So TSMC's 5nm isn't going into production next month then? I guess i missed that bombshell, Apple will be most unhappy. Zen 4 (possibly a Zen 3+) and RDNA 3 are on 5nm next year (ish). So yes, there is another die shrink available to AMD. You also seem to glance over the fact that AMD are releasing RDNA 2, which with all the improvements is as much of a new architecture as Ampere is to Turing.
So both companies are dropping new arches on new nodes at the same time. Let the pure numbers talk, the playing field is level.
As for Nvidia, arguably they're the ones without the shrink available. They tried to play TSMC and failed so most of Ampere is pegged for Samsung 8nm, which is a smidgen behind TSMC's standard 7nm, let alone EUV, and with AMD taking up most of TSMCs 7nm capacity, only Nvidia's super top end halo cards (3090/Titan) will be produced on 7nm. So they can't move to Samsung 7nm because it's woeful and delayed, TSMC's 7nm production in all forms is full, and Nvidia won't get a look in on TSMC 5nm for years.
I also have to chuckle at your waxing lyrical about Nvidia have dedicated ray tracing hardware like it's somehow any good. It's not. We've had 2 years of seeing just how woeful Turing's beta test actually is. So I would fully expect the RDNA 2 consoles to outperform 2080 Ti in ray tracing because the 2080 Ti is just rubbish in the first place. Consoles will be nowhere near Ampere's ray tracing if the latest information holds true.
What are you talking about? This is another example of you jumping into a conversation to cause an argument with me, to try and put me down without actually understanding the conversation at all.
AMD are on 7nm now, Nvidia are on 12nm now. AMD's RDNA 2 will be on a slightly improved 7nm node. It's not a die shrink. Don't you get that? Most of the performance increase is going to be from the new architecture. That's facts.
Now Nvidia, their cards are on 12nm. We have no solid info yet about what manufacturing process they are going to be using. The only solid facts we have is that Nvidia used TSMC's 7nm for their GA100 card. They might be using Samsung 8nm or TSMC 7nm, it's all rumours so far. More than likely the GPU production will be split between the two and that's the reason for the different rumours. In either Case Nvidia's ampere cards will be on a die shrink.
Are you still following? Quick summary.
AMD - No die shrink but new Architecture.
Nvidia- Die Shrink + New Architecture.
Die shrink always brings better performance than new architecture. Just look through your GPU history.
Your second paragraph, I shouldn't really comment on it. Sounds like you just making stuff again to be honest. Nvidia have pre-booked TSMC 5nm for Hopper, that was reported a couple of months ago.
Waxing lyrical about Nvidia's dedicated hardware solution for Ray Tracing? Can't say that I have. Just talking facts. Hardware dedicated to a certain function in most cases performs better than a hybrid solution. I don't think you understand how much processing power Ray Tracing needs because you guys keep saying it's terrible. I agree the price was terrible but how can you say their solution was woeful? Ray Tracing is extremely demanding, it's only now that GPUs can even attempt it. AMD's solution looks interesting from the Patents and will probably be the best solution down the line when GPUs are powerful enough to have Ray Tracing and Rasterisation using the same resources.
How powerful do you think the new consoles are going to be? You think they are going be more powerful than the 2080Ti? And I don't mean in Ray Traced performance, just ordinary raster performance.
