AMD don't need a fundamental redesign in architecture. They are immune to Meltdown precisely because they do things the correct way in that it costs more performance to run a permission check before rather than after the crucial point that lets Intel chips be vulnerable. Same goes for variant 2 of Spectre, AMDs fundamental design makes them effectively immune to it. Intel needs a fairly large change to fix both these issues and is very vulnerable to both due to architecture. Variant 1 is a little unclear, AMD only seem to be effected in one situation with an incredibly easy fix.
Also it's highly unlikely Intel will fix these in their next generation chips, it takes 18 months to tape out a cpu once design is finalised, meaning a chip due to come out a year from today required finalised design 6 months ago and it will likely take some time to fix. I think the concept for the fixes will be quite simple, the problem is everything is set to work in a specific order in a specific way so even a small change that they know how to make for the fix will likely cause lots of other smaller things needing to be worked out.
That will be the really big pain for Intel, AMD will presumably use the '12nm' process for EPYC as well and will be churning them out as fast as possible, making them slightly smaller, potentially slightly faster, slightly better power/performance ratio and that might be from say mid next year for Epyc. Up margins a little as demand is likely very high. Then you have Zen 2 at 7nm potentially early in 2019, before Intel has a chance to launch a fixed architecture AMD will seemingly have 48 cores per EPYC chip, higher clock speeds, higher single thread performance, likely higher memory support and higher infinity fabric speed all leading to improved performance and much improved core density in a rack. It wouldn't be entirely surprising if AMD are planning to move towards 4 socket systems with Zen 2, its a lot of work to validate so makes sense AMD skipped it this gen but as demand increases and maybe vastly more than they thought before this, then 4 socket could be extremely lucrative and a potentially easy jump into high market share due to the security issues. IE if and when they become available companies might be desperate for them. Without the security issue AMD would be launching 4 socket systems into a market 100% dominated by Intel so less potential for the payback. Now they could potentially launch 4 socket systems up the margins considerably and still have companies desperate for them.
For AMD if Intel doesn't have this fixed by Zen 2, it will likely be a genuinely massive massive year for AMD in 2019.
. Meltdown is a big explosion that's almost completely in Intel-land while Spectre is a smouldering fire almost everywhere.
