• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Microsoft is preparing DirectX for Neural Rendering

I'm wondering: Will these features work on previous RTX series cards?

Alan Wake 2 will be updated soon and Nvidia has written that it will support Mega Geometry on all RTX series cards, which seems to suggest that neural rendering will not be restricted to RTX50 series.


rkMfZHx.jpeg
I'm curious to see this by myself, wondering what AMD answer will be on this lol
 

DoubleClutch

Gold Member
There are a few, very important differences.
FG on these games, can access several buffers, like depth, colour and motion, to produce a more accurate image, with fewer artifacts.
Something that a FG that works on the final frame, delivered by the console, can't do.
The second difference, is that all games that use Nvidia FG also have Reflex, to keep input latency in check.

They’re still fake frames, and you add a lot of input delay regardless of reflex. Not to mention the artifacting.

Frame gen shouldn’t exist for gaming.
 

3liteDragon

Member
I'm wondering: Will these features work on previous RTX series cards?

Alan Wake 2 will be updated soon and Nvidia has written that it will support Mega Geometry on all RTX series cards, which seems to suggest that neural rendering will not be restricted to RTX50 series.


rkMfZHx.jpeg
Well, DLSS & FG are forms of neural rendering as well, it's just this time they're injecting neural networks at the shader level, many different kinds of shaders make up all the stuff you see in a video game. All the RTX Kit features (texture compression, neural materials, mega geometry) will work on all RTX GPUs, it's just that the performance deltas between generations will be huge going forward when games implement it. The 20 series won't render all that as fast as the 50 series cards especially since the 50 cards are also the first GPUs that are built to support neural shaders on an architectural level.
NVIDIA said:
With up to 92 billion transistors, Blackwell is the most powerful consumer GPU ever created. The Blackwell streaming multiprocessor (SM) has been updated with more processing throughput, and a tighter integration with the Tensor Cores in order to optimize the performance of neural shaders. Blackwell is enhanced by several hardware and software innovations to improve Shader Execution Reordering. The reorder logic is twice as efficient, increasing the speed and precision of reordering which accelerates the performance of neural shaders.
NVIDIA said:
New Blackwell Tensor Cores have been built with a massive amount of AI processing horsepower, and support accelerated processing of FP4 precision models. With FP4, Blackwell Tensor Cores can not only process models faster, but can do so while using less graphics memory.
The RT cores on the 50 cards are designed to support Mega Geometry as well, doesn't mean it won't run on all RTX cards but it'll be much faster on the Blackwell cards due to architectural support.
NVIDIA said:
Blackwell also features brand new RT Cores designed to ray trace massive amounts of detailed geometry. The RT Cores have 2X the ray triangle intersection rate of the previous generation, and enhanced compression designed to reduce memory footprint. This allows Blackwell GPUs to ray trace levels of geometry that were never before possible.
 
Last edited:
Top Bottom