STiFU

I actually fixed it now. My 3060 Ti OC used to overheat. I RMA'd it, but Gigabyte was having none of it. So, I had to undervolt it to prevent it from overheating, and it turns out, that undervolt is simply not stable in some modern games, most notably UE5 games. I had tested many games and benchmarks with that undervolt and never had problems. Then two years later, problems suddenly arise. I ruled out that the undervolt could be responsible for the crashes because I saw the game crashing even with "MSI Afterburner" (the undervolt / overclock tool) closed. However, this conclusion was simply incorrect, because closing MSI Afterburner does not revert the GPU to stock state, the undervolt will stick. I am now running a super conservative undervolt with a slight underclocking of my GPU and the game is absolutely stable and I can now even raise many settings to high or epic, which used to produce the crash much sooner. I finished the game by now and I too am blown away. Here's my review, if you care for it. I omitted all my technical issues with the game, because those were simply on my stupidity.

Man, these crashes are getting out of hand, I am in the finale of act 2 and it keeps crashing on me... It's so annoying that I am thinking about quitting, although I really enjoy the game and I am so close to the finish-line. It's a huge bummer! Some people reported that switching to DX11 renderer would help with the constant crashes, but it didn't work for me. I tried various ini tweaks etc. Just now I even tried disabling DLSS and setting everything to low. Still crashes eventually. It seems to dependent on the area, as I had sections of the game without any crashes for a whole evening. Then again, there were sessions were I had 10 crashes. I eventually created a "kill and restart" batch so that I at least could quickly get back into the game. By the way, it turns out that those LOD issues were due to me using the DX11 renderer.

Avowed looked like a 9/10 for me at first, because exploration, visuals and combat were great. However, after some time, I found the pay-off of the exploration was not very high, as the loot is mostly meaningless (yaaaay, a tree branch). Then there's the problem with the lacking enemy variety and, as a result, also lacking combat variety. The few basic enemy types are also grouped up in the wildest combinations, which kind of eats away from immersion. The soundtrack also gets old rather qickly, which led me to put on the "Epic Fantasy Music"-Playlist on Spotify in the background... Eventually, I stopped playing for these reasons, but would still rate it 7/10. For Anti-woke-warriors, this game will probably be more of a 4/10 or something like that...

I like Clair Obscur quite a bit, but it crashes fairly often for me. Also, in Act 2, some of the models look really broken, as if an LOD-model was loaded in rather than the high-poly model. It is weird, because I know that the game uses Nanite. Did they actually intentionally include such low-poly models or is this bugged?

Sifu is free on Epic right now, but the offer ends in 50 minutes, so be quick. It is an awesome game. It adapts the combat system of Sekiro and extends it in interesting ways.

I was always put off by the visuals of vampires survivors. Eventually, I tried it on my cell phone and it was a lot of fun. I think the phone is also a very suitable platform for this game, because the controls are extremely simple and you don't mind the bad visuals as much as on PC or console.

I was just about to post here that Death's Door is free on prime, when I saw its thumbnail on Twitch. However, apparently it is actually not on prime gaming right now. Hopefully, it will be in the future, it is fantastic: "A darker and more difficult adaption of the Zelda-Principle. It is a lot of fun, has a great soundtrack and an atmosphere that compares to the also brilliant Hollow Knight. The only criticizm I have for this game is that it ended too soon after 11 hours of playing." EDIT: I just realized that half the Screenshot is German, but I guess you get the gist.

If we were able to convert our existing materials to "good enough" PBR materials, we could make the switch to a PBR renderer. This would allow us to successively improve our materials and new contributions could be made directly with PBR. If we had to first fix ALL materials manually up for PBR before making the switch, I don't think this small project would ever take off. That's the motivation for the investigation if the materials can be converted in a sensible way.

The main issue seems to be to derive metalness. What does our current specular map do? It basically encodes reflection intensity at a constant roughness. Isn't reflection intensity basically metalness?

Congrats! Have to give this a shot soon!

I have visited Marienburg (Malbork Castle) in Poland this summer. It is the biggest brick building in Europe and was built and used by the German Tuetonic Order. It was later used by the Nazis and many parts were destroyed during the war, but it has been faithfully rebuilt. Here are the pictures I took: https://drive.google.com/drive/folders/17Vd-2kJG6sLfBi0cCJ6-g9wBT733T5lZ?usp=sharing

I played Fallout 2 back then and it was a great game. I vastly prefer the oldschool c-rpg Fallouts. Sadly, due to 16 bit color depth and low resolution, these games did not age too well. Here's to wishing for a Remaster.

Yeah, I was not very optimistic about my idea to begin with. So, we'd have to painfully go through all materials (and custom FM materials) and convert them manually. Makes me wonder if it would be feasible to support both PBR and traditional materials in a somewhat efficient way in our renderer, so we could still have a functioning game while not all assets have been converted. @stgatilov?

PBR sure would be a major upgrade for our game. Problem is, all our assets would have to be converted and that includes custom assets from FMs. I do wonder, whether there is any research on automatically converting traditional materials to "good-enough" PBR-materials, or if we should experiment ourselves with such a process. I imagine a heightmap could be "approximated" from Normalmap (I know this is not accurate, but would maybe be good-enough for starting), metalness could be derived from thresholded specular map (or material name), glossiness could be derived from normalmap roughness and maybe specular map and as albedo we could simply use the existing diffusemap (maybe try to get rid of any brightnessvariations on that map). @stgatilov care to comment?