Digital Foundry Article Technical Discussion [2022]

[Flappy Pannus]

Why would they have to hook into another platform and be storefront specific though? It's literally just uploading a file(s) generated by the game to a server..

Who's server though? Who pays for that?

and the developer takes those files, generates a list of all the PSOs by all the users, and then recooks the package again, and pushes the update out to multiple storefronts.

Yeah, I get how it works. But like...networks aren't free.

Basically my joke was they could get away with just using Valve's infrastructure for uploading the PSO data files. They wouldn't need to do it separately or for each store.

Well yeah, but of course they would need to do it separately, as Valve just doesn't hand out access to their servers for nothing.

 

[Remij]

Who's server though? Who pays for that?



Yeah, I get how it works. But like...networks aren't free.



Well yeah, but of course they would need to do it separately, as Valve just doesn't hand out access to their servers for nothing.

Maybe I'm not being clear?

Valve's server. Valve provides the service to developers, developers use service. 30% cut continues to do its work...

And no... they wouldn't need to do it separately for each storefront.. The files the game would generate would be completely storefront agnostic... It's literally: Files uploaded to a server where the developers can access them... and then the developers take those files, and recook the game.. and send out the patches to all storefronts.

 

[Flappy Pannus]

And no... they wouldn't need to do it separately for each storefront.. The files the game would generate would be completely storefront agnostic... It's literally: Files uploaded to a server where the developers can access them... and then the developers take those files, and recook the game.. and send out the patches to all storefronts.

So basically, you're proposing renting out shader butlers, with Valve being the intermediary service to facilitate this crowdsource network. Ok.

The concern I have with this approach is it's still relying on the developer to initiate this process every time they need new PSO's generated. So DLC's, patches, new characters - expecting them to rent out a service offered by Valve to do this every time leads up back into the problem of developer involvement. This is something I want to ultimately decrease, I can see a benefit for very small developers that don't have adequately sized QA teams to generate these with their own playthroughs before launch perhaps, but the problem has also come from large studios who just don't bother. I don't trust them to keep going back to this hypothetical service for game updates.

You're expecting developers to keep going back to pay for a third party service to do what they're not bothering to get their own internal QA teams to do. My approach would be to first and foremost provide the hooks for the engine to be able to import/export the PSO list continually, it's something that gets done automatically as people play through the game. Developers don't have to 'recook' the game and 'send out' patches, it's done continually, just as Valve does it now with compiled Vulkan caches.

 

[Remij]

So basically, you're proposing renting out shader butlers, with Valve being the intermediary service to facilitate this network. Ok.

The concern I have with this approach is it's still relying on the developer to initiate this process every time they need new PSO's generated. So DLC's, patches, new characters - expecting them to rent out a service offered by Valve to do this every time leads up back into the problem of developer involvement. This is something I want to ultimately decrease, I can see a benefit for very small developers that don't have adequately sized QA teams to generate these with their own playthroughs before launch perhaps, but the problem has also come from large studios who just don't bother. I don't trust them to keep going back to this hypothetical service for game updates.

No. You're really overthinking this.

Basically I'm saying I think it would be a great service to developers, if PSO caches could be automatically uploaded to a server they could access... so that they could use it to update the game to provide the best PSO coverage possible for users to pre-compile as many shaders as possible.

The game is just logs and creates the files necessary as you play it.. those files go into a folder, and then when you close the game, that file gets uploaded to the server. Then they could take those files, update the game with them, and push the update out. Whether that's on Steam alone, or all other storefronts... wouldn't matter at all. I mean, do you think the developers of The Callisto Protocol made different PSO caches during development for the different storefronts? lol..

You're talking about renting out services and all this stuff lol... no.. it's not that complicated.

But let's just forget that I said anything, it's not worth arguing about.

 

[Flappy Pannus]

No. You're really overthinking this.

Or not enough.

The game is just logs and creates the files necessary as you play it.. those files go into a folder, and then when you close the game, that file gets uploaded to the server.

TBC, the contention has never been of this from the start. I get how the PSO gathering process works.

Then they could take those files, update the game with them, and push the update out.

That's my minor quibble/desire though - I don't want to necessarily have this responsibility be up to the developer to push these out as it will eventually atrophy, I'd much prefer a system when the updated PSO list is just continually uploaded and downloaded by the community with no further input required from the developer. Basically acts exactly like how Valve distributes Open GL/Vulkan compiled caches now, but it would just be the generated PSO list. I mean, Valve already supplies the architecture now, so my thinking is why not use it.

However your proposal definitely provides the least friction. Having the developers ultimately be the ones to distribute this gets around the security concerns of a peer to peer system as well, which is certainly something I've seen raised when it comes to distributing compiled caches.

I think just focusing on the gathered PSO list as the distributed data instead of the compiled shaders is a great idea though, so kudos for that.

 

[trinibwoy]

Is DF going to cover the Witcher 3 update on PC?

 

[techuse]

Is DF going to cover the Witcher 3 update on PC?

Alex will be I’m sure.

 

[BRiT]

Is DF going to cover the Witcher 3 update on PC?

They did a preview of it, so I would assume they'll cover the full release too.

 

[Pete]

Is DF going to cover the Witcher 3 update on PC?

From Alex himself: “Portal with RTX and The Witcher 3 are my priority before holiday.”

 

[mr magoo]

So.. the Blizzard WoW client from twelve years ago? Yeah, kind of nuts how some companies jumped on P2P tech to distribute data really quickly and others - like 97% of the industry - have seemingly stuck with CDN.

With one important addition you don’t only download/upload content but also generate content. As everybody else.

 

[Cyan]

Is DF going to cover the Witcher 3 update on PC?

gotta play the game once again and complete -again- my favourite expansion, Blood and Wine. The download weighs around 50GB now -it was like 29GB before, expansions included-.

 

[SmooTh]

 

[davis.anthony]

I still can't get over the resolutions in that game as it's not even that special looking.

Only thing I can think of it the draw distance being especially taxing as the grass is drawn farther out than I've seen in a lot of resent games so the game has a high transparancy load?

 

[Inuhanyou]

Great video @Dictator I can see now how path traced gaming is so different from individual RT effects or even a few a once.

In the video you mention the limit of the current technology for this game is 4 bounces. Doesn't reality bend light infinitely in real life? I wonder if a future proofing mechanism could be to unlock ability for more bounces in the future

 

[Dictator]

In the video you mention the limit of the current technology for this game is 4 bounces. Doesn't reality bend light infinitely in real life? I wonder if a future proofing mechanism could be to unlock ability for more bounces in the future

Reality does not bounce light infinitely, the energy and photons are eventually absorbed after a few bounces. The amount of real bounces depends on the surface properties. Mirrors like surfaces can bounce photons around tons of times before absorption, something like dirt? Not as many.
Regarding "4 Bounces" I state in the video. "4 Bounces" is not the limit in Portal RTX - that is just what the ultra setting is. You can set it up to 10 beyond Ultra. It is extreme diminishing returns though for visual differences beyond 4 tbh in portal.

 

[Inuhanyou]

It is not the limit - that is just what the ultra setting is. You can set it up to 10 beyond Ultra. It is extreme diminishing returns though for visual differences beyond 4 tbh in portal.

I see. So there is a setting beyond the default 4 ultra provides. Got it

 

[Karamazov]

 

[Cyan]

Great video @Dictator I can see now how path traced gaming is so different from individual RT effects or even a few a once.

In the video you mention the limit of the current technology for this game is 4 bounces. Doesn't reality bend light infinitely in real life? I wonder if a future proofing mechanism could be to unlock ability for more bounces in the future

great video as usual! Quite surprised as to how DLSS Ultra Performance looks in this game. Can only imagine a fully path traced Witcher or Skyrim. This game could be my first Portal game to date but I am not getting it for now. It crashes in my PC's GPU if you don't play using Vulkan. Has anyone tried Portal RTX in Vulkan mode with an nVidia card?

Apparently the game has a lot of graphics glitches when using Vulkan, even on nVidia cards.That's why I ask, I am not in a hurry to get the game.

 

[Inuhanyou]

Reality does not bounce light infinitely, the energy and photons are eventually absorbed after a few bounces. The amount of real bounces depends on the surface properties. Mirrors like surfaces can bounce photons around tons of times before absorption, something like dirt? Not as many.

Science and physics never was my strong suit but of course it makes total sense with just an bounce of brainpower. "Rough materials" like matte surfaces obviously aren't shiny or reflective.

Reality itself is very weird. I've also heard the claim made that "people can't see beyond x arbitrary fps" but different parts of the eye process information at different speeds and scales to begin with and dynamically adjusts. Otherwise high fps gaming would be a thing to begin with as we would not be able to process it right?

 

[Dr. Nick]

Excuse me @Dictator could you please try it on an ARC gpu to see how well it performs in comparison to the AMD GPU? I'm betting the game will do better in it's current state on Intel GPUs.