Support NeoGAF

[Shogmaster]

Am I missing something in the kep vs mlid debate? When I hear things like 10x ray tracing, while it obviously affects performance, I'm assuming most gains will go towards the quality of the image. I'm expecting 60fps (or 90/120fps in some cases or with MFG) to still be the target for next-gen consoles and for devs to focus any uplift more into image richness, npcs, physics, new ways of play, etc. I don't hear 10x ray tracing and think 10x the frames- just the availability to use more, quality, and performance will be higher.

I always look forward to new consoles as it gives devs a new standard baseline to develop for. At this point with AI upscaling and MFG I do wonder what 25% will contribute to, since traditionally more power in consoles has gone towards resolution, frames, sharpness, and texture quality. AI tools might make image sharpness look the same despite relatively low power differences. New effects or density would be nice as you move up or into pc. I'm excited for next-gen and the efficiency gains.

They're just having lovers' quarrel

 

[dgrdsv]

I wouldn't be shocked if future titles RT on vs off shows little to no performance gains but the RT version looks better.

The only way this could happen is if a game would have s/w RT which would produce the exact same image as a h/w RT option.
In all other cases a non-RT version will always run better just because it's omitting a lot of RT related compute on both the CPU and GPU.

 

[Black_Stride]

The only way this could happen is if a game would have s/w RT which would produce the exact same image as a h/w RT option.
In all other cases a non-RT version will always run better just because it's omitting a lot of RT related compute on both the CPU and GPU.

I take it you you didnt watch Nvidias GDC talks?
Effectively primary lighting using RT is getting so fast gen on gen that eventually there will be no real benefit to using rasterization for primary lighting.
They are all free by the way the talks that is, no vault lockout.

^Forgive the poor screenshot.
Imma try getting the full paper presentation to capture better.

 

[LiquidMetal14]

 

[damiank]

25% faster and (at least) 50% more expensive, fuck yeah. We are xbots, we are so back.

 

[dgrdsv]

I take it you you didnt watch Nvidias GDC talks?
Effectively primary lighting using RT is getting so fast gen on gen that eventually there will be no real benefit to using rasterization for primary lighting.
They are all free by the way the talks that is, no vault lockout.

^Forgive the poor screenshot.
Imma try getting the full paper presentation to capture better.

I've seen this, and this applies to RT vs RT scenario, not non-RT vs RT - why would you have "primary rays" otherwise?
One reason why this trend is apparent lies in the fact that rendering is becoming more complex over time. RT primaries are cheaper in a more detailed scene on the h/w which can do more rays per second - which is what the wireframe shows.
Also doing them with the usual Z fill is still faster even on Blackwell, and there aren't a lot of benefits to doing them on RT h/w - sure, you can do some camera lens like deformations but do we even want that in our games?
All in all this won't mean anything for RT performance, on the contrary it could mean that RT is going to become even slower cause that's how you arrive to Z fill not being as fast as doing it via RT h/w.

 

[viveks86]

I've seen this, and this applies to RT vs RT scenario, not non-RT vs RT - why would you have "primary rays" otherwise?

Though the previous topic was about primary rays, that particular test is an end to end frame time comparison of raster versus RT. Not RT with primary and RT without. They don't even talk about "primary rays" at that point. Goal was indeed to show that rasterization may become obsolete very soon. Obviously it's a very basic, static test and there is a ways to go before we can overhaul everything, but it demonstrates that it's an important discussion for the future.

 

[dgrdsv]

Goal was indeed to show that rasterization may become obsolete very soon.

It may become obsolete in the context of what they are showing - i.e. ray tracing where primary rays are being "shot" with rasterization h/w right now, by using the good old Z buffer.
It's quite obvious that rasterization h/w itself won't be obsolete for quite some time still - although it is possible that at some point it will be easier to emulate it on generic compute, but this is a completely different topic.

 

[Hemingwayoffbase]

Who gives a shit? Multi platform games all have to conform to the lowest common denominator (I.e. the weakest system). This "extra 25% power" will literally never be used.

 

[J Required]

Who gives a shit? Multi platform games all have to conform to the lowest common denominator (I.e. the weakest system). This "extra 25% power" will literally never be used.

25% in this day and age is imperceivable.

 

[manzo]

First generation where I just don't give a flying shit anymore. Will wait for Pro models even if then - only Expedition 33 had been a game I have played this gen that hasn't been available on PS4.

If the same slop keeps coming, will stick to PC for those rare new games (Ys series) that I have anymore interest in.

I'm more waiting for AES+. Just returning to play more 8-bit all the way until PS2/DC/Gamecube. Enough excellent games to finish there than play most of the dreck that comes out.