NVIDIA's Morgan McGuire predicts that the first AAA game to REQUIRE a ray tracing GPU ships in 2023

For a big budget game that does sound optimistic, even though if new console generation will have RT acceleration in hardware.
I wouldn't be surprised if some developers would try to make it happen as cascaded shadows can be really painful in many cases. (And have been somewhat broken in UE4 for a long time.)

For small scale game I certainly could see it to happen earlier.
 
I think for that to happen, the next NVIDIA GPU series will need to at least double RT performance.

3070 needs to perform like the Titan RTX in terms of RT performance. By 2023, I think 4K displays will be common enough that 4K will be mid tier spec and not high end spec.

So I hope the next gen GPUs bring a Pascal like gain.
 
He speaks about stochastic lightcuts for Global illumination inside his SIGGRAPH presentation and about decoupled sampling from visiblity for great motion blour and defocus blur/depth og field) and could be useful for AA too.:D
 
I wonder why they declare the first Geforce as a hallmark when it was functionally identical to the earlier Riva TNT2 except for T&L which could be done in the CPU at decent speeds until much later.
Also, the same slide suggests Doom 3 was the first title to demand hardware T&L, 5 years after releasing in the Geforce 256. But I thought the game had the Geforce 3 and Radeon 8500 as minimum requirements, both of which already had pixel shaders.
Besides, until well into 2001 a bunch of graphics cards without T&L were released and some were fairly successful, such as the PowerVR Kyro I/II, Voodoo 3/5 and Savage 2000.

Seems to me that programmable shaders were a much more important hallmark than T&L ever was. Especially pixel shaders 2.0 which right after they came out no one else dared to release new dGPUs without tem.

In 2023 we're probably looking at 3nm nodes. At that point we should have the equivalent of a gigantic TU102 (754mm^2) in ~150mm^2 chips, i.e. similar to a modern day GTX 1050 Ti.
Honestly I think it depends on how far VR/AR gets until there. If VR gets too much attention, developers will be more concerned with very fast rasterization speeds for high-resolution headsets than with higher detail on computer monitors / TVs.


Regardless, if 2004's Doom 3 really was the game that regarded T&L as "mainstream" and made full use of it, then no one should have invested in a Geforce 256 because of T&L. By the time Doom 3 arrived the Geforce 256 had terrible performance.
Which in turn doesn't sound like a very good incentive to push people to purchase a Turing GPU because of raytracing.
 
REQUIRE ray-tracing in 2023? Man, that sounds a bit optimistic to me unless it's a total TWIMTBP deal with the game developer.

Requiring cutting edge tech can cut your own throat if the anticipated body of users/hardware doesn't materialize, the developer would need nVidia as a safety net for failure.
 
TNT and GeForce 256 have more differences than T&L. For example real trilinear filtering was added, and register combiner functionality overhauled.

They might see Doom3 as a turning point because it has a special rendering path for the NV1x GeForce cards. I think it's the first game to require DirectX 7 level functionality? It also doesn't really support any DirectX 7 cards from other companies.
 
Last edited:
I couldn't stop thinking about this one. Wouldn't the hypothetical game have to have started development just about when the first RTX cards came out and be built up around whatever RT language nVidia is using or is there such a thing as DX RT now?

Seriously, I can't see ANY game requiring ray tracing for quite some time. Hell, look how long it took them to adopt to multicore CPUs! Does anyone really think nVidia's RT is a paradigm changer to the extent that everyone has just fully embraced it?

I'm really wondering here, am I missing something?
 
It's not 20 years ago, there's no cutting edge PC games market that can innovate freely and then shoe horn in a console port later just for the money as iD did with every game until the PS360 gen.

If console support dedicated hardware ray tracing, it will come sooner. If not, looking at the 5 years it took for hardware T&L to be required 20 years ago isn't really comparable.
 
I'm really wondering here, am I missing something?

Quite a few things, I'm afraid.

1- There's no such thing as Nvidia's RT language/API. The API used by RTX cards is Microsoft's DXR, an integral part of DX12 since last year.
2- There's already a few games shipped using ray-tracing, with much more coming in the near future.
3- Even if they had to start making a game when the first RT enabled card was launched, perhaps in orther to make it revolve around it instead of being a bolt-on feature, wouldn't 5+ years be enough time for development? I'm pretty sure most games' development time is much lower than that nowadays, unfortunately.
 
.
3- Even if they had to start making a game when the first RT enabled card was launched, perhaps in orther to make it revolve around it instead of being a bolt-on feature, wouldn't 5+ years be enough time for development? I'm pretty sure most games' development time is much lower than that nowadays, unfortunately.

No. Nintendo platform says "Hello", again.

He said every gaming platform would offer accelerated ray tracing. Everyone should have legitimate doubts that Nintendo will offer that.
 
No. Nintendo platform says "Hello", again.

He said every gaming platform would offer accelerated ray tracing. Everyone should have legitimate doubts that Nintendo will offer that.

Sorry, but I don't see how my 3r point relates to your reply? Or any of my points really? Or Nintendo?
 
Sorry, but I don't see how my 3r point relates to your reply? Or any of my points really? Or Nintendo?

No amount of time spent in development will change the fact that Nintendo likely will not have accelerated ray tracing in 2023.

Also, I don't see how what you said relates to the topic at all, namely the quote of "I predict that the first AAA game to REQUIRE a ray tracing GPU to run will ship in 2023, and every gaming platform will offer accelerated ray tracing by that year." Everything else is secondary.
 
No amount of time spent in development will change the fact that Nintendo likely will not have accelerated ray tracing in 2023.

So? Again, how does that relate to my post? I was replying to digitalwanderer.

Also, I don't see how what you said relates to the topic at all, namely the quote of "I predict that the first AAA game to REQUIRE a ray tracing GPU to run will ship in 2023, and every gaming platform will offer accelerated ray tracing by that year." Everything else is secondary.

Digitalwandered expressed his concern that game development time would be an obstacle to the realization of that prediction. I pointed out how, IMO at least, it probably wouldn't be a problem in any case, since 5 years seems enough time to even start from scratch, and the fact that games have already been released using ray-tracing at some capacity.
 
Anyway, I'll bite your beloved Nintendo topic. How can you be so sure the next Nintendo console will lack RT? I wouldn't even assume Nintendo would have a say in the matter if they go the same route as with the Switch...
If Nintendo were to release a new console in 2023, which kinda makes sense, being 6 years after the Switch, and they repeat with Nvidia, they'll likely choose whichever 2 year old SoC Nvidia has around at the time. Will that 2021 Tegra SoC have RT acceleration of any sort? How integrated into future Nvidia GPU architectures will RT acceleration be? Well idk, nobody here does, but Nintendo being Nintendo wouldn't be relevant at all.
 
I was approaching it from the more efficient argument that none of this matters because this won't be the case at all.
 
Back
Top