Fab/die size and heat

[MadAd]

Weve seen the heat spreaders coming out, and thats because of the newest die size becoming too small to easily work with. So the HSp is made to the size of the old dies.

But if u stare long enough at the base of a Heat Sink, and your P3/Tbird core (ommmm anyone?) you might get the silly idea that as far as heat goes, its a huge bottleneck.

Why arent cores being made the size of the socket? Either by heatspreaders, or by design. (like instead of a cube block, make 4 slices, and put them together in 2d- voila, a flat chip)

And has anyone experimented with (perhaps) silver caps for the core, kinda like a DIY(Home improvements) way of making your own heatspreaders?

And if we had flat head cores up to 4x size, how much more cooling would that allow- not 4x of course- but roughly what would be gained?


Thanks everyone - just another thought that seems to defy normative opinion.

 

[rimshaker]

Financial and performance factors.

Bigger transistors require more power, and much slower because of larger gate lengths. You'd NEVER get to 1GHz if you enlarged the core to the socket size.

A fabrication facility costs several billion dollars these days. It costs the same amount of $$$ to produce 10 cores or 1000 cores on a single wafer. If the cores were enlarged, can we say..... bankrupt?

 

[CTho9305]

<< Financial and performance factors.

Bigger transistors require more power, and much slower because of larger gate lengths. You'd NEVER get to 1GHz if you enlarged the core to the socket size.

A fabrication facility costs several billion dollars these days. It costs the same amount of $$$ to produce 10 cores or 1000 cores on a single wafer. If the cores were enlarged, can we say..... bankrupt? >>



I would say performance more than finance. Consider the P4... TWO of its pipeline stages exist solely to get something across the die. If you had a core the size of the socket, you'd most likely need even more, resulting in a major performance hit.