It looks like they are going to have stacked memory on the chiplets. That would explain the 32MB+ L3.
What if ray tracing is a separate card running PCIE4.0?
Tensor Core Equivalent?
Vega 20 GPU patents?
Apparently AMD already has 7nm Vega in their lab.
I think AMD is working on implementing their exascale apu architecture with vega 20 as outlined by AMD Research in their paper.
It looks like...
Perhaps mGPU is more for Ravenridge CPUs with Vega GPUs.
That would be for next year.
Rumor is that GT200b is cancelled.
A better solution would be to have one card start at the top and one at the bottom and work your way to the middle.
Radeon HD 4870X2 in the nude, 15% better than 4870 CrossFire
Thats what I'm talking about.
- lower cost
- higher yields
- better performance
If they can share the same framebuffer then there goes your arguement.
Perhaps with an interconnect they can treat two GPUs as one from a driver perspective.
As we will soon see multiple chips will eliminate the need for 512-bit memory...
I think the real efficiency gains will occur with the 4870X2.