Yeah, I know it's too soon for Navi 23, I was just teasing.
But eventually, it will (hope at least) make it to a new MPX module to replace good ol' Vega.
RT seems to be a given, let's hope the info is correct.
I think you skipped over most of what I wrote. Here is some AMD specific context. ( over a year old so you need to 'adjust' what "next year"
"... the
AMD Navi release next year. “We are looking at the MCM type of approach,” says Wang, “but we’ve yet to conclude that this is something that can be used for traditional gaming graphics type of application.” ...
....
....
It seems, however, that the MCM approach is only an issue in the gaming world, with the professional space more accepting of multi-GPU and potential MCM designs.
“That’s gaming” AMD’s Scott Herkelman tells us. “In professional and Instinct workloads multi-GPU is considerably different, we are all in on that side. ....
"
https://www.pcgamesn.com/amd-navi-monolithic-gpu-design
AMD slapping some "Nvidia killer" on something along with a raytrace hardware is probably the 'gamer' monolithic chip being pitched here as Navi 23 GPU. That huge gamer targeted chip isn't pragmatically the follow on to Vega. Or at least Vega's sweet spot.
I think Apple is going to get Navi with some other number. ( or perhaps another name for the separate computation/workstation track or perhaps simply just even numbers; 22 , 24, etc. ). The MPX Duo is pretty close to a multi-chip module now. If AMD puts most of it on a bigger package and cuts down the design work for Apple ... all the more so Apple will probably track that way.
Leaving the ray trace hardware off would make it easier to put two relatively smaller dies into the same package. Trying to stuff everything under the Sun into one die is going to make it big which is fine if need all that for the same shared output ( e..g,. content for frame buffers at high frame rate.). Navi23 could basically be Navi21 with more stuff (ray trace + etc. ) piled on 'top' that just expands out to a bigger die.
Kind of what Nvidia has done at more than a couple of process nodes is just roll out a much bigger die than AMD.
I don't think Metal is going to rapidly track some proprietary ray trace hardware. Nor is that huge spend on transistors going to make FPX or Logic (or a wide array of audio video processing apps/workloads ) go any faster. So I doubt it will be high on their "gotta have" list. I might get some drivers for 3rd party cards but that doesn't seem like a primary target for the niche Apple is shooting for.
Now, it makes me wonder if I should go with the 580 for now and next year, when the Navi MPX module is available, upgrade with one of those.
Have to wait on the specs for the Navi 23 GPU I guess.
I suspect what will happen in 2020 is that the Navi 10 will trickle down to the entry MPX half width options. ( Not the navi23.). if anything "new" perhaps a lower end Navi 21 is the thermals and price are better for a half width module.
"Vega 20" ( the foudation for Vega II) I suspect will be something other than Navi 23 ( if that is big Navi with gobs of transistors thrown at ray tracing. ) because it is on the 'wrong' track than what Apple will be following for the large MPX modules.
AMD has been trying to follow both gamer and GPGPU with the same arch that was suppose to be "Jack of all Trades" . If they are trying to build an Nvidia 'killer' they can't do that. Primiarily because Nvifia is already following the two track arch. AMD would have to pick one of those. and 'attack'. Attacking both with just one instance isn't going 'defeat' either competitor instance. AMD is less poor at this point so they'll need new tactics. Intel is jumping into the market also. ( so just undercutting Nvidia prices is going to have less traction also. Intel is likely going to be "down there" also. )
