Computer Architecture Podcast cover image

Ep 2: Domain-specific Accelerators with Dr. Bill Dally, Nvidia

Computer Architecture Podcast

00:00

The Cost Model for Channels Offship

The problems are the same. When you are constructing a network, you af topology, you have routing outim yo, ha, full control. All those things are the same, a on chip and off chip. Now, the answers are often differet. The reason why the answers are different is tha cost models are very so on chip, you're signalling with semos logic. And it actualy es a cost model where the ene g is proportional to distance. So the longer a channel is, the more energy you consume on that channel. Once you go off chip, you pay a big up front cost for asertes,. For io driver on the chip to drive a

Play episode from 15:17
Transcript

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app