somebodyspecial
Honorable
The article forgets to mention they've been restructuring now for 5 years, and have previously laid off 30% of their workforce over said time. So not exactly a drop in the bucket when you consider they KEEP doing it repeatedly. I really hope ZEN is a hit and they chose a DIE size that is equal to Intel's whole cpu/gpu die, RATHER than just matching Intel's cpu side. The whole size would mean victory for a while and pricing power perhaps for a few years, while matching the cpu side just means Intel will price it to death until they come up with something faster again. If ZEN's die is huge (~1/2 of Intel's current dies are GPU), Intel's only choice would be to go 8 core. In a world where 4 faster cores are better than 8 (in most cases, except stuff like ripping etc) for most apps/games, they will end up like AMDs 8 core vs. Intel's 4 right now. That would make them very profitable for a few years as Intel would end up 125w etc vs. AMD's 85 or so (like the reverse is true now). IE, when my Devils canyon is ripping it's hitting 60-68w (gpu off), so double that if Intel's response is two cpus.
If AMD went small or barely bigger than Intel's cpu side, Intel can still cherry pick (ala devils canyon but a bit better of course) and still price them down. AMD needs a large die to come back big. Anything less is failure. Saving a few dollars per die and losing is nothing compared to winning and charging $350-400 instead of $160-200 and being slower in most crap. You need to win, and do it DECISIVELY. I would gladly pay $400 for a die size that is the total size of skylake (or whatever Intel is running then) but far faster at like watts. Lopping off the gpu portion and a die shrink + Finfet should allow AMD to win in EVERYTHING if they chose wisely. If not, I guess I'll be buying Intel again for my dad's PC. I couldn't put off my own pc any longer, too much ripping, but hope to buy AMD this time. Last time they had a king for 3yrs they couldn't produce more than ~20% of the market. That would not be the case this time, as AMD could have multiple people producing chips (GF, TSMC, possibly samsung at some point since they share tech with GF). This is the perfect time to strike as Intel has been concentrating on battling ARM and likely has no HUGE die chip they've been designing (quad I mean) as a backup plan if AMD pulls this. It would take a few years no matter how much money Intel threw at the problem to fix this. I really hope this die is >200mm^2, heck I hope it's near 250 Go big or go home. Intel's dual core broadwell is ~133 IIRC, so ~200+ sounds good. The 9590 is ~320 IIRC, so they've been larger before (yeah different process, but you get the point, go big!).
If AMD went small or barely bigger than Intel's cpu side, Intel can still cherry pick (ala devils canyon but a bit better of course) and still price them down. AMD needs a large die to come back big. Anything less is failure. Saving a few dollars per die and losing is nothing compared to winning and charging $350-400 instead of $160-200 and being slower in most crap. You need to win, and do it DECISIVELY. I would gladly pay $400 for a die size that is the total size of skylake (or whatever Intel is running then) but far faster at like watts. Lopping off the gpu portion and a die shrink + Finfet should allow AMD to win in EVERYTHING if they chose wisely. If not, I guess I'll be buying Intel again for my dad's PC. I couldn't put off my own pc any longer, too much ripping, but hope to buy AMD this time. Last time they had a king for 3yrs they couldn't produce more than ~20% of the market. That would not be the case this time, as AMD could have multiple people producing chips (GF, TSMC, possibly samsung at some point since they share tech with GF). This is the perfect time to strike as Intel has been concentrating on battling ARM and likely has no HUGE die chip they've been designing (quad I mean) as a backup plan if AMD pulls this. It would take a few years no matter how much money Intel threw at the problem to fix this. I really hope this die is >200mm^2, heck I hope it's near 250 Go big or go home. Intel's dual core broadwell is ~133 IIRC, so ~200+ sounds good. The 9590 is ~320 IIRC, so they've been larger before (yeah different process, but you get the point, go big!).