Jump to content



Photo
* * * * - 10 votes

The VP 10.4 beta thread


  • Please log in to reply
740 replies to this topic

#61 nFozzy

nFozzy

    Pinball Fan

  • Members
  • PipPipPipPip
  • 553 posts

  • Flag: ---------

  • Favorite Pinball: Pinbot

Posted 11 August 2017 - 07:18 AM

I'm trying not to use any static elements anymore, but aliasing is still a big problem. JP makes a good point that high contrast env lights aren't helping - it's the only way to get the metal on the habit rails to pop, but it also creates those really hard lines that are jaggy as hell. Maybe there's room for improvement in the materials?



#62 toxie

toxie

    VPF Veteran

  • VP Dev Team
  • PipPipPipPipPipPip
  • 5,734 posts
  • Location:berlin, germany

  • Flag: Germany

  • Favorite Pinball: AFM

Posted 11 August 2017 - 08:09 AM

Let me clear some things up here about anti-aliasing(AA)/killing jaggies (as its actually one of my expertises in real life ;))..

 

First of all, lets analyze where jaggies can come from in VP:

a) geometry: everything that is very detailed and has lots of variation (like wireframe ramps or high poly primitives,toys,etc), but it can even be as simple as a non-pixel aligned flipper, as this amount of detail has to be catched by the rasterization process of the GPU, which is usually impossible at a rate of 1x sample

b) lighting: if one uses a high dynamic range (HDR) environment image, with a lot of rather small bright spots, which also heavily interacts with a) and c)

c) normal maps: this is a rather similar problem as a) but mainly gets you jaggies usually because of b)

d) plain texture maps/images: not a huge problem nowadays, as long as you have anisotropic filtering enabled in VP

 

Now, how is this usually handled in games:

a) geometry oversampling: things like MSAA and similar techniques, which catches most effects of aboves a), but not aboves b) & c)

b) oversampling everything: DSR (or the similar brute force 4xAA in VP), which is IMHO the best one can and should do (maybe in combination with other techniques though), as it can resolve all of the above

c) postprocess AA: FXAA, (the simple versions of) SMAA, etc, which blurs all things locally where the filter thinks that jaggies are there (e.g. error prone and cannot resolve all of the above equally good)

d) temporal AA: using all kinds of information from previous frames and combining these with the current one (TXAA and similar variants), which is usually what everybody does nowadays, but is a huge effort to get implemented correctly so to not cause all kinds of artifacts

 

What does VP do:

It only can use b) and/or c) at the moment, as a) and d) would require a fully dynamic rendering engine to make this work (without getting severe headaches otherwise), whereas VP still has some things rendered statically for performance reasons

 

As for Pinball Arcade: There i would guess that they simply use something like MSAA or FXAA. Especially as with their old rendering engine (maybe even with their DX11 one) they do not get HDR effects that they'd have to handle.

Plus @DJRobX: There is of course yet another effect for your ipad3 setup: As you have a much higher pixel density on there, as when standing in front of your cab, one can even get away without doing any AA, or just some plain FXAA.

This would be similar to driving your cab at 4K res, or even higher.

 

EDIT: TL;DR version: Yes, we could do better in VPX, but it would result in a lot of pain to get it working correctly on all kinds of hardware setups/video preferences combinations. So we'd better wait for VP11 and do it like all others do nowadays, at the cost of performance when NOT doing any AA, but at a severe performance benefit (hopefully ;)) when doing anti-aliasing.


Edited by toxie, 11 August 2017 - 08:22 AM.


#63 hauntfreaks

hauntfreaks

    Real 4K Tables Since 2013

  • VIP
  • 2,522 posts
  • Location:South Jersey

  • Flag: United States of America

  • Favorite Pinball: -<3<3<3 All Real Pins



Posted 11 August 2017 - 08:36 AM

Haunt, in your screenshot above when you say extreme close-up did, you zoom the table size in VPX first then render and screenshot the result?  If that's the case, then it's not an apples to apples deal as anything zoomed in on the table via editor / view settings creates more geometry per screen size and smooths everything out all that much more (I and others have done that before with screenshots for sections of tables and with 4xAA on it can look super smooth).  But that doesn't help when we have to zoom back out to view / play the whole table.  I see you say you have brute force 4xAA on so if you're also zooming in the table view / size settings on a 1080p monitor you'll be getting potentially the equivalent of like 8k or 16k resolutions if it was rendered for the whole table. But in any case, can you clarify how exactly you zoomed in / got the "extreme close-up".  

 

I still see a lot of jaggies myself even with brute force and using the HV style view even with brute force 4xAA and 2560x1440 resolution.  I think the more angled items in desktop view and HV / FSS views show the problem of VP and lack of AA more than the FS views, which are closer to rendering a table in a more linear and straight fashion.  I can't do VP at all without brute force 4xAA on and have had to throw money and hardware at the problem as well as time on overly unoptimized tables to try and get to play properly and smoothly.  To be fair though, my setup is using a true 120hz monitor so "costs" more and even though I'm using 1440p resolutions, the HV view and sizing of the backglass and tables makes it a whole lot less demanding then a FS table would be at 1440p.

 

In all fairness, current brute force 4xAA in VP is extremely demanding on GPU power (especially the higher resolution you go), but I'm glad at least we have the option over just FXAA.  More up to date AA will be welcome and if it has to wait, but also can be done with VP 11, then at least part of the performance drop offset by going all dynamic in that version can be gained back by getting real AA working undoubtedly much more efficiently and not having to use brute force 4xAA.  

 

 

here is an example as is from the cab... only jaggies here are in the highlights of the transparent ramp.... which is possible to be there in the real thing...
also what you see also gives me smooth play

36328373562_06a40ce572_o.png

 


 26794541816_30ca1cca80_o.gif 43109635392_fc11af1a57_o.gif


#64 toxie

toxie

    VPF Veteran

  • VP Dev Team
  • PipPipPipPipPipPip
  • 5,734 posts
  • Location:berlin, germany

  • Flag: Germany

  • Favorite Pinball: AFM

Posted 11 August 2017 - 08:58 AM

What you see there is a combination of a) and b): High local geometric curvature along with the bright spots from the environment lighting.

One could tone down this in the shader/material code, but my previous experiments always lead to a decrease in crispiness of the lighting for all kinds of situations, not just for the evil ones, so i never included that.



#65 dark

dark

    3D model-man

  • VIP
  • 1,936 posts
  • Location:Toronto

  • Flag: Canada

  • Favorite Pinball: Star Wars, AbraCadaBra,MB, LAH,JPark...too many to choose!

Contributor

Posted 11 August 2017 - 10:33 AM

 

 

 

 

As toxie mentioned that has nothing to do with it. Look at Body dump or flupper tables. They dont use low res PF images and have jaggies. How about ramps? They have jaggies galore. 

 

I noticed if a table is using VPX wire ramps, I can see jaggies on the curves, but when primitive wire ramps are used , I get no jaggies....

what flupper table are you referring to??... never notice jaggies on any of his....??

here an example screenshot (extreme closeup) from my cab with an old gtx760ti and only Brute Force 4xaa on...

36255835846_4b9af99011_o.png

 

Even with the 'zoomed' view bent is right, I can see jaggies too (@1080p),  looks closely at the wireform especially the area where it overlaps the creature plastic on the left and to a lesser extent the sling plastic, the section on about a 45 degree angle has some rough edge as well.  It's made more obvious by the lighting and this is probably worse when not zoomed up like that.

 

[edit] Also I'm not sure who's BTTF build that is, but they need to use my Data East prim apron that you can grab from vpinball. :)


Edited by dark, 11 August 2017 - 10:36 AM.


#66 Slydog43

Slydog43

    Pinball Wizard

  • Platinum Supporter
  • 3,008 posts
  • Location:Hackettstown, NJ

  • Flag: United States of America

  • Favorite Pinball: Addams Family, All Williams 90's Games

Posted 11 August 2017 - 01:58 PM

Is anyone using DSR, I had trouble getting it going with my setup.  I changed the resolution in the registry, but it ran in a window instead.


Edited by Slydog43, 11 August 2017 - 01:59 PM.


#67 sliderpoint

sliderpoint

    Pinball Fan

  • Members
  • PipPipPipPip
  • 760 posts
  • Location:Spokane, WA

  • Flag: United States of America

  • Favorite Pinball: Metallica

Posted 11 August 2017 - 04:57 PM

So... is that a 'no' on the 'block transmit'...  lol.



#68 krille81

krille81

    Enthusiast

  • Members
  • PipPipPip
  • 197 posts

  • Flag: Sweden

  • Favorite Pinball: Junk Yard, White Water, Star Wars (Data East)

Posted 11 August 2017 - 05:58 PM

Is anyone using DSR, I had trouble getting it going with my setup.  I changed the resolution in the registry, but it ran in a window instead.

You have to run exlusive fullscreen mode. Then you can change to higher resolution direct in vpx


Skickat från min E2303 via Tapatalk

#69 hauntfreaks

hauntfreaks

    Real 4K Tables Since 2013

  • VIP
  • 2,522 posts
  • Location:South Jersey

  • Flag: United States of America

  • Favorite Pinball: -<3<3<3 All Real Pins



Posted 11 August 2017 - 06:01 PM

Even with the 'zoomed' view bent is right, I can see jaggies too (@1080p),  looks closely at the wireform especially the area where it overlaps the creature plastic on the left and to a lesser extent the sling plastic, the section on about a 45 degree angle has some rough edge as well.  It's made more obvious by the lighting and this is probably worse when not zoomed up like that.

 

 

LOL.... stop  ;)


 26794541816_30ca1cca80_o.gif 43109635392_fc11af1a57_o.gif


#70 Thalamus

Thalamus

    Pinball Wizard

  • Platinum Supporter
  • 4,983 posts

  • Flag: Norway

  • Favorite Pinball: GOT, Alien Star, LOTR, TOM

Posted 11 August 2017 - 06:50 PM

I agree with HF here ... wtf ?! Do you actually play pinball or start the table and stand there looking for the smallest of issues ?

We can't even look at the same picture and agree if we see anything ... so c'mon :) There surely must be other areas where the VP devs can be put to better use than this ?


From now on. I won't help anyone here at VPF. Please ask Noah why that is.


#71 hauntfreaks

hauntfreaks

    Real 4K Tables Since 2013

  • VIP
  • 2,522 posts
  • Location:South Jersey

  • Flag: United States of America

  • Favorite Pinball: -<3<3<3 All Real Pins



Posted 11 August 2017 - 06:55 PM

i'm just shocked anyone could see that microscopic amount of pixels to care.... I'm guessing they must be running 4xaa and the highest AA to blur the shit out of everything on the playfield....LOL
hey to each there own.... but like Thalamus said, these guys have way more cool stuff there working on..... 


Edited by hauntfreaks, 11 August 2017 - 06:56 PM.

 26794541816_30ca1cca80_o.gif 43109635392_fc11af1a57_o.gif


#72 toxie

toxie

    VPF Veteran

  • VP Dev Team
  • PipPipPipPipPipPip
  • 5,734 posts
  • Location:berlin, germany

  • Flag: Germany

  • Favorite Pinball: AFM

Posted 11 August 2017 - 07:32 PM

So... is that a 'no' on the 'block transmit'...  lol.

 

not so sure, that's why i didn't answer yet.. maybe you can elaborate a bit more on the exact mechanism you have in mind here?



#73 dark

dark

    3D model-man

  • VIP
  • 1,936 posts
  • Location:Toronto

  • Flag: Canada

  • Favorite Pinball: Star Wars, AbraCadaBra,MB, LAH,JPark...too many to choose!

Contributor

Posted 11 August 2017 - 08:10 PM

The pixels aren't that hard to miss, I think you guys just maybe haven't seen many modern games?  A jagged edge here and jagged edge there and suddenly the overall presentation suffers for it.  It's been discussed sufficiently enough though already, you won't see a fix for it until vp does away with static rendering....ie vp11.

 

I run 4xAA and Quality AA, it helps, but it's far from perfect.



#74 ganjafarmer

ganjafarmer

    Enthusiast

  • Members
  • PipPipPip
  • 278 posts
  • Location:Poland

  • Flag: Poland

  • Favorite Pinball: Attack from Mars

Posted 11 August 2017 - 08:35 PM

Not long ago you've been playin' VP9 and now we got VPX and some of you complain? Pinball is made for playing not to look good. But it is always nice when it looks good :)

 

I've noticed when i zoom in or zoom out (Z axis offset) while camera is moving the so called 'jaggies' are tottaly gone, most noticable on wire ramps, do you noticed that? What's causes that?

 

Again: it is possible to add ball bounce/jump posibility?


One day i'll build Vitalik statue out of used AMD gpu's.


#75 fuzzel

fuzzel

    spaghetti code

  • VP Dev Team
  • PipPipPipPipPip
  • 2,818 posts

  • Flag: Germany

  • Favorite Pinball: yes I have

Posted 11 August 2017 - 08:38 PM

Not long ago you've been playin' VP9 and now we got VPX and some of you complain? Pinball is made for playing not to look good. But it is always nice when it looks good :)
 
I've noticed when i zoom in or zoom out (Z axis offset) while camera is moving the so called 'jaggies' are tottaly gone, most noticable on wire ramps, do you noticed that? What's causes that?
 
Again: it is possible to add ball bounce/jump posibility?

Not in the near future. The physics engine is really a mess and even adding a correct kicker ball movement is more a try'n error thing

#76 toxie

toxie

    VPF Veteran

  • VP Dev Team
  • PipPipPipPipPipPip
  • 5,734 posts
  • Location:berlin, germany

  • Flag: Germany

  • Favorite Pinball: AFM

Posted 11 August 2017 - 08:40 PM

vanishing jaggies: see my explanation, it's due to geometry AND lighting. and when the camera moves, the lighting/reflections also change.

 

ball bounce/jump: i have a WIP physics mod since ages around that would (in theory) also add this (as it's full 3D and not just 2.5D physics), but it still needs way more work and testing.. the thing is that the VP physics code is still pretty fragile as it's full of heuristics, so changing stuff without breaking other things is time-consuming..



#77 sliderpoint

sliderpoint

    Pinball Fan

  • Members
  • PipPipPipPip
  • 760 posts
  • Location:Spokane, WA

  • Flag: United States of America

  • Favorite Pinball: Metallica

Posted 11 August 2017 - 09:04 PM

 

So... is that a 'no' on the 'block transmit'...  lol.

 

not so sure, that's why i didn't answer yet.. maybe you can elaborate a bit more on the exact mechanism you have in mind here?

 

 

Sure,  I have a light object that's height is setup below a prim bumper cap.  I have the transmit on at .5.  It causes the bumper cap to light up nicely.  but it also transmits to a primitive ramp that is above the bumper cap as well.  I would like to not have any of those transmit to the prim. ramp.

 

some images

The light objects:

lights.png

 

The lights on with no bumper caps:

onnobumpercaps.png

 

With bumper caps, lights Off:

bumpersoff.png

 

Now lights on:  You can see how it looks if the light transmits on to the front bumpercap.  But there is a white washing now on the prim ramp that hides the other bumpers:

bumperson.png

 

I can shoot you a link to the current WIP of this table to check out. I also had a problem with enabling/disabling a trigger that I messaged you about a while back as well.

 

-Mike



#78 toxie

toxie

    VPF Veteran

  • VP Dev Team
  • PipPipPipPipPipPip
  • 5,734 posts
  • Location:berlin, germany

  • Flag: Germany

  • Favorite Pinball: AFM

Posted 11 August 2017 - 09:11 PM

ah, i see.. that should be rather straight forward by adding a flag to primitives to disable that optionally.. i'll have a look..

 

as for the trigger: i saw that, looked at it briefly, then forgot about it again, sorry..  :/

i guess fuzzel is more the specialist there..



#79 sliderpoint

sliderpoint

    Pinball Fan

  • Members
  • PipPipPipPip
  • 760 posts
  • Location:Spokane, WA

  • Flag: United States of America

  • Favorite Pinball: Metallica

Posted 11 August 2017 - 09:19 PM

Thx for checking into it!  I'll ping Fuzzel.

 

-Mike



#80 bent98

bent98

    Pinball Fan

  • Members
  • PipPipPipPip
  • 1,077 posts
  • Location:NY

  • Flag: United States of America

  • Favorite Pinball: Roadshow, Haunted House, Safe Cracker

Posted 11 August 2017 - 09:54 PM

I agree with HF here ... wtf ?! Do you actually play pinball or start the table and stand there looking for the smallest of issues ?
We can't even look at the same picture and agree if we see anything ... so c'mon :) There surely must be other areas where the VP devs can be put to better use than this ?

It's subtle in the pic but not when you have the table displaying on a 40+ display. The tables being produced with all the 3D work is just amazing. Take a look at Jurassic park and white water WIP. The screenshot look great inside of 3D programs used to create the objects but when you import them into vp , you really can see all the jaggies, it really detracts from all the authors hard work and artistic vision they are trying to portray . It's like detailing a car and making the paint look like glass and then the chrome and trim are rusted and pitted. If you can't see that then it's time to get glasses.


I appreciate the authors and devs hard work of course. I am stating my opinion that this is a glaring issue that needs to get addressed one day.

Edited by bent98, 13 August 2017 - 12:25 AM.