Thermal Challenges And Moore's Law

[@laynemccormic9102]

Excellent video very interesting

[@geonerd]

Thanks, Ed. This series is most wonderful! :)

[@kevinbauce7972]

Great video, the audio was a bit too low IMO though. I have a question. When do you think graphics cards will obligatory be watercooled for customers ? The cooling solutions have become more and more complex and sophisticated but a large heatsink with big fans and heat pipes has some downsides (limitation for the heat dissipated, weight, size...)

[@dominikliberda4017]

I don´t think water cooling will be obligatory anytime soon, cause there are also many downsides to it: Pump endurance Leakege Liquid evaporation Various corosion risks Air cooler will pretty much last forever (maximaly you replace broken fan. There´s a risk of evaporation of the liquid that is inside heatpipes, but that takes really long). But with water cooling there are many things that can go wrong and causes various reliability issues even after few years. Water cooling will become obligatory once (if ever) we find out a way to get the cooling liquid flowing literaly inside the chip through channels under transistors - I believe water cooling directly inside the chip has a big benefit in decreased differences in temperatures across the chip. Today even if a chip is running at comfortable 60°C, you might have a hotspot with 80°C. And this delta increases as you advance to smaller nodes. New AMD GPUs report GPU Edge temperature (temperature at the edge of the die) and Junction temperature (highest temp across all sensors, which might not be *the highest* temperature on the die but it comes close to it) and there are big differences between these two. Water cooling channels inside the chip would really get rid of these hotspots in my opinion.

[@fleXcope]

Good explanation

[@FullFledged2010]

No no. That asus 5950 ultra you're holding is from 2003 not 2000/2001 😉