>CPUs plateaued after SandyBridge! Is this a meme or is there

File: 1487013222251.png (79KB, 556x454px) Image search: [Google]

79KB, 556x454px

it's when diminishing returns really started yeah, but "progress" in general really started slowing around the mid 2000s on the hardware level and practical gains from each new generation is even lower because it turns out that the same old algorithms we perfected decades ago are agnostic to the passage of time and don't get any hungrier for system resources as they get older... really depolarizes the synapses

>>59024413
Pretty much, yeah. If you're not playing gaymes you will hardly tell the difference between a CPU made in 2010 and one made in 2017. You can even go as far back as 2007 if you use top of the line CPUs from that time.

>>59024413
There is maybe a ~30-35% improvement in performance between a Nehalem and a Kaby Lake CPU with the same clock rate. While performance has stagnated, Moore's Law hasn't stopped since technically they're fitting more transistors per unit of space (by making the die smaller) which gives very good improvements in power consumption.

But yes, the days of Moore's law being an indicator of performance growth is over. Silicon will reach its physical limits in the next 5 years and there's nothing concrete planned after that. It's just gonna be moar cores.

>>59026810
Silicon-Germanium might become widely adopted if they are able to keep the costs down.

File: 1452616536122.gif (2MB, 500x500px) Image search: [Google]

2MB, 500x500px

>>59024413
It's the famed frequency wall,

A hurdle Computer Engineers saw coming with the FET structure.

There are 2 physical barriers forseen
1)The Frequency wall: electrons can only go so fast
2)The Silicon Wall: FETs can only get so small

we reached the first part...hence why we haven't seen speeds above 4Ghz and started multi-coring...

After this will be another hurdle of FETS not getting smaller

>>59028517
Nah we didn't hit the electron wall. The problem is making gates switch faster requires higher voltage gradients, which increases power usage, which increases heat output.