Doom High Res Sprites - Kickstarter?

Doom High-res Sprites   37 members have voted

  1. 1. Tea Monster, the guy who makes the game-ready, next-gen Doom models, is interested in how much response there would be to creating a kickstarter to make some new Doom sprites from 3D models. They would be accurate to the sprite and of high quality. I would ask for kickstarter backing to help with software costs and I would be taking time away from commercial projects to make these.

    • Yes, I'd contribute to a kickstarter to make some high-quality, accurate Doom sprites
      3
    • No, Doom should be free. I'll wait for someone else to do it.
      13
    • While it sounds nice, this commercial venture would be shut down by Bethesda instantly.
      22

Please sign in or register to vote in this poll.

Well, so far everything seems to point towards that direction, so the only ones potentially affected are Chillax BFG spammers -and those couldn't care less about the sprites being hi-res In the first place. On a completely academic note, just how big are those hi-res imps?

Edited by Maes

Share this post


Link to post

They 634x310, so it's more then 8x bigger than original ( I know it should be in power of two, but remember that it is only test sprite). Coming back to the most important question - how big sprites will behave in sloughter maps. So I do another quick tests. Here we go:

 

(in)famous NUTS.wad + usd3mod

At start everything is ok, but when all monsters are activated there's one big lag and we have 0 fps :)

F7EQOdS.png

 

 

But! I make some map with 1200 monsters visible at time in one sector. 

On start everyting is ok, but when fights go on I notice framedrop to something between 60-24 fps, in most cases more then 50 fps. So it's fully playable.

3mzZjVl.png

0pimeDs.png

taflhND.png

QA401Hv.png

 

I test this map with 600 hires imps...

TNxKXdS.png

5H9tSmO.png

 

...and don't notice any framedrops.

 

Here's my imp if you want to test it yourself:

https://cp.sync.com/dl/be0be1b10#4waejt5z-rb7ng5vk-8b3uqaih-je5yx3pn

1 person likes this

Share this post


Link to post

Well then, at least for GPU-accelerated ports, it seems that big sprites are not a problem.

 

That being said, there's something comical about those screenshots....

Edited by Maes
1 person likes this

Share this post


Link to post
7 hours ago, Reinchard said:

Here's my imp if you want to test it yourself:

https://cp.sync.com/dl/be0be1b10#4waejt5z-rb7ng5vk-8b3uqaih-je5yx3pn

Thanks for posting your imp! I love that imp!

 

I'm glad you're getting good performance out of it. Unfortunately, your tests don't really target the cache issue we were discussing a few posts back, because it showcases frames for only 1 hi-res monster.

 

To *really* test performance, you'd need to have hi-res sprites for all 17 Doom II monsters, and feature a map with all of them in a room. You don't need 1,200 of them, either. Maybe 12 each, or so. It's more about the number of *distinct* hi-res frames, not necessarily the number of monsters.

 

But, because you don't have hi-res images for all the monsters, you can use *any* hi-res images for this test. If it were me, I'd make copies of the hi-res imp images, and rename them to replace the other monster's frames. For example, you could copy TROOA1 16 times, renaming each copy to "BOSSA1", "HEADA1", "POSSA1", etc. This should not take much time to do. (an hour or so, maybe less?)

 

This is a very good, realistic test, and it creates a "worst-case scenario" for the Doom renderer, and for the CPU. When you go into a room with all 17 monsters, Doom will have to load hundreds of hi-res frames to paint the scene. If performance is acceptable in this test, most maps should play fine.

 

If anyone tries this test, be sure to try it in GL mode and software mode, because the potential problems are very different for hardware vs. software rendering.

Share this post


Link to post
5 hours ago, kb1 said:

To *really* test performance, you'd need to have hi-res sprites for all 17 Doom II monsters, and feature a map with all of them in a room.re rendering.

That's exactly what I did two posts above - besides of imp tests there is a test with 1200 hires monsters from Doom3 monster pack for Doom. Have you missed it? The test is reliable, proves that evey wad except nuts-like ones can handle it. Besides of all hires monsters, there is a lot of hires gfx associated with monsters, like fireballs, teleporting effects, smoke. 

Share this post


Link to post

Keep in mind that this is still on a GPU, so caching in the traditional sense might not even exist (older GPUs may not have any at all). There might even be some different handling of overdraws, compared to the software renderer.

Share this post


Link to post

I don't think Tea Monster creating 3d models and hi-res stuff with software rendering in mind. He probably thinking about mods like Doomsday or GZDoom.

Share this post


Link to post

When it comes to hi-res models or sprites, there's always some ambiguity as to what port(s) these are supposed to target, or even left as an afterthought. OK, so Ascension targeted Doomsday right off the bat, at least that's clear-cut even if it leaves a huge chunk of the Dooming community out. Now, unless there's some cross-port, broadly compatible way of making either of them work, such projects will always be left in a Limbo of sorts, because who would like to work with the premise of "make the assets first, worry about the target port and testing later"?

Edited by Maes

Share this post


Link to post

I don't know what exactly Tea Monster planned, I'm just doing test to check if this is even possible. So i prove that this is possible on ports like GZDoom or Doomsday. I suppose that if he is planing some extra graphic stuff, he's target are mods with extra graphic addons for sure, not Chocolate Doom for example. 

Share this post


Link to post

If I'm not mistaken ports like the classic ZDoom, ZDaemon, prBoom+ and others allow using sprite scaling and extra viewing angles, which in theory would make them targets for such a mod too, but with more restrictions (e.g. only 8-bit resources in the case of wall textures and sprites). A quick "downgrade" of the imp pack and a test in one of those ports would be most revealing... IMO, a quality hand-drawn sprite with no more than 2x scaling would be improvement enough for most players, even in software mode.

 

Besides, there's a minimum distance beyond which you normally don't get to see textures or sprites, so choosing a resolution/scaling factor also depends on your target audience. It's one thing making sure you'll get little to no pixelation at 320x200, at 640x480 etc. and another at 1080p or 4K.

Edited by Maes

Share this post


Link to post

I agree that 2x bigger sprites are enough. That solution requires less of work and should not produce any slowdowns.

 

Share this post


Link to post

The biggest sprites (e.g. Cyberdemon, SM) are already beyond a scaling factor of 2x compared to most common ones, esp. zombies and imps, so a test using a DEHACKED sprite swap of every monster's frames for the Cyberdemon's in NUTS.WAD would also be interesting, if a little unsettling.

Share this post


Link to post
16 hours ago, Reinchard said:

That's exactly what I did two posts above - besides of imp tests there is a test with 1200 hires monsters from Doom3 monster pack for Doom. Have you missed it? The test is reliable, proves that evey wad except nuts-like ones can handle it. Besides of all hires monsters, there is a lot of hires gfx associated with monsters, like fireballs, teleporting effects, smoke. 

Oh, I didn't realize that the other monsters were hi-res too - my mistake. Yes, in that case, your test is good...for GPU rendering. All of that previous technical discussion was us theorizing about the effects of hi-res sprites on the software renderer. The possible issues with hi-res sprites are completely different when using the software renderer.

 

15 hours ago, Maes said:

Keep in mind that this is still on a GPU, so caching in the traditional sense might not even exist (older GPUs may not have any at all). There might even be some different handling of overdraws, compared to the software renderer.

Yes, I think the sprite images get pushed to the graphics card, and stay cached on the card. So, as long as the card has enough memory, you're good. I don't know why, but I've always felt like using graphics acceleration was cheating, a bit, as far as Doom is concerned. Please, no offense to GL programmers intended - what you've done with Doom is truly awesome. But, I expect Doom to run well on GPUs. What always fascinated me was that Doom could run at all, when powered completely by the CPU. This was Doom's claim to fame, back in the day, after all.

Share this post


Link to post
12 hours ago, Maes said:

a test using a DEHACKED sprite swap of every monster's frames for the Cyberdemon's in NUTS.WAD would also be interesting, if a little unsettling.

Coming from a guy that lusts after cyberasses :)

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now