*speculation*
AMD will do a big Navi with the RDNA2 architecture on 7nm+, it will surpass RTX 2080TI by 20% but nVidia will move to 7nm+ themselves and come right back at AMD with an even faster card.
You know what, that's fine by me, AMD don't have to do to nVdia what they did to Intel, they just have to get back to being competitive.
The frustration is AMD could drop a 5800 XT right now and match, even beat, the 2080 on RDNA
1. TSMC have been producing big 7nm GPU dies for well over a year and Navi is scalable, so it's not like there's any technical issue preventing the bigger cards from landing (power draw maybe?). The 5700 XT is only a 40 CU part and TSMC could easily produce something bigger with, say, 48 or 56 CUs. Hell, Radeon 7 was EOLed to make room for such a card (we all know it was only ever a marketing stunt and stopgap)
But AMD won't make it because they won't shift TSMC's capacity away from EPYC (and now clearly Threadripper). How much ranting, raving and frustration have we seen from members here about the lack of 3900X and 3950X stock? Well given EPYC and Threadripper are plentiful, it's obvious where the best chiplets are being allocated. But we'd get the same levels of frustration - even more so - if the 5800 XT is announced with 56 CUs, wipes the floor with the 2080 Super but Gibbo has to resort to grey market tactics just to stock 10 of them.
So it does look like we're waiting until 7nm+/6nm production ramps up for Zen 3 before we see anything else from Navi/RDNA 1. That does suggest though AMD could stay competitive with Turing with RDNA 1 and keep up the fight with RDNA 2 when Nvidia release their next arch. Nvidia milking Turing and TSMC focussing on CPUs might actually play to AMD's advantage - Nvidia are in no rush to release a 7nm product and the longer they wait the sooner Zen 3 comes out which frees up 7nm resource.
But anyway, back to the paper launch of Intel's latest embarrassment...