Can we finally admit that anything below 28nm is a meme?
Intel have seen zero meaningful performance gains from 32nm Sandy Bridge to 14nm Skylake, which also use just as much power and get just as hot. And shills have been saying for a year that everybody should wait for 14/16nm GPUs, because it's going to be this massive leap forwards and the 900/300/Fury series would be BTFO hard. Yet what we got was underwhelming garbage from both AMD and Nvidia, offering the same performance as previous gen cards, just for $100 less. Wow!
Die shrinks are just a tool that companies use to trick tech-illiterates into buying new technology that's no better than the stuff they already own. Prove me wrong.
>>55598990
Hey, you watch your mouth. The next -gen GPUs did bring something meaningful.
980ti below $400
>>55598990
Pretty much. You won't ever see large performance gains linked to a die shrink ever again.
>>55598990
>Intel have seen zero meaningful performance gains from 32nm Sandy Bridge to 14nm Skylake
A 6700T has about the same performance as a 3820 with a TDP of 35 W vs. 130 W. And that's where the market heads at the moment: Less power draw, not more performance.
>>55599161
This. We reached a point where performance per CPU core started stagnated, but performance per watt kept on moving. Most devs can't into moar cores, so hardware shifted to minimizing cost and power consumption.
>>55598990
The fastest Sandy Bridge chip produced was the Xeon E5-2687W with 8 SNB cores @ 3.1ghz.
The fastest Broadwell chip is the Xeon E5-2699 v4 with 22 Broadwell cores (~22% IPC increase over SNB) @ 2.2ghz.
There's been massive gains in CPU performance, but only if you're willing to spend thousands of dollars on a server chip. IPC gains are still shit, but the overall computational gains have been huge.
If AMD had any legitimate competitive presence above the $120 mark, those server chips would go way down in price (and would probably see releases in Intel's consumer processor line).
