I think there's a chance they can beat the 1080 Ti (at least in DX12) but the issue is that they have a massive ~520 mm^2 die and to only beat GP102 by a small amount makes it trivial for Nvidia to release an even larger die if big Volta won't be out soon enough.
Well, unless they have their whole design totally modular and macro'ed, I don't see it being easy (read fast) to make a new design between GP102 and GP100. The industry may be heading that way (Intel's Embedded Multi-Die Interconnect Bridge (EMIB), the hints Raja dropped about (presumably) Navi being modular, even the Infinity Fabric design with Ryzen might allow AMD to swap out one CCX for a Vega CCX, and similar things), it's not here yet. So while GP102 @ 478mm² and assuming the max size is 600mm², it's not as simple as taking GP102 and adding another 120mm² of 'stuff' to it. The design, verification and tape-out would take months (maybe 1.5 years?).
Since Freesync is at the scaler level, isn't it possible that a Gsync monitor could support Freesync too? That is without too much effort? Obviously, Nvidia may not certify such a monitor but might it just work or maybe there are even firmware hacks for monitors?Seems no one is talking about Freesync/Gsync.
With Freesync and Gsync monitors out now, some of us are vendor locked. Whether you like it or not, Freesync is a VENDOR LOCK. End of STORY. So I have a Freesync monitor, and hte only high end GPU I can get is Vega. So I'm getting Vega. There are a lot of markets for Vega. Not DOA at all.
EDIT: (Obviously that doesn't help those who have a Freesync monitor, but it might make some future Gsync monitors not actually have vendor lock-in. Also, looks like Scorpio supports Freesync (but PS4 Pro doesn't, strange), so Nvidia may eventually be forced to support Freesync too.)
Last edited: