workloads
  • 50:30 – How is Ampere lasting when other companies quit
  • 55:30 – Supply chain outlook
  • 58:30 – Risk of a cyclical downturn?
  • 1:01:30 – Lightning round and edge vs. cloud

102 حلقات