Polymer lives again
#211 Posted 15 August 2017 - 04:16 AM
#212 Posted 15 August 2017 - 05:07 AM
blizzart, on 15 August 2017 - 04:16 AM, said:
Well, I've never, ever seen Polymer running well on anything so I'd love to see some benchmark results.
#213 Posted 15 August 2017 - 02:38 PM
Zaxx, on 15 August 2017 - 12:36 AM, said:
Weird, EDuke32 classic renderer always runs butter smooth on my system. And definitely runs better than the WT classic mode.
#214 Posted 16 August 2017 - 01:39 PM
Zaxx, on 15 August 2017 - 12:36 AM, said:
Ah so I should just quit mod development now then, since we're throwing non-Windows support out in your world. That's silly. You don't adopt a poor close-source renderer that only works on one platform instead of doing proper improvements.
Edit: I don't know what your performance is but I'm getting over 100fps in Hollywood Holocaust at 1920x1200 in classic and 60fps or higher in the Polymer with the Polymer HRP (which obviously also has models and a bunch of additional lights).
And that's without closing potentially CPU hogging applications like browsers and such.
#215 Posted 16 August 2017 - 03:57 PM
Mblackwell, on 16 August 2017 - 01:39 PM, said:
I did not say that, I only said that there are clear benefits to DX11 support that could shed some light on how to move things forward in OpenGL. There is nothing wrong with having multiple renderers from a user perspective.
Quote
Please: benchmarks. Polymer goes to anything between 30 and 120 on my rig and in the case of classic I get around 100 too... the problem is that in WT's classic mode I have around 200.
#216 Posted 16 August 2017 - 05:11 PM
#217 Posted 16 August 2017 - 05:47 PM
Polymer is another matter since it is much further entrenched into GL, what with the use of several dozen extensions (IIRC) and plenty of GLSL shaders.
#218 Posted 16 August 2017 - 11:50 PM
Zaxx, on 16 August 2017 - 03:57 PM, said:
Please: benchmarks. Polymer goes to anything between 30 and 120 on my rig and in the case of classic I get around 100 too... the problem is that in WT's classic mode I have around 200.
Yes, but what resolution? And what hardware for that matter? And i walked around the entire map and told you the general framerate, would you say you did the same?
#219 Posted 17 August 2017 - 02:05 AM
Mblackwell, on 16 August 2017 - 11:50 PM, said:
Here's the EDuke32 classic renderer:
https://www.youtube....h?v=7qgy6Gmk4E8
And here's World Tour running in classic mode at the same 1080p resolution:
https://www.youtube....h?v=fIxEqoA6Cro
As you can see I have sometime double the fps in World Tour.
Hardware: GTX 970, i7 2600, 16 gigs of RAM, Windows 10 64 bit.
This post has been edited by Zaxx: 17 August 2017 - 02:05 AM
#220 Posted 17 August 2017 - 10:19 AM
This is World Tour's performance on Hollywood Holocaust with the true 3D rendering and ambient occlusion features turned on:
https://www.youtube....h?v=Tu0rNI6jkew
And this is EDuke32 Polymer running World Tour with normal maps turned on:
https://www.youtube....h?v=3D9RuwhGwxY
Running just Polymer with the base game would of course show a lot better results but I see no point in that in a comparison with World Tour since WT does manage to use normal maps with great performance. I also did a quick playthrough of the first level of Alien World Order just to show the real struggle of Polymer. Only dropping below 60 fps a few times on one of the simplest maps of the game (Hollywood Holocaust) is not a huge accomplishment.
This post has been edited by Zaxx: 17 August 2017 - 10:21 AM
#221 Posted 17 August 2017 - 12:16 PM
The fact is that it runs well enough under the right conditions to be useful for many projects. And yes, we would all like to see a Polymer optimization or PolymerNG.
#222 Posted 17 August 2017 - 12:39 PM
Trooper Dan, on 17 August 2017 - 12:16 PM, said:
Well sure, if you're running the base game on polymer then it runs fine... but what's the point? The base game barely uses any of the features Polymer was designed for.
Let's say you release a game, a fast paced FPS where high graphical settings destroy any hardware on the market to the point where nothing can run your game - the fast paced FPS - at a high framerate. On low it runs great but looks nothing like what the screenshots advertised. You'd hate the guts of that product.
Polymer shits itself the moment you throw anything advanced at it while WT's true 3D rendering setting enriches the visuals in a tasteful, colourful way that very much fits the original visual aesthetic of the game. And it does this without crippling performance.
This post has been edited by Zaxx: 17 August 2017 - 12:41 PM
#223 Posted 17 August 2017 - 12:48 PM
Zaxx, on 17 August 2017 - 12:39 PM, said:
Let's say you release a game, a fast paced FPS where high graphical settings destroy any hardware on the market to the point where nothing can run your game - the fast paced FPS - at a high framerate. On low it runs great but looks nothing like what the screenshots advertised. You'd hate the guts of that product.
Polymer shits itself the moment you throw anything advanced at it while WT's true 3D rendering setting enriches the visuals in a tasteful, colourful way that very much fits the original visual aesthetic of the game. And it does this without crippling performance.
You are presenting a false dichotomy: either use Polymer for the base game and don't use most of its features and it runs fine, or push it hard and it runs like crap. What I'm saying is that there is a middle-ground, where you can enjoy a lot of polymer features and it still runs pretty well. There are released projects that are good examples.
As for WT: gearbox is not releasing the source code so it's a moot point.
#224 Posted 17 August 2017 - 12:49 PM
Zaxx, on 17 August 2017 - 02:05 AM, said:
https://www.youtube....h?v=7qgy6Gmk4E8
And here's World Tour running in classic mode at the same 1080p resolution:
https://www.youtube....h?v=fIxEqoA6Cro
As you can see I have sometime double the fps in World Tour.
Hardware: GTX 970, i7 2600, 16 gigs of RAM, Windows 10 64 bit.
dnrate cheat or r_showfps 1 in the console
Not FRAPS.
#225 Posted 17 August 2017 - 12:53 PM
Mblackwell, on 17 August 2017 - 12:49 PM, said:
Not FRAPS.
Sorry, won't make a new recording just because you don't like FRAPS. It was recorded with ShadowPlay, FRAPS was only used to display the framerate (I like it: nice big yellow numbers with a great update rate, the update rate of EDuke32's built in fps counter is quite poor in comparison) so the performance loss on a machine such as this is negligable. In fact I've yet to see performance loss because of FRAPS.
This post has been edited by Zaxx: 17 August 2017 - 01:01 PM
#226 Posted 17 August 2017 - 01:01 PM
#227 Posted 17 August 2017 - 01:04 PM
Mblackwell, on 17 August 2017 - 01:01 PM, said:
These are real numbers, directly from the game.
Also please tell me how should I monitor the framerate of World Tour without the use of external tools such as FRAPS, RTSS or the Steam Overlay (ShadowPlay does not record the last one as far as I know). Using the same application for both versions of the game provides the 100% fair comparison.
This post has been edited by Zaxx: 17 August 2017 - 01:05 PM
#228 Posted 17 August 2017 - 01:05 PM
It's been a cheat since Duke 1.3d
I already said that.
And no it's not a fair comparison if WT is doing things like rendering hud text on the GPU and outputting as a texture.
Edit: More explanation below.
#229 Posted 17 August 2017 - 01:06 PM
#230 Posted 17 August 2017 - 01:14 PM
TerminX, on 17 August 2017 - 01:06 PM, said:
I checked it out by setting up the FRAPS counter on one side and the built in fps monitor on the other: the numbers match.
#231 Posted 17 August 2017 - 01:18 PM
Zaxx, on 17 August 2017 - 01:14 PM, said:
FPS number alone isn't enough though and FRAPS isn't made for benchmark purpose and is very limited. Check this example out to get the idea how people are making benchmarks:
This post has been edited by Sledgehammer: 17 August 2017 - 01:24 PM
#232 Posted 17 August 2017 - 01:24 PM
Sledgehammer, on 17 August 2017 - 01:18 PM, said:
I don't want to sound like an elitist but that's a really old card and WT has significantly higher system requirements than EDuke32. WT seems like a fairly low budget "console port" and that stuff is optimized for newer hardware. WT runs like shit on my old laptop too while EDuke32 flies on it.
Edit: As for using Afterburner / RTSS: I only wanted to show raw performance, I did not care about frame pacing and such.
This post has been edited by Zaxx: 17 August 2017 - 01:26 PM
#233 Posted 17 August 2017 - 01:27 PM
Trooper Dan, on 17 August 2017 - 12:48 PM, said:
As for WT: gearbox is not releasing the source code so it's a moot point.
That's sad to hear since in his conversation with the community Randy pointed to the possibility of the contrary. Fuck 'em if that's true.
#234 Posted 17 August 2017 - 01:35 PM
Zaxx, on 17 August 2017 - 01:24 PM, said:
Yeah, and I have even much older than that.
But anyway, looks like I misunderstood you. I thought you were arguing about polymer performance and new WT render.
#235 Posted 17 August 2017 - 01:39 PM
Trooper Dan, on 17 August 2017 - 12:48 PM, said:
Is this official? Or a speculation? In the first case, another sad story...
#236 Posted 17 August 2017 - 01:41 PM
Zaxx, on 17 August 2017 - 01:24 PM, said:
But weren't you just arguing that WT had better performance? (also, GTS450 is still around 4 times faster than GT610 which is recommended for WT on the Steam storefront)
#237 Posted 17 August 2017 - 01:51 PM
Phredreeke, on 17 August 2017 - 01:41 PM, said:
And it has better performance (except when compared to polymost)... on newer hardware.
As for the system requirements of WT I don't think you can believe those specs for a game like this. What is "recommended" when it comes to the remastered version of Duke 3D? 30 fps? 60 fps? One of those two. At what resolution? Only Gearbox knows but if I'd have to guess I'd say those are just random general specs made up by looking at the game's performance and dialing it back.
This post has been edited by Zaxx: 17 August 2017 - 01:52 PM
#238 Posted 18 August 2017 - 06:59 AM
(even if the built-in may be not accurate by bugs or something but unlikely in most cases)
I always using Afterburner/RTSS overlay when play games, but I never say its fps counter would be superior, especifically when the game have built-in one...
Whatever you believe or not, using anything 3rd party software does caused some performance impacts, maybe just very, very minor.
This post has been edited by Player Lin: 18 August 2017 - 07:01 AM
#239 Posted 18 August 2017 - 09:36 AM
Player Lin, on 18 August 2017 - 06:59 AM, said:
(even if the built-in may be not accurate by bugs or something but unlikely in most cases)
I always using Afterburner/RTSS overlay when play games, but I never say its fps counter would be superior, especifically when the game have built-in one...
Whatever you believe or not, using anything 3rd party software does caused some performance impacts, maybe just very, very minor.
After seeing what it does I'd say dnrate is just as good as FRAPS but EDuke32's built in counter is just not great.
Anyway I started benchmarking the two versions in an attempt to pinpoint why WT runs better. I suspected multi threading but the game's not great at all when it comes to that, there is some of it but it mostly relies on one core.