Recent content by GuardHei

Actually you are right. There should be lumen. Long story short, I feel odd about not having lumen showing up in all the frames I captured via RDC, so I contacted my friend at GS today. I got the confirmation that Lumen is indeed in use if GI is set to cinematic without Raytracing. It took us...

So… I did some in-depth dive into one frame of BlackMyth using RenderDoc, the findings are pretty unexpected. Everything I talked about is based on Cinematic Quality on a rtx 3080 with 2k output rendered at 50% up scaled by FSR2. No raytracing. Here to confirm a few points already discussed in...

You are not wrong. Answer is they still heavily rely on screen space raymarching to produce most results in diffuse & specular gi passes. HWRT is more like one of the fallback solutions when the ss technique fails. In fact since there're no small vegetation and decoration props in the RT BVH, ss...

I think the reason CPU pressure scales with resolution is due to the way RE engine handles occlusion culling. At least in RE2:RE, occlusion culling is handled via CPU based software rasterization on a low res buffer. It is reasonable to assume that the occlusion test res gets increased slightly...

Finally we have Relink covered. And yes that camera stutter is super annoying and perceptible across the entire game. Also I believe they use the same engine from Platinum games, as the game assets shared the same format as in Nier Automata, with some improvements like using Granite to stream...

To give FSR2 credits, the devs do iterate the algorithms quite a lot over the year. There are quite a lot of additions now compared to what I've seen in summer 2022, mainly focusing on transparency and motion recognition. Although these hand-tuned solutions are understandably limited. And...

I think UE5's TSR sorts of walking in this direction as well. From the shader codes, they seem to apply a form of morphological AA based on Luma whenever the history sample is rejected (could be FXAA/SMAA, or their in-house edge search function) Although I would argue these spatial data feel...

As much as I'm not satisfied with the quality and performance of FSR 2, I don't think CBR outperforms FSR 2. At least the OG implementation, which is a simple 1/2 internal res rendering with diagonal samples and using the differential blend to reconstruct the info spatially. The only reason I...

Ya, this is what I mean by "via MSAA". For the TAA issue, I guess with more games using vis buffer, resolving subsamples would be a lot easier given instance and primitive ids.

You can't increase the shading rate via VRS. The upper rate limit is bounded by the original resolution, which is a one-one pixel mapping. If you have a shading rate > 1, you don't have extra geometry data to evaluate. However via MSAA you get that extra data to evaluate on selective instances...

That last question feels like it pops up from nowhere

Unlike in forward shading, MSAA in deferred shading has a lot of different implementation, ranging from old school to effectively selective super sampling. For example, in RDR2, their MSAA implementation will select pixels whose MSAA subsamples have high normal/depth divergences and run lighting...

Also TAA has more variations in implementation than probably all those other AAs added together. From the way you jitter the frames, the way you accumulate the history buffers, the way you reject and rectify the history buffers, the way you resample the history buffers, the way you treat...

Without TAA you not gonna see those giant visual leaps in the PS4 era. Give up TAA and run ssao, ssr, godrays, froxel fogs, parallax occlusion mapping, any forms of raytracing at full res and ultra high sample count. Now look at the single digit framerate.

those are all relics from the past now. Different forms of the almighty TAA have ruled the realm 🏰