Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
Lets imagine this scenario, that can be extrapolated from Matt's words.

Sony uses a diferent RT solution. It's questions are not random, but directed at AMD's and DXR weak spots that sony has covered.

The problem is: those listed 'weak spots' are not those of DXR, AMD or NV, they are just general to raytracing.
This makes the assumption 'Sony has super powerful RT' unlikely.

More probable is assuming they have different RT, because they started work on it prior to AMD.

But the outcome could end up like you speculate, e.g. if they had additional HW solution for photon mapping, radiance caching or whatever. But then we would talk about more than just RT. That's far fetched.

Another option would be addition reordering and / or BVH builder HW. This would remain in realm of RT, ImgTec already did it, and DXR would be partially incompatible with that (as is currently). So this could make sense, and it could compensate 2-3 TF less than competition, btw.
Still, it's not that you can not do multiple bounces, or transparency etc. using DXR. The difference would be only in performance, and with this in mind i would phrase my poking differently than he did.
The tweed is just too confusing to give us more than wild speculations.
 
Odium is apparently verified as a..."QA tester". I mean are we just ignoring that? That's a game tester=low level lackey in my book. But apparently some people just go "oh, he's got the verified stamp, nobody dare question 100% legit!". What's next, verifying gamestop clerks?

Again, of course. Low level lackeys are the only ones going to post repeatedly on Neogaf/resetera and break NDA's right and left.

Imagine how seriously sued, or more likely fired I guess, any "real" insider would be in danger of being. Including having to show identifying info to the mods to get verified.

You're mistaking o'dium for someone else..
o'dium is currently developing an indie idTech FPS, but AFAIK he worked on other games for large devs before and he hangs out with people from his former teams.

And QA tester AFAIK doesn't mean "game tester = low level lackey". Regardless, I don't know who gl0w is. I don't follow anything on reeesetera.
 
You're mistaking o'dium for someone else..
o'dium is currently developing an indie idTech FPS, but AFAIK he worked on other games for large devs before and he hangs out with people from his former teams.
So i'm understanding correctly:
to qualify RDNA1.0 and RDNA 2.0 you need to know what both look like.
You need architectural diagrams of both. 1 is available the other is not
Then you need architectural diagrams of both XSX and PS5 both of which are not available.

Then as he claims:
a) does not know any values for PS5 clock speed or TF values
b) has only heard things about it
c) likely does not have access to XSX by every understanding if he doesn't have access to PS5.

d) but claims and I quote:
Both systems are rdna 2, but not. They are essentially rdna 1.9 so to speak, with features bolted on, meaning that you can call it the next generation of rdna, you could even call it rdna 2 if you wanted. I wouldn’t myself but that seems to be what they are doing so ok."

then goes on to say:
What I can’t ascertain however, is if either are using a split memory pool for OS level tasks. I haven’t had a single straight answer on that one, other than “its complicated, but yes, and no”. That leads me to believe that possibly something else is going on here. If the PS5 has a separate look of let’s say 4gb DDR4 then they can use the full 16gb for games opposed to less for the SX. But what’s been alluded to for me is a bit strange, almost like you don’t have to worry about it as much. Which I can’t understand at all? Any ideas?

but somehow he has in-depth knowledge of the differences between all 4 gpus to make a statement that XSX and PS5 are not RDNA2.0 as per AMD spec, but instead something else. Cause it sounds like to me he hasn't any idea what the SoC diagrams look like.

I just want to be clear on something: You know people can be wrong.
When I was at Build 2015 asking questions in a Unity DX session, I asked about XBO feature set levels and I got 2 different answers. One told me 12_1, the other told me 12_0. People can be wrong. Even straight from the horses mouth.
 
At this point I'm just hoping RDNA2 has something like Nvidia's mesh shaders and texture space shading, but if they revealed VRS as a feature, which is really something only gpu nerds would understand, then they probably would have mentioned those other things as well.
Yeah i don't have hope for it. Has never been mentioned, not even a rumor, nothing.
If NV extends task shaders also for compute, the result would be a real game changer. I hope they do so and the competition has to react quickly at least after that.
As long as it is only about triangles, i think we can live without it.

I really hope there's some hardware ML capabilities in their on both consoles.
With Navi12 already having those dot products instructions, i guess it's only a question if MS / Sony wants it or not?
Seeing not so much application for ML in games for now, AMDs approach of adding just those dot products seems a much better area compromise than full matrix ops as in tensor cores IMO.
 
The problem is: those listed 'weak spots' are not those of DXR, AMD or NV, they are just general to raytracing.
This makes the assumption 'Sony has super powerful RT' unlikely.

More probable is assuming they have different RT, because they started work on it prior to AMD.

But the outcome could end up like you speculate, e.g. if they had additional HW solution for photon mapping, radiance caching or whatever. But then we would talk about more than just RT. That's far fetched.

Another option would be addition reordering and / or BVH builder HW. This would remain in realm of RT, ImgTec already did it, and DXR would be partially incompatible with that (as is currently). So this could make sense, and it could compensate 2-3 TF less than competition, btw.
Still, it's not that you can not do multiple bounces, or transparency etc. using DXR. The difference would be only in performance, and with this in mind i would phrase my poking differently than he did.
The tweed is just too confusing to give us more than wild speculations.


If Sony has "their own" RT solution (collaborating of course), I'd be afraid. And it could be the reason PS5 is allegedly undesrpecced in other areas.

I feel like Sony's lack of ability to admit they can no longer engineer console silicon is what led to the Ps3 semi-disaster. They wanted to be involved, in some way, any way, so they made Cell, even needed IBM help. Probably wasn't a great idea. Even worse if they'd used as GPU too as supposedly they originally wanted.

That said why do we think Sony might have their own RT solution? Surely not the mighty github leaks. But that's the only thing I can figure where they started....since that test didn't include any RT test on the alleged PS5.
 
If Sony has "their own" RT solution (collaborating of course), I'd be afraid.
According to Odium, since PS5 and XSX are both RDNA 1.9, so to speak, then they both should have the same RT and VRS blocks from RDNA 2.0 back ported to RDNA1.9.

So by this admission of evidence Sony is using AMD's solution and not Power VR or photons or their own custom solution.
 
Last edited:
According to Odium, since PS5 and XSX are both RDNA 1.9, so to speak, then they both should have the same RT and VRS blocks from RDNA 2.0 back ported to RDNA1.9.

I don't follow this logic.
If A and B don't belong to X, then A and B must both belong to Y?
 
So they scrapped their plans on going for cheaper at $399 ?
 
I don't follow this logic.
If A and B don't belong to X, then A and B must both belong to Y?
Odium says both are RDNA 1.9, not RDNA 2.0
We know XSX has both VRS and RT blocks from RDNA 2.0. But even with that it is still missing 0.1 for whatever reason unknown.
If PS5 is also 1.9, it must also be missing that same 0.1 for whatever reason that is.

So by default PS5 must have the same RT blocks. because not having it would conclude it to be a significantly less of a number. Like a RDNA 1.5 or less.
 
So they scrapped their plans on going for cheaper at $399 ?
Someone suggested radeon7 was the result of sony switching plans in 2017. Same 1800mhz boost clocks. Rather than wasting money spent on r&d, they repurposed it into a desktop gpu to fill the gap that was left open by a delayed navi card.
 
Status
Not open for further replies.
Back
Top