Thread replies: 9
Thread images: 2
Thread images: 2
File: Northwood_130nm_die_text_1600x1200.jpg (761KB, 1600x1200px) Image search:
[Google]
761KB, 1600x1200px
How high can the clock speeds on consumer level processors get before desync becomes an issue? Ie: what's the maximum distance a signal has to travel inside a processor in a single clock cycle? I know that light itself in a vacuum travels about a foot in one billionth of a second (one clock cycle at 1ghz). So at 4ghz, light can travel 3 inches, and considering that electricity in a conductor doesn't travel as fast as light in a vacuum, it seems to me, being not very knowledgeable on this, that we're pretty close to desync becoming an issue on the highest clocked processors.
I'm just curious because quantum tunneling is going to be an issue in a few node shrinks, and we might not have a commercially viable alternative to silicon by the time we've reached the limits of silicon, so it seems natural to me that higher clock speeds would be how we squeeze every bit of performance out of silicon before we're done with it.
>>
Interesting question, but before electron speed becomes an issue we will be using photons instead
This already sort of exists but the tech still needs to advance to be miniaturized and mass-producable, look up optical processors
>>
The section of the CPU die that runs at the frequency you're talking about is unbelievably small, 3 inches is a pretty ridiculous number to pull out of your ass
But yes the limit isn't far, that's why multicore will be important and why die shrinks are still relevant
There's a reason why we're not using 8GHz celerons though, speed isn't the only way to squeeze performance out of silicon
>>
>>60092004
Obviously speed isn't the only way to increase performance, but it is a way to increase performance when traditional methods of doing so are becoming harder and we don't have a commercially viable alternative yet.
>>
>>60092050
It's not like people haven't seen this coming, I'm sure a few companies have been working on non-silicon for a while at this point, probably just waiting for the economically best time to put it on market
Or you know, we get the electronic stability and heat under control and we can keep pushing moar cores as long as efficiency keeps going up and die size keeps going down, that'll extend traditional CPU tech life for a while
>>
>>60092090
Honestly if CPU tech cant start moving towards a progressive direction soon enough, I could imagine igpu monsters or special hardware encoders being the next big thing instead of performance
>>
>>60092110
Kinda like how Intel advertised Kaby Lake as "LOOK AT THIS CPU RUN 4K YOUTUBE OMFGGGHHG!"
>>
File: 1489631912311.png (171KB, 765x624px) Image search:
[Google]
171KB, 765x624px
>>60092110
So basically a bunch of ASIC chips for their own particular purpose on a die instead of a main core?
>>
Clockspeed has bigger issues before signal clarity, degradation and synchronization come into play.
It's voltage.
Silicon doesn't respond to with kind voltages at high frequencies, that's the biggest problem.
Thread posts: 9
Thread images: 2
Thread images: 2