About Reshade Ray Tracing

[aykekarallo]

Surely many of you already know this, but lately I have been experimenting with Ray Tracing for Reshade (aka RTGI) developed, for some time, by Pascal Gilcher, occasional Nvidia collaborator and who offers it on Patreon where you can find the latest version (google "RTGI Reshade").

Previous versions have been leaked, I don't know if intentionally, on the net.
To use it, obviously you have to install Reshade (latest version) and download the shader.

 

Now let´s go with Reshade:

It cannot be said that it is a "pure" Ray Tracing,but it is close enough and works quite well in game.
In the photos above the difference may seem subtle (my settings are moderate).Changes are best appreciated in the details.

In practice the difference is quite obvious. The lighting is more realistic and the characters seem better integrated into the scene thanks to the finer gradient of shadows.

And no, you don't need an RTX graphics card. It should work fine with any relatively modern graphics card, including AMD.
There is a cost in performance, of course. In my case about 10% of fps (GTX) but I suppose that this can vary in each case.

[erikku90]

I gave it a go and for me, the effect is very minimal, It is probably because I am using several shaders. The only thing I noticed is a slight shadow on the skin and hair.

[hfg2]

Don't know what other reshade filters you are abusing of, but to me the reshade screenshots are actually worse than pure hook5 because it seems you lost a lot of details in postprocessing.

If you want to make a compare then do only a hook5 vs rtgi.

[aykekarallo]

2 hours ago, hfg2 said:

Don't know what other reshade filters you are abusing of, but to me the reshade screenshots are actually worse than pure hook5 because it seems you lost a lot of details in postprocessing.

If you want to make a compare then do only a hook5 vs rtgi.

That is a matter of taste.
In this case, the filters give depth to the scene and color temperature (vibrance and some more, not too many). I don't care, I insist,on this case. details are lost.

But yes, it is true that a direct comparison H5 clean vs rtgi would have been more appropriate.

[ddman2]

RTGI+H5 video: https://mega.nz/file/QY41HJBC#Bm3kUYzcPFlFPzyZbnC3K2_5NOW3_Ty0GdlwUQDIKa8

[aykekarallo]

2 hours ago, ddman2 said:

RTGI+H5 video: https://mega.nz/file/QY41HJBC#Bm3kUYzcPFlFPzyZbnC3K2_5NOW3_Ty0GdlwUQDIKa8

This is how to make a comparison of a graphic injector and not what I have done 

[x17]

3 hours ago, ddman2 said:

RTGI+H5 video: https://mega.nz/file/QY41HJBC#Bm3kUYzcPFlFPzyZbnC3K2_5NOW3_Ty0GdlwUQDIKa8

Honestly, this looks pretty damn good regarding the visual improvement and loss of performance. The gain is considerable, the loss isnt too big. I didnt know about this in particular (the name of author rings a bell though ...)

Now, this is something that could be already worked with...and considering RTX GPU isnt mandatory, thats a pretty big enhancement.

[ddman2]

4 minutes ago, x17 said:

Honestly, this looks pretty damn good regarding the visual improvement and loss of performance. The gain is considerable, the loss isnt too big. I didnt know about this in particular (the name of author rings a bell though ...)

Now, this is something that could be already worked with...and considering RTX GPU isnt mandatory, thats a pretty big enhancement.

Yes it is, but only for rooms like this one, where there are lamps, tv screens and other sources of light. The performance impact is massive, because there is no DLSS or tensor cores.

One thing I didn't show, because it didn't work, is "Render a frame". It will render a still image at max ray tracing quality, and then you really notice the details. For people that like to take pictures, this feature is very valuable because the performance impact becomes irrelevant.

[ddman2]

Here is another video (GPU DM P: VRAM used by Tk, Ram P: RAM used by tk) GPU: 1080TI undervolted a bit (default max clock 1950Mghz). From 100FPS down to 50FPS.

https://mega.nz/file/hQwzEAxY#NtZGLbSW0cIgwFU7pvEvA8Rrgbq_jaI_RX5CDWrdB00

[x17]

2 hours ago, ddman2 said:

Yes it is, but only for rooms like this one, where there are lamps, tv screens and other sources of light. The performance impact is massive, because there is no DLSS or tensor cores.

One thing I didn't show, because it didn't work, is "Render a frame". It will render a still image at max ray tracing quality, and then you really notice the details. For people that like to take pictures, this feature is very valuable because the performance impact becomes irrelevant.

Thats definitely more than just 10 % decrease ... and I think it wouldnt utilize Tensor cores in RTX GPUs, as this is different approach.

Not so good as I thought at first sight ... sadly. 

[ddman2]

23 minutes ago, x17 said:

Thats definitely more than just 10 % decrease ... and I think it wouldnt utilize Tensor cores in RTX GPUs, as this is different approach.

Not so good as I thought at first sight ... sadly. 

You can use cheat engine's speed hack feature to gain back performance lost. The downside is that the game runs slower. Everything will move slower, animation, UI, physics.

I managed to get from 50fps to 80 fps at 0.82 speed(cheat engine).

[hfg2]

Reshade RTGI works in software, you don't need an RTX videocard for that. But yes, a RTX videocard is new hardware anyway, so usually means better than GTX gen.