I'm pretty sure he was including architectural differences. I trust him honestly. Yes, GPU progress has slowed but by too much.
Would a RDNA 1/2 combo still be possible you think?Some news for the most optimistic members in this thread:
RDNA2 is not a topic for next gen. Why I say that because the most recent AMD Corporate Slide Deck talks about 7nm+ and RDNA2 as "In Design" and not as "Design complete". Time now would be to short for expecting this new technology in next-gen consoles.
In the grand scheme of things, Xbox One X and PS4 Pro are selling what, 20% of total Xbox and Ps4s sold? That's like 20Mish total combined when this gen is over, maybe 25M, compared to a total of 150M+ combined PS4/Xbox sales, I think 80% of most consumers would not even notice if they were only 8-9TFs
This "engine" is dedicated RT h/w. You may call it a "core" or not, it doesn't matter in the slightest.
Also there are no discernible "RT cores" on any die shot of any Turing GPU so it's anyone's guess where they are located. But the fact is that the number of "RT cores" in 20 series cards is linked to the number of active SM units so it's highly likely that NV's "RT core" is a part of Turing's SM (CU).
What shots I've seen don't show much difference between these. The SMs look kinda similar and there's nothing apparently present on a TU10x chip which is absent from a TU11x one. NV's RT cores are likely just a functional unit inside Turing's SM, they are no more "dedicated" than any other unit inside an SM or CU.
That wouldn't surprise me. People are expecting these big cores when it reality it's a unit that generates inputs for the other ALU hardware already there.
Because the new designs were on the same process node. RDNA2 and RDNA are on different process nodes. But we will see ... I am little bit more sceptic here .... I also believe that the RT implementation will be different for MS and Sony, but that is just a gut feeling ...
It's not that simple ;)
I don't think they would either. In terms of process development it makes more sense to have a unified solution, considering AMD is doing the design. AMD will want to reuse and leverage the Process IP as much as possibleS
Closer to 8% ;) To be honest I thought it would be little more.
Someone could, but I wouldn't trust anyone, especially someone like Mark Cerny that definitely knows better, that seriously tried to do that as part of corporate PR. By that metric, virtually all computing hardware supports raytracing, so it's a completely meaningless statement. It's like saying "hardware supports web browsers" or "hardware supports climate simulation" when it's just a 486 running Windows 3.11. Which, by the way, the hardware I most used ray tracing software on was a 486 with POVRAY.
As amd only recently have patent for hybrid rt implementation I think many sony and microsoft engineers could help with own solutions so I think there can be difference in rt approach.
Some news for the most optimistic members in this thread:
RDNA2 is not a topic for next gen. Why I say that because the most recent AMD Corporate Slide Deck talks about 7nm+ and RDNA2 as "In Design" and not as "Design complete". Time now would be to short for expecting this new technology in next-gen consoles.
Developers (respected ones at that) told us specifically to not expect it.
Yeah no tensor denoising... yet. I cannot wait till that day though!
thats because people see nvidia shots with that machine learning thingy and assume that must e the RT cores... I don't even know who conned that phrase.
ImgTec architecture and patents are available for licensing. It will be interesting to see if they have any takers.
These announcements are often made after the fact. That doesn't mean I think it's likely to show up in consoles. ImgTec is hurting since losing Apple so they need to leverage their IP any way they can.
Let's say you are right, wonder how that could affect it on one console more than the other. Like, could it be more complicated on one than the other that we'd get lesser results? MS does have some kind of implementation on Windows I think with Direct X.
I used to agree that it could be read that way, but the more I think about it the less I can support that position. The exact quote is
I've learned that devs always underestimate certain things (and overestimate on other things too of course). I feel most devs right now aren't really that enthusiast (based on interviews) about next-gen. A lot of people think devs know everything about the equipment their working with. Everybody, devs and players, will be surprised!
Hmmm but I was disappointed with the physics even in SP games this gen. This also doesn't bode well for my hope that sports games get revolutionized physics.
It's more up to devs than hardware limitations, if that's what you mean.
Yeah, didn't the article say that the focus on MP in games affected advancements in physics? If that's the case then I don't see Madden or 2K getting new physics engines
So for that purpose, fast storage and more main memory should solve that issue right?
I've never played any of those sports games but last gen vs current the leap was still substantial:
Even in Gears 4 to Gears 5, physics were toned down in certain areas. But Gears 5 had higher-res textures and poly-counts. I'm guessing that as asset quality is increased, more memory is required to have them persist?
Always love when technological limitations are worked into the story/gameplay. I recall that in Jak and Daxter, Jak would trip if he was running and the next area hadn't been loaded yet.
