02-19-2015, 01:16 AM
so broadwell just launched and intel is gonna launch skylake just 6 months later?
W-H-A-T-?
that is completely crazy man,
craziest thing i ever seen.
What is going on? Is broadwell a top to bottom series? will they finally bring 6 or 8 core CPUs to the mainstream?
Is this 7% performance increase a clock per clock advantage or is it a result of the smaller node allowing a higher clock speed?
How much has power consumption improved? Its kind of scary/crazy to see that not even intel is able to gain much in performance per watt with their advance fabs and word-class nodes. They have been barely budging forward in real world power consumption and since then has invented new power rating called SDP which they tried to be slick about.
But considering that massive improvements from 45nm to 32nm, we really didnt see too much going from 32 to 22nm. And now 14nm, which should be a massive improvement and it is not. Especially in the high performance sector. It is gonna take a major change in the core architecture. This is where intel has got the small gains it did get. Going from sandy to ivy to haswell, all the improvements came from the architecture and very little from the node. Very very little.
But this means that GPUs are gonna be in the exact same boat. We already see that 20nm is not even worth it and it is being skipped. But, i am starting to think that this is a bigger deal than just this one node. Moving forward is gonna be very very tricky for GPUs. I think the LP is doing okay and the smaller nodes do allow for more transistors in the same area. But we are running towards a narrowing tunnel. Perhaps GPUs will have to tone it down in the future. It is sad to think, but GPUs are facing the same fate as CPUs and will really start creeping forward. Until there is some break through.....somewhere
W-H-A-T-?
that is completely crazy man,
craziest thing i ever seen.
What is going on? Is broadwell a top to bottom series? will they finally bring 6 or 8 core CPUs to the mainstream?
Is this 7% performance increase a clock per clock advantage or is it a result of the smaller node allowing a higher clock speed?
How much has power consumption improved? Its kind of scary/crazy to see that not even intel is able to gain much in performance per watt with their advance fabs and word-class nodes. They have been barely budging forward in real world power consumption and since then has invented new power rating called SDP which they tried to be slick about.
But considering that massive improvements from 45nm to 32nm, we really didnt see too much going from 32 to 22nm. And now 14nm, which should be a massive improvement and it is not. Especially in the high performance sector. It is gonna take a major change in the core architecture. This is where intel has got the small gains it did get. Going from sandy to ivy to haswell, all the improvements came from the architecture and very little from the node. Very very little.
But this means that GPUs are gonna be in the exact same boat. We already see that 20nm is not even worth it and it is being skipped. But, i am starting to think that this is a bigger deal than just this one node. Moving forward is gonna be very very tricky for GPUs. I think the LP is doing okay and the smaller nodes do allow for more transistors in the same area. But we are running towards a narrowing tunnel. Perhaps GPUs will have to tone it down in the future. It is sad to think, but GPUs are facing the same fate as CPUs and will really start creeping forward. Until there is some break through.....somewhere