fatslob-:O said:
It's still not in production yet ... (And it won't be up until 2020/2021) |
That doesn't mean they haven't invested billions into that fab already... They halted it's construction remember.
| fatslob-:O said: It wasn't big at the time for desktop users but it was a big thing in servers, it was the biggest contributor to how AMD destroyed the monopoly Intel had on x86 servers at the time ... |
Can't argue with that.
But server isn't exactly stupidly high volume like other market segments. It does have higher profit margins though.
| fatslob-:O said: Die size says otherwise ... (GP102 at 471mm^2 vs Vega 10 at 484mm^2) |
We both know how behind AMD is on GPU efficiency right now.
Their chips are slower, consume more energy and are large.
Vega was AMD's catch-up part. Polaris was a "stop gap" measure.
The Radeon Group got made into it's own seperate group, there was allot of distruption. Hopefully after Navi AMD switches things into a more competitive gear.
****
Also something interesting about Vega is that, with it's clocks normalized to Fury... It doesn't seem to have gained much in terms of efficiency.
You would expect there to be some big efficiency gains due to the draw-stream binning rasterizer, improved colour compression, caches, improved geometry, ROPS. etc'.
So either AMD hasn't enabled the technology in it's drivers already, or Vega has some bottlenecks at play.
http://www.gamersnexus.net/guides/2977-vega-fe-vs-fury-x-at-same-clocks-ipc
I would have liked for them to included Frostbite so we can have a taste of an extremely popular deferred renderer. But it's still some interesting data points.
www.youtube.com/@Pemalite
