>>53708293
People expecting massive performance increase aren't paying attention to state of silicon, Intel's Haswell to Skylake performance increase should have been a warning flag. Node shrinks are getting diminishing returns

>>53708293
>not getting a 9500 and modding it to a 9700

>>53708483
>not getting nvidia

>>53708498
>getting an nvidia FX
nice meme, too bad their DX9 performance sucks

>>53708483
>not getting the 9800se and unlocking the 4 shader pipelines using softmod drivers

>>53708293
Performance per watt and performance per clock (IPC) are integrally related, champ.

>>53708293
How has the dust settled?

There's a whole bunch of rumors about polaris, bullshit on vega and nothing on pascal.

All we know is both AMD and Nvidia have been talking 2x the performance/watt, and there's no reason to assume they're going to be halving TDPs.

>>53708447
Except we're not talking going down a tiny step here, we're going all the way from 28nm down to 14/16 which is a massive jump all by itself combined with being a new generation designed with modern features and on top of all of that we're getting GDDR5X or HBM2.

>>53708571
> 28nm down to 14/16 which is a massive jump
Consider that 32nm Sandy Bridge to Skylake was still a shit increase, even with multiple shrinks and new architectures.

>on top of all of that we're getting GDDR5X or HBM2.
First Pascal and Polaris GPUs coming out Spring will have the old GDDR5.

GDDR5X isn't being mass produced until summer, its sampling began only this week. First GDDR5X models can come out late 2016. HBM2 samples won't be out until 2017, mass production is unknown.

>>53708571
14/16 both use a 20nm BEOL so the penultimate electromech characteristics are more similar to somewhere between 20 and 15 nomenclature.

I dont think we will see a true "xx nm" whatever it may be from anyone but Intel until other companies can go beyond tri-gate designs.