3889 replies to this topic

[The Loafer]

Big performance difference, AO should be set off by default IMHO.

I don't know about hunk of junk, but saying everyone needs to upgrade is not accurate either. The only thing we know for sure is that vp10 is still in beta and therefore it's always possible some performance optimization are still coming. Another thing too is just how much authors use some of these features. I mean there are vp7/8/9 tables that were released that brought many a hardware to its knees... At least until VP9.9.0 came out and breathes new life to mediocre hardware !

I think people forget that a 6 year old $700 GPU probably performs worse than a $100 current GPU because of better directx 9 driver support. I think instead of words like hunk of junk or power hardware, the word "current" is probably more accurate. Just an opinion of course

[The Loafer]

We can't tell yet but we try to keep it compatible. If the changes are too deep/heavy we might break the compatibility.

I don't think most of us will have a problem if there is a compatibility problem. After all, VP 10 in in beta, and changes are expected between beta versions. 

 
@Praticedummy, It could be an issue the dev teams plan to release 10.0 Final and then work on the rendering engine and come out with a 10.1 version which could break 10.0 tables. I dont have an issue with this approach as we need to continue to enhance vp and not worry about compatibility.Who knows maybe it wont cause any compatibility issues, but if it does the table authors can always adapt their tables to the new version.

I think this would always depend on exactly how much work is entailed In having a VP10.0 author update a table for VP10.1. Some authors have put in hundreds, not dozens, but hundred of hours into their VP recreations. Using the suffix .1 usually indicates compatibility, so if moving up to dx11 is a next step and it causes big changes that necessitates dozens of hours for each table to be updated, not the end of the world, but that should become its own release (VPXI), not an incremental release from VP10 to VP10.1.

you are right though, at some point, new development certainly leads to some compatibility sacrifices, this probably can't be helped. Just hope it's not so soon because if the Devs ever say "ok it's sooner rather than later to go to dx11" then I would advise holding off on vp10 final release and incorporating dx11 in that. No one would want authors to think they have wasted their time if lets say there is potential for a sooner rather than later release. This is NOT A REQUEST to delay vp10 nor is this a request to add dx11 or drop it entirely, I don't care either way, IMHO there is much to enjoy as is. I am just saying "think about this before going either way.

I am not an author and surely some won't agree with what I say but ... Am I wrong?

[bent98]

@loafer - I dont think they are migrating to DX11. They are staying with DX9 just migrating a few things on rendering engine that will help driver based AA. When the time comes to goto a new Direct X it should be 12 not 11.

[fuzzel]

Hehe right. DX11 is nice and has some cool new features but when we would port VP to use DX11 say bye-bye to WinXP support I'm happy with DX9 at the moment and it has enough features for a good looking pinball simulation. The only problem is that the core design of the engine is really old. We improved the engine a lot but some fundamential behavior hasn't been changed yet (read static vs dynamic rendering of elements). It's pretty easy to change the engine to render everything each frame but you will get a lot stutter on mid range cards even on high end cards maybe. One reason for this is that the tables aren't designed for such things. One golden rule is to put as much elements into one primitive as possible and use one texture for this object. Instead of rendering 20 separate posts with the same texture combine all posts into one mesh and render only this single primitive. The same goes for pegs, ramps, walls and so on as long as they are not animated.  Midrange gfx cards have enough bandwidth to render even 20.000+ polygons in one render call but the performance will drop if you have to switch textures or render states (transparent/non-transparent) or vertex buffers.

[javier1515]

and... be prepared for a major hardware update, by the way...

I'll stick to VP 9.xx.

pinball for everybody, not for a a few "hardware power users".

thanks,

Mario,

Buenos Aires, Argentina

 

Mario como ya te contestaron algunos.. esto es como los juegos de compu, mientras mas graficos, mas compu te exige.

De todas maneras esto es una beta y si comparamos la primera que lanzaron a esta ultima ya se optimizo y se seguira optimizando mas y mas.

Por ejemplo la Moster Bash de Unclewilly con la primer beta no funcionaba en mi PC de escritorio y con la ultima beta ya juego casi fluido.

 

Tenemos que tener un poco de paciencia y no tener miedo a lo nuevo.

 

Un Saludo desde Villa Maria Cordoba..

[vampirolatino2]

 

and... be prepared for a major hardware update, by the way...

I'll stick to VP 9.xx.

pinball for everybody, not for a a few "hardware power users".

thanks,

Mario,

Buenos Aires, Argentina

 

Mario como ya te contestaron algunos.. esto es como los juegos de compu, mientras mas graficos, mas compu te exige.

De todas maneras esto es una beta y si comparamos la primera que lanzaron a esta ultima ya se optimizo y se seguira optimizando mas y mas.

Por ejemplo la Moster Bash de Unclewilly con la primer beta no funcionaba en mi PC de escritorio y con la ultima beta ya juego casi fluido.

 

Tenemos que tener un poco de paciencia y no tener miedo a lo nuevo.

 

Un Saludo desde Villa Maria Cordoba..

 

 

Asi es, oh y en enero estuve por Mendoza, Argentina .... ME ENCANTO TODO!!!

Saludos para ambos!

[bent98]

Hehe right. DX11 is nice and has some cool new features but when we would port VP to use DX11 say bye-bye to WinXP support I'm happy with DX9 at the moment and it has enough features for a good looking pinball simulation. The only problem is that the core design of the engine is really old. We improved the engine a lot but some fundamential behavior hasn't been changed yet (read static vs dynamic rendering of elements). It's pretty easy to change the engine to render everything each frame but you will get a lot stutter on mid range cards even on high end cards maybe. One reason for this is that the tables aren't designed for such things. One golden rule is to put as much elements into one primitive as possible and use one texture for this object. Instead of rendering 20 separate posts with the same texture combine all posts into one mesh and render only this single primitive. The same goes for pegs, ramps, walls and so on as long as they are not animated.  Midrange gfx cards have enough bandwidth to render even 20.000+ polygons in one render call but the performance will drop if you have to switch textures or render states (transparent/non-transparent) or vertex buffers.

 

Progress cost $$$. Need to upgrade video card to play VP 10 so be it. To be honest you can pick up a GTX 970 4GB for $309 now,

[Practicedummy]

Big performance difference, AO should be set off by default IMHO.

I don't know about hunk of junk, but saying everyone needs to upgrade is not accurate either. The only thing we know for sure is that vp10 is still in beta and therefore it's always possible some performance optimization are still coming. Another thing too is just how much authors use some of these features. I mean there are vp7/8/9 tables that were released that brought many a hardware to its knees... At least until VP9.9.0 came out and breathes new life to mediocre hardware !

I think people forget that a 6 year old $700 GPU probably performs worse than a $100 current GPU because of better directx 9 driver support. I think instead of words like hunk of junk or power hardware, the word "current" is probably more accurate. Just an opinion of course

Scapino's tables come to mind. Before I went high end on my computers, his tables would run at slideshow speeds on me. Back then, you practically needed a 4 GHz CPU to run them.

[bent98]

 

Big performance difference, AO should be set off by default IMHO.

I don't know about hunk of junk, but saying everyone needs to upgrade is not accurate either. The only thing we know for sure is that vp10 is still in beta and therefore it's always possible some performance optimization are still coming. Another thing too is just how much authors use some of these features. I mean there are vp7/8/9 tables that were released that brought many a hardware to its knees... At least until VP9.9.0 came out and breathes new life to mediocre hardware !

I think people forget that a 6 year old $700 GPU probably performs worse than a $100 current GPU because of better directx 9 driver support. I think instead of words like hunk of junk or power hardware, the word "current" is probably more accurate. Just an opinion of course

Scapino's tables come to mind. Before I went high end on my computers, his tables would run at slideshow speeds on me. Back then, you practically needed a 4 GHz CPU to run them.

 

 

Thats because he was reel happy.

[Argo]

I agree with bent98, the next step should be dx12! This will bring a massive performance improvement.

[bent98]

And who cares about win xp compatibility

[Shockman]

I have things set in the preference menu. In the table menu which is set to override those, everything displays the same value except AO which replaces the zero with a minus 1. I know I don't have it, I was just unsure if the code is running or not. For all I know, because it don't work with my gpu I can not even set it, not even from default 1, to zero. That is how it seems.

[Talantyyr]

I agree with dx12. IF you're going to upgrade to a new dx version anytime, it should be dx12. It provides huge fps boosts compared to dx11 and this would definately be worth the trouble loosing xp support or upgrading the table.

Btw, windows xp was released 2001 and is 14 years old. It's not supported by microsoft any more and it's slow. I can't imagine a single reason why people would wanto to stick with xp.
I'm going to use Win8.1 for my cab, as it's the newest windows and i would even upgrade to windows 10 (it's free!) if everything works fine. I've tested all pinball programs on my main pc (windows 8) and had no problems at all.

[toxie]

Just to get this out of your heads yet again: Just switching to DX10,11 or 12 will bring no performance wonders (yet).

 

The current performance 'issues' are in the engine itself (and to some extent also how most authors design their tables), as Fuzzel already pointed out. GFX cards want to have huge chunks of work without any interaction at all with the CPU (if possible). Currently though VP pushes multiple 100s of rather tiny objects with a lot of state changes to the card, which could mostly be optimized of course (either semi-automatically by the engine or to some extent already by the authors).

 

One should also never forget that current high-end game engines are not just fast because the engine itself is so super optimized and based on a recent DX version, its also to a rather large part caused by the content pushed into it and quite some pre-processing. Thats one of the reasons why 100s of educated artists have to work on these titles and tweak dozens of knobs in an advanced UI to design the levels and optimize the content for the engine/hardware.

 

VP shouldn't go that route though IMHO, it should do stuff as automatic as possible to ease the use for non-professional/hobbyist artists. Its not impossible, but its also a lot of work, and stating the obvious again, we do this all in our (limited) spare time. So we can either optimize the engine to death (implementing only a basic set of actual features) or do a practical compromise of actual features and optimizations (like IMHO we did for VP10). But i'm pretty sure we can squeeze out some more performance for 10.1.

 

 

EDIT: as for the default of AO: its initialized to off already

Edited by toxie, 17 March 2015 - 08:16 AM.

[alessio]

VP10 not is compatible with old VP9 tables again?

Edited by alessio, 17 March 2015 - 10:28 AM.

[fuzzel]

Yes VP10 breaks with a lot of old legacy stuff used in VP9/8. You can load VP9 tables into VP10 but they won't play without errors or strange behavior.

[alessio]

I have test Theatre of Magic but error "stopskake" and "Pipos"

Script Flash not compatible.......

Is only my problem?

Edited by alessio, 17 March 2015 - 01:08 PM.

[fuzzel]

You need a new VP10 table of Theater of Magic.

[Shockman]

Maybe after the release of 10, between the time you are gathering a broader block of feedback and starting the next major revision, we together, can get a comprehensive guide to all the optimization techniques. It could include things like what you mentioned about separate posts and pegs, etc. being built from one primitive, something I didn't even know was possible, making just a couple, if there is one area with a whole lot of lights, be able to be reflected by the ball, and that kind of thing.

 

I'm not one of the experts, but I suspect the ball , as dynamic as it is now, rolling, using much more physics, and reflecting light, and maybe even the unfortunate constant changing of shape, would account for the lions' share of the fps difference.

 

I can Vsync constant to 60 on most of the testing and beta tables with the more conservative post processing aa effects, with my fair cpu and sub mediocre gpu, so I don't consider VPX a monster even now. I seem to be at a point where I would have a remainder to use for higher quality post process, with a doable level of optimization with the program, whether it in the code, and/or building technique, and most certainly with a gpu purchase.

 

.

[toxie]

Balls are pretty much for free, unless you maybe play A13 multiball.. 

 

But the main performance drainers are usually insanely detailed (animated) primitive objects, and all kinds of light effects, so basically everything that can potentially be animated and dynamic, and is most of the time visible.

Then, based on this, the next rule of thumb is: The more of these objects you have, the lower the performance will be (roughly).

 

You can use the output of the F11 debug vis as guidelines:

The amount of drawcalls is the amount of objects drawn each frame (lower = better).

State, texture and parameter changes you cannot really influence (unless you wanna go insane, i guess ), but basically a simplified rule here is: The fewer images you have on the table, and the fewer materials you use on the table, the lower these numbers should be (which again is good, but not as important as all the other stuff mentioned above).