TR Forums

Says Jim Keller: https://www.youtube.com/watch?v=oIG9ztQw2Gc

It's just sleeping?

It's pining for the fjords...

Beautiful plumage on the Norwegian Blue.

We people are living in a fantasy world of "atomic clocks" and stuff, and we think two years have passed. Atomic ... hahaha. Meanwhile, in the real world of Moore's Law, it's only been a couple months!

"Moore's Law" has already been been invalidated. It is just trying to shift goalposts around to keep the meme alive. We are reaching the physical and economical limits of straight miniaturization with semiconductors. I doubt we'll find an alternative that will fare as well.

The game just shifted towards software, architecture and material changes. The returns aren't going to be great as before and it'll cost more to R&D and mass-produce. That'll unfortunately limit its applications.

Oh, I think we'll find something viable to replace silicon eventually. It'll take a while though.

All the big players have something cooked up and ready to launch in their biggest top secret underground lab. Just waiting till it's Time's Up for silicon.

just brew it! wrote:

Oh, I think we'll find something viable to replace silicon eventually. It'll take a while though.

I doubt it'll able to miniaturize as much though. The magical smoke has been used up.

Igor_Kavinski wrote:

All the big players have something cooked up and ready to launch in their biggest top secret underground lab. Just waiting till it's Time's Up for silicon.

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

just brew it! wrote:

Igor_Kavinski wrote:

All the big players have something cooked up and ready to launch in their biggest top secret underground lab. Just waiting till it's Time's Up for silicon.

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

What, you don't think some mad genius has a carbon-nanotube chip fab in his garage?

K-L-Waster wrote:

just brew it! wrote:

Igor_Kavinski wrote:

All the big players have something cooked up and ready to launch in their biggest top secret underground lab. Just waiting till it's Time's Up for silicon.

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

What, you don't think some mad genius has a carbon-nanotube chip fab in his garage?

He very well might. But even if it exists, it ain't gonna scale to the point where it can produce commercial volumes of reasonably priced chips any time soon.

K-L-Waster wrote:

just brew it! wrote:

Igor_Kavinski wrote:

All the big players have something cooked up and ready to launch in their biggest top secret underground lab. Just waiting till it's Time's Up for silicon.

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

What, you don't think some mad genius has a carbon-nanotube chip fab in his garage?

And disguising all his work as refurbing old video arcade games?

Blahpony wrote:

K-L-Waster wrote:

just brew it! wrote:

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

What, you don't think some mad genius has a carbon-nanotube chip fab in his garage?

And disguising all his work as refurbing old video arcade games?

Hmm, sounds like a gerbil we may have read about...

just brew it! wrote:

K-L-Waster wrote:

just brew it! wrote:

I'm sure they're working on stuff... but "ready to launch"? No frikkin' way. It'll take billions of dollars in investment to build a new fab and several years in transition to ramp from laboratory proof-of-concept to volume production. If a replacement for silicon was "ready to launch" we'd know about it, since it's kind of hard to hide a commercial-scale fab based on revolutioonary new tech, not to mention the entire supply chain that would need to be created to supply the equipment and raw materials (whatever they are).

What, you don't think some mad genius has a carbon-nanotube chip fab in his garage?

He very well might. But even if it exists, it ain't gonna scale to the point where it can produce commercial volumes of reasonably priced chips any time soon.

Well, if you're going to insist on bringing reality into the discussion, this just isn't fun anymore...

As we all know, Moore's Law isn't just about physics and engineering -- it's also about economics.

I like the figure/table at the bottom of the page here: https://en.wikichip.org/wiki/technology_node

Now that GloFo is out, we're down to Intel, TSMC, and Samsung. Who will be next to drop out of the race? I'm pretty confident it won't be TSMC. I'm less sure about Intel and Samsung, but I lean towards thinking Samsung will hold out longer than Intel, out of ego if nothing else.

Moore's law has been pining for the fjords in Santa Clara since January of 2011, at the latest.
https://www.technologyreview.com/s/6152 ... oores-law/

In retrospect, AMD's decision to ditch their own fabs was a very good move. Let TSMC/Samsung/etc. figure out the process issues, and have your own people focus on CPU architecture.

The move to MCMs also couldn't have come at a better time, to deal with the inevitable yield issues that come with transitions to smaller process sizes and the slowdown of Moore's law.

just brew it! wrote:

The move to MCMs also couldn't have come at a better time, to deal with the inevitable yield issues that come with transitions to smaller process sizes and the slowdown of Moore's law.

Do you think the smaller process sizes will force the engineers to space the chiplets further apart from each other so they don't melt each other from their heat? That's going to end up increasing latency. Do you think they might shift to multiple sockets to solve this issue?

Igor_Kavinski wrote:

Do you think the smaller process sizes will force the engineers to space the chiplets further apart from each other so they don't melt each other from their heat? That's going to end up increasing latency. Do you think they might shift to multiple sockets to solve this issue?

I think it is more likely that it will result in some changes to how we cool chips. Maybe coolant channels embedded in the heatspreader or within the chip package itself. Or perhaps a return to bare-die CPU packages (like we had with Socket A), to eliminate the extra thermal interface.

I'm with JBI. Cooling integrated at the chip design level is where we're headed.

Captain Ned wrote:

I'm with JBI. Cooling integrated at the chip design level is where we're headed.

I've seen it demoed more than a few times in the past decade or so. How to manage erosion, on the other hand, seems to be the difficult part.

just brew it! wrote:

In retrospect, AMD's decision to ditch their own fabs was a very good move. Let TSMC/Samsung/etc. figure out the process issues, and have your own people focus on CPU architecture.

Decision? My recollection is they had no choice. They overpaid for ATI by several billion dollars, made bad design choices for their CPUs, and HAD to sell off their manufacturing assets to avoid bankruptcy.

It's kind of like giving someone credit for being hit by moped because it prevented them from being hit by a bus later.