Testing a new rendering backend (New tests needed)

Here, developers communicate stuff that does not go onto the main News section or the front page of the site.
[Dev Blog] [Development Builds] [Git Change Log] [GZDoom Github Repo]

Moderator: GZDoom Developers

Re: Testing a new rendering backend (New tests needed)

Postby MartinHowe » Wed Mar 31, 2021 2:21 pm

Ironically, I am a kinda tinkerer, or soon will be.

My plan to buy a much more modern computer than I have now, for day to day gaming, will still go ahead; but soon I will also be the proud owner of an AlphaStation with dual EV56s and PCI (not PCIE) slots; so very limited graphics cards.

I was thinking of trying to build a simpler form of Doom on it as there ain't a hope in Hell of getting GZDoom to work on any video card that is old enough to only have PCI. While PrBoom works on AXP via OpenVMS, I still bear the scars of VAX/VMS admin and programming and do not ever want to touch VMS again with a barge pole :p So linux or a very old RC of Win2K for AXP. I doubt even GLES would be supported on any card like that and no drivers for Win2K so would have to be Linux; Gentoo is still supported for AXP but the installation will be a pig to do. I guess ZDoom 2.8.1 would be doable (pure software render) but it has no ZScript.

Still, seeing any form of Doom running on what, but for the ineptitude of Digital and the mendacity of Intel and Compaq, should have been the mainstream desktop CPU by now (well, at least on Mac) would be cool!
User avatar
MartinHowe
In space, no-one can hear you KILL an ALIEN
 
Joined: 11 Aug 2003
Location: Waveney, United Kingdom

Re: Testing a new rendering backend (New tests needed)

Postby Rachael » Wed Mar 31, 2021 2:46 pm

My initial plan with this project was to try and integrate this directly into GZDoom as a new vid_preferbackend option. And trying to get some feedback from folks with regards to how well it runs on their potato systems was a part of that, seeing if it would still be worth doing such a thing.

But seeing as how people with potato PC's are becoming more rare (as was expected in the first place), it might be better off kept as its own separate project which can be updated alongside GZDoom so long as the code continues not to cause merge conflicts. The moment it becomes more of a chore to keep this running though, where renderer improvements are so great that it no longer becomes feasible to maintain this project, then my desire to do so will go with it.

I think the Intel HD4000 result showed one thing: This certainly does have its benefits.
User avatar
Rachael
Admin
 
Joined: 13 Jan 2004
Discord: Rachael#3767
Twitch ID: madamerachelle
Github ID: madame-rachelle
Operating System: Windows 10/8.1/8/201x 64-bit
OS Test Version: No (Using Stable Public Version)
Graphics Processor: nVidia with Vulkan support

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Wed Mar 31, 2021 3:09 pm

What benefits do you see on the HD4000? The numbers were all the same, except for that distance clipped Frozen Time. This is also a GL 4 chipset running on the modern render path in LZDoom.
But what's missing here is running actual GZDoom as well to see how LZDoom stacks up to that
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby Redneckerz » Wed Mar 31, 2021 3:37 pm

Graf Zahl wrote:Can we please leave this nonsense behind? Every time the user share of old hardware gets mentioned some retro-fanatic starts with this vast miraculous group of users living beyond the internet, using ancient toaster systems and not caring for any advancements.

Believe or not, they exist. People like PhilsComputerLab, the Vogons forum, and what not. They care far less about GZDoom in general, but they will like GZDoom if it runs on their custom retro rig from 2002 and does so smoothly.

They don't account for real user share because GZDoom's intent from your end of the bargain is not that. But clearly a need exists, otherwise none of the lower end builds of ZDoom and now Beloko's work wouldn'' exist.

Graf Zahl wrote:And regarding Vogons, that's really a group of users we do not need to cater for - at all! It's a major difference to give some people who cannot afford to upgrade an option or such tinkerers who normally have better hardware at their disposal.

I never said that you should cater to it. But you also said that (in general) that if people wanted features in beyond your scope, they should add them themselves, in a way that it is either easy for you to maintain, or the person doing this feature is willing to do that maintenance for you.

Beloko's work is exactly that.

If this is too little gains for too little pays, than i reckon Beloko will just keep his own build and just change the executable name accordingly. But as Rachael says: There are benefits. Now it has to be seen how easily integrated it can be and if it is easy to maintain, i'd wager.

Its no surprise that GLESZDoom works best on potato hardware - Current mobile GPU's barring flagship phones are pretty much in the league of a HD4000 these days. Anything lower and performance should go up.*

*Atleast for now in theory until i can test this out.
User avatar
Redneckerz
To it's ports i may have seen
Spotlight Team
 
Joined: 25 Nov 2019
Discord: Redneckerz#8399
Operating System: Windows Vista/7/2008 64-bit
Graphics Processor: nVidia (Legacy GZDoom)

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Wed Mar 31, 2021 4:01 pm

Redneckerz wrote:Believe or not, they exist. People like PhilsComputerLab, the Vogons forum, and what not. They care far less about GZDoom in general, but they will like GZDoom if it runs on their custom retro rig from 2002 and does so smoothly.


Sorry, but these people are irrelevant in my book. They run old rigs for the fun of running old computers, not for using them as their daily driver. Whether they can or cannot play GZDoom on their toys if of zero consequence. If they care about playing Doom they always have the option to use their primary computer for it.

Redneckerz wrote:They don't account for real user share because GZDoom's intent from your end of the bargain is not that. But clearly a need exists, otherwise none of the lower end builds of ZDoom and now Beloko's work wouldn'' exist.


No, there is no need. If we do support for older hardware it is not for letting some tinkerers toy around with it but to allow people with no access to better hardware to play the game.


Redneckerz wrote:I never said that you should cater to it. But you also said that (in general) that if people wanted features in beyond your scope, they should add them themselves, in a way that it is either easy for you to maintain, or the person doing this feature is willing to do that maintenance for you.

Beloko's work is exactly that.



No, Beloko's work was not to service such a fringe group but to let the engine run on mobile phones with older graphics hardware.


Redneckerz wrote:If this is too little gains for too little pays, than i reckon Beloko will just keep his own build and just change the executable name accordingly. But as Rachael says: There are benefits. Now it has to be seen how easily integrated it can be and if it is easy to maintain, i'd wager.


It will remain a separate project. The changes are just too invasive to make it all work. Also, right now there are no benefits. The HD4000 is not a bit faster than the fully featured version.

Redneckerz wrote:Its no surprise that GLESZDoom works best on potato hardware - Current mobile GPU's barring flagship phones are pretty much in the league of a HD4000 these days. Anything lower and performance should go up.*


So far we have nothing that supports this. I know how poorly GL2 hardware for PC works with shaders and would not be surprised if it totally tanks compared to LZDoom which uses fixed function instead. What I can tell you outright is that my rather weak 10 year old HTC Desire handles shader based rendering better than the Geforce 6800 in that 16 year old computer I still own (no, I don't use it for gaming!) For the HTC Desire it doesn't matter whether I use GLES 1.1 with fixed function rendering or GLES 2.0 with shaders - but for the Geforce it means being playable or being a slideshow. The shaders I tested on it back then were simple colorization shaders for the invulnerability colormaps
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby simc » Wed Mar 31, 2021 4:17 pm

A bit late reply, but was it intented to run both engines in the same sized windows? For me the test-LZDoom runs in 1024x640 but GZDoom defaults to oddly measured smaller windows, such as 1008x561 window on a 1366x768 full screen and 1012x565 window on a 1680x1050 full screen.

https://i.imgur.com/adH9sKW.jpg

i3-6100u, 4GB, Intel HD 520
a6-5200, 4GB, Radeon HD 8400
also sempron 2200, 2GB, GeForce 6150se - 32-bit Vista sp1, but it couldn't run GZDoom. Only LZDoom.
You do not have the required permissions to view the files attached to this post.
User avatar
simc
 
Joined: 03 Feb 2012

Re: Testing a new rendering backend (New tests needed)

Postby Redneckerz » Wed Mar 31, 2021 5:52 pm

Graf Zahl wrote:Sorry, but these people are irrelevant in my book. They run old rigs for the fun of running old computers, not for using them as their daily driver.

I haven't said anything to the contrary. I said they exist, not that it is your target group. This bears no reptition

Graf Zahl wrote:No, there is no need. If we do support for older hardware it is not for letting some tinkerers toy around with it but to allow people with no access to better hardware to play the game.

Feel free to read anything into the quote. I am not contesting it.

Graf Zahl wrote:No, Beloko's work was not to service such a fringe group but to let the engine run on mobile phones with older graphics hardware.

That's the primary incentive yes. The secondary is mentioned already.

Graf Zahl wrote:So far we have nothing that supports this.

The theoretical conclusion is exactly that. This wouldn't hold merit otherwise. You want proof however. Ill deliver if possible
tomorrow.

Graf Zahl wrote: I know how poorly GL2 hardware for PC works with shaders and would not be surprised if it totally tanks compared to LZDoom which uses fixed function instead. What I can tell you outright is that my rather weak 10 year old HTC Desire handles shader based rendering better than the Geforce 6800 in that 16 year old computer I still own (no, I don't use it for gaming!) For the HTC Desire it doesn't matter whether I use GLES 1.1 with fixed function rendering or GLES 2.0 with shaders - but for the Geforce it means being playable or being a slideshow. The shaders I tested on it back then were simple colorization shaders for the invulnerability colormaps

The Adreno 200 was the first unified shader design from Qualcomm. It's feature set is more current-thinking than GF's 6x arch.
Ill get back with results.
User avatar
Redneckerz
To it's ports i may have seen
Spotlight Team
 
Joined: 25 Nov 2019
Discord: Redneckerz#8399
Operating System: Windows Vista/7/2008 64-bit
Graphics Processor: nVidia (Legacy GZDoom)

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Thu Apr 01, 2021 12:13 am

Redneckerz wrote:The Adreno 200 was the first unified shader design from Qualcomm. It's feature set is more current-thinking than GF's 6x arch.
Ill get back with results.


Yes. So here's the problem we're facing with this branch - it is ultimately geared towards far more modern hardware than anything that still requires OpenGL 2.x on desktop systems - which according to Steam's hardware survey are Geforce Series 6 and 7 and Intel GMA. For Geforce Series 6 I wouldn't expect any shader based approach to work and for the other two we got no samples yet.
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby XLightningStormL » Thu Apr 01, 2021 12:33 am

Graf Zahl wrote:Depends on how many "anyone"s there still are. Remember: This requires a 10+ year old computer with low end graphics hardware! When we had the last survey with GL 2 support it was somewhat around 3%. But that was over 2 years ago! How many such systems do you think are there now? We also ran some performance tests back then - Intel's GL 2 chipsets were slow enough that anything but vanilla levels at 640x480 turned out too slow.

Don't forget that such a low end port also needs to be maintained. If there is no or merely marginal interest it may simply turn out uneconomical.

Do you seriously think, most players will seriously care about a miniscule little poll beyond maybe a certain % of players, and the "GZDoom Elite"? Polls are too easy to either fraud (case in point your "greatest enemy" has stuffed plenty of online ballots for things as miniscule as taylor swift going to a school for the blind, or Dream stuffing the ballot of the Minecraft Mob Vote to vote for glow squid because his horde of 12 year olds have only 2 braincells), or be ignored, because the "silent majority" doesn't really care at the end of the day.

Graf Zahl wrote:Sorry, but these people are irrelevant in my book. They run old rigs for the fun of running old computers, not for using them as their daily driver. Whether they can or cannot play GZDoom on their toys if of zero consequence. If they care about playing Doom they always have the option to use their primary computer for it.


That's a pretty dense assumption, you are assuming that only some boomers with ancient rigs care about compatibility? What about all the players and modders from Asia, Russia and South America, especially Brazil which GZDoom has a rather large community including Cosmobyte, Endie, etc - Even He Who Shall Not Be Named, and the unfortunate reality is Brazil isn't a 1st world country like the US, Germany, UK or Australia, so a lot of these people don't have the tech outside of a mobile phone because Computers can be a complete rip off, and outright archaic by our western standards, this is why Zandronum is still popular with South American players - because it still support their systems, and not everyone has a Vulkan-ready or supporting Chipset

I understand this is GRAF ZAHL DOOM, but I think you need to look at your player base, every part of it, and stop looking at this as a commercial product for specific set of people, you are already on "top" your kind of thinking is seriously going to bite you in the butt, because at some point, a more competent, fresh (maybe even actually sociable) programmer is going to raid your pantry and take over your house, and GZDoom will just be another "historical artefact of yesteryear" all because he let negligence to certain communities take full control, and said communities ran off and blew up another Doom Sourceport.

Supporting those people means they'll come to GZDoom, which means more clout for you...How is that a bad thing? Doom Engines made specifically for certain groups of people never really succeed.
User avatar
XLightningStormL
Former DUSK Wiki Master
 
Joined: 09 May 2016
Discord: XLightningStormL#7461
Twitch ID: XLightmingStormL

Re: Testing a new rendering backend (New tests needed)

Postby inkoalawetrust » Thu Apr 01, 2021 1:40 am

Alright I was able to run the benchmark this time since you also implemented a 32-Bit version for LZDoom. Here are some of the specs of my computer.

CPU: Intel Pentium E5300
GPU: Intel Q45/Q43 Express Chipset
RAM: 2GB DDR2
OpenGL Version: 2.1

As you can tell from above, this computer is pretty ancient and probably the crappiest and most underpowered PC tested here so far. Which in this case is perfect for the benchmarks.

Oh also GZDoom crashed when I tried benchmarking it with Beloko's renderer, so I included the crash log/dump instead, if I were to guess it has to do with GZDoom AFAIK not having been designed with 32-Bit systems in mind for the past few years, but I'm not really a programmer so I wouldn't know.
You do not have the required permissions to view the files attached to this post.
User avatar
inkoalawetrust
 
Joined: 26 Aug 2019
Discord: inkoalawetrust #9783
Operating System: Windows Vista/7/2008 32-bit
OS Test Version: No (Using Stable Public Version)
Graphics Processor: Intel (Legacy GZDoom)

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Thu Apr 01, 2021 1:48 am

XLightningStormL wrote:stuff


Typical. Let's ignore hard numbers and waste valuable time on wishful thinking. I can assure you that our survey had an engagement ratio of 50% (half as many reports as downloads) so it definitely was representative.

Also, please stop this nonsense about the "silent majority". I really have no idea where people get this stupid notion that there's some silent, anonymous group out there whose preferences are so radically different from those which care to report back and who need some big consideration for development. You can rest assured that this group does not exist. It's just a made-up pretext to discredit customer research as a valid tool for decision making.

There's also that *little* problem, that the hardware this is about (mainly Intel GMA integrated chipsets) is so slow that it barely runs a mid-sized map at 640x480 and will tank entirely once you enable any advanced rendering feature like even dynamic lights. The cold hard truth here is that this hardware is fundamentally incapable of doing anything that requires a little bit of performance.

10 years ago I had a Geforce 8600, which currently is the lowest end of what full GZDoom still supports. Even on that card I could barely run larger maps and on anything exceeding vanilla limits I had to switch off dynamic lights. And please don't even start talking about resolutions above 1024x768.

So the math is simple for even weaker hardware. Good luck running the game at 30fps!
Do you really believe that sinking time into supporting that will magically bring in new users?
3 years ago, before GL 2 support was dropped, it had a user share of 3-4%, but natural attrition will most certainly have reduced this to less than 1% by now.
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Thu Apr 01, 2021 1:50 am

inkoalawetrust wrote:Alright I was able to run the benchmark this time since you also implemented a 32-Bit version for LZDoom. Here are some of the specs of my computer.

CPU: Intel Pentium E5300
GPU: Intel Q45/Q43 Express Chipset
RAM: 2GB DDR2
OpenGL Version: 2.1

As you can tell from above, this computer is pretty ancient and probably the crappiest and most underpowered PC tested here so far. Which in this case is perfect for the benchmarks.

Oh also GZDoom crashed when I tried benchmarking it with Beloko's renderer, so I included the crash log/dump instead, if I were to guess it has to do with GZDoom AFAIK not having been designed with 32-Bit systems in mind for the past few years, but I'm not really a programmer so I wouldn't know.


I don't think the crash is 32 bit related. It's more likely that it still tries to access a feature not available on old OpenGL. Beloko said himself that it still wasn't compatible with GL 2.1.
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby XLightningStormL » Thu Apr 01, 2021 2:21 am

Graf Zahl wrote:
XLightningStormL wrote:stuff


Typical. Let's ignore hard numbers and waste valuable time on wishful thinking. I can assure you that our survey had an engagement ratio of 50% (half as many reports as downloads) so it definitely was representative.

Also, please stop this nonsense about the "silent majority". I really have no idea where people get this stupid notion that there's some silent, anonymous group out there whose preferences are so radically different from those which care to report back and who need some big consideration for development. You can rest assured that this group does not exist. It's just a made-up pretext to discredit customer research as a valid tool for decision making.

There's also that *little* problem, that the hardware this is about (mainly Intel GMA integrated chipsets) is so slow that it barely runs a mid-sized map at 640x480 and will tank entirely once you enable any advanced rendering feature like even dynamic lights. The cold hard truth here is that this hardware is fundamentally incapable of doing anything that requires a little bit of performance.

10 years ago I had a Geforce 8600, which currently is the lowest end of what full GZDoom still supports. Even on that card I could barely run larger maps and on anything exceeding vanilla limits I had to switch off dynamic lights. And please don't even start talking about resolutions above 1024x768.

So the math is simple for even weaker hardware. Good luck running the game at 30fps!
Do you really believe that sinking time into supporting that will magically bring in new users?
3 years ago, before GL 2 support was dropped, it had a user share of 3-4%, but natural attrition will most certainly have reduced this to less than 1% by now.



Laughs in 10 FPS on E1M1 in GZ 3.5.0 on OpenGL

Half the engangement ratio = fully representative

Understandable have a great day.
User avatar
XLightningStormL
Former DUSK Wiki Master
 
Joined: 09 May 2016
Discord: XLightningStormL#7461
Twitch ID: XLightmingStormL

Re: Testing a new rendering backend (New tests needed)

Postby Graf Zahl » Thu Apr 01, 2021 2:54 am

Oh, I didn't know you had a graphics card from 2000! :twisted:

Yeah, have a nice day indeed!
User avatar
Graf Zahl
Lead GZDoom+Raze Developer
Lead GZDoom+Raze Developer
 
Joined: 19 Jul 2003
Location: Germany

Re: Testing a new rendering backend (New tests needed)

Postby _mental_ » Thu Apr 01, 2021 2:57 am

XLightningStormL wrote:Laughs in 10 FPS on E1M1 in GZ 3.5.0 on OpenGL

Half the engangement ratio = fully representative

Understandable have a great day.

Tens of thousands don't have this problem. Maybe it's something wrong with your PC.
For no apparent reason, some people think that everyone has a particular issue just because they have it.
_mental_
 
 
 
Joined: 07 Aug 2011

PreviousNext

Return to Developer Blog

Who is online

Users browsing this forum: No registered users and 1 guest