2087 replies to this topic

[pincade]

I love the latest test9, would like to use it in my cabinet as the primary exe but as I said a few days ago I must set the windowed/full screen mode to 1650 (now it's 1920) someone gives me an advice and in regedit I've changed the height and width parameters but this only affects the"official" vp920.exe and not the dx9 test.. Tips?

[jimmyfingers]

Yah, I was a bit surprised no one else had posted about the ball colliding / overlap issues either.

 

As far as stutter, it's much improved over test8a, however, I still can see at least a small difference between test9 and test7.  I imagine it's trivial to note that it is seemingly related to the ball rendering changes, which maybe a can of worms to potentially tackle.  However, it seems to be something you feel is important and / or in the right direction for VP's future but hopefully is something that will still be able to have some continued improvement on the stutter (or blur, as it can appear as, when stutter get's small enough yet still present).  With the depth you went into using GPUview maybe there is more to be discovered or unlocked and also with other bottlenecks in general with VP and all it's underlying functioning involving the various hardware components.  

 

On an improvement side, I had noticed a positive aspect to pre-test9 you had me test with that made it better seemingly then any other previous version and that was in the realm of the spikes in max frame times being not as high or as frequently noticeable (still need to test more with test9 but I believe the same will hold true).  It seemed that other parts of the game play were levelled out and major table events were not having as much as an effect on the draw time variance.  For that aspect it seems like it's yielded more consistency but still the smoothest moments of test7 are still better it appears then the smoothest moments of test9, again on a small but still perceivable scale.

 

I personally, still like the clarity / smoothness of test7 over even this latest build or the previous intermediate version you had me test with even if test7 was an odd (VPish) way of doing things.  I noticed that the FPS in test9 is back at one it seemed to be previously (as opposed to the intermediate test version I had) so that is good, however, it also now shows that it wasn't the FPS alone that had added that slight bit more stutter / blur vs. test7.  But it's very close now and if the last bit of stutter can be refined and in a way that has the program working in a more graspable / defined fashion yet yield the previous DX9 smoothness (test7 and before) then that would seem to be a win-win.  I don't quite feel we're there yet though and have some more testing to do that may also help identify some areas to check out further.  More testing on my overclocked 76Hz monitor, which helps to show these variances, and cross-comparison with my 2ms development monitor as well as DLP projector will be beneficial and maybe help me highlight some more instances / ways to isolate and confirm some of this.

 

I did a quick test today using the new frame limiter function and purposely without FXAA enabled to see the impact of lower FPS rates (300 for testing) but without FXAA in the mix and I could still see moments where things went out of sync and stutter was witnessed (on my mini-cab / highest-end testing platform).  This was just using the VP default table and was something I think is a concern considering it was that basic of a table (no VPM) and FXAA was off yet periodically the ball would stutter (maybe once every minute or two for about 3 to 4 seconds).  I need to test that a bit more but even now I would suggest checking out that combination of settings yourself as I believe you will see it too (it's just time consuming and eye straining to focus highly while banging the ball around on that test table and wait for the one in every couple minutes slight stutter).  It's there though in that regard for sure at least on my one system and I will confirm more about this more simplified / non-FXAA way of testing for stutter from 1 or 2 other systems I have.  Wanted to mention it here quickly along with these other thoughts.

 

Still, as always, huge appreciation for your work, time, and dedication to this overall process and specifically in getting VP into it's potentially best playable state ever experienced since it was created. 

[mukuste]

I love the latest test9, would like to use it in my cabinet as the primary exe but as I said a few days ago I must set the windowed/full screen mode to 1650 (now it's 1920) someone gives me an advice and in regedit I've changed the height and width parameters but this only affects the"official" vp920.exe and not the dx9 test.. Tips?

 

I wouldn't recommend putting this version on your cab because of the multiball bug. Otherwise, I don't know, but if you put the proper Width/Height values into HKCU\Software\Visual Pinball\Player, then it should pick them up. Just make sure not to open the Video Settings dialog as that will reset them most likely.

[Shadowsclassic]

Just FYI, on my older PC with the 1 meg video card.

 

Monster Bash PC Killer version, 16fps average With anoying but playable stutor.

Monster Bash PC Killer version with SweetFX 117fps average smooth as silk perfect game play.

 

What manor of sorcery is this?????

[mukuste]

Just FYI, on my older PC with the 1 meg video card.

 

Monster Bash PC Killer version, 16fps average With anoying but playable stutor.

Monster Bash PC Killer version with SweetFX 117fps average smooth as silk perfect game play.

 

What manor of sorcery is this?????

 

Uh...?

 

Are you sure these were both run with the same VP version? 16 fps vs 117 fps?

 

Also I don't think you have a videocard with 1 MB of memory. What card is it, what's your hardware and OS?

[Arcade4]

Test 8a is running so good right now, I think I will wait on the next release this time. 

[Shadowsclassic]

 

Just FYI, on my older PC with the 1 meg video card.

 

Monster Bash PC Killer version, 16fps average With anoying but playable stutor.

Monster Bash PC Killer version with SweetFX 117fps average smooth as silk perfect game play.

 

What manor of sorcery is this?????

 

Uh...?

 

Are you sure these were both run with the same VP version? 16 fps vs 117 fps?

 

Also I don't think you have a videocard with 1 MB of memory. What card is it, what's your hardware and OS?

 

Okay ran it again.

The first time is using the latest build vpinball 9.2.1 r932 fps 16 used graphic memory 228.36

The second was using VP9_DX9_test9, 98.5 avr. 96.7 Draw calls 375 state change 2796 texture 175 ish  (lower fps than before maybe do to the fact that it was a frest boot before and my system has been of several hours now.  Still big difference.

 

I also tried it with South Park.

with r932  I can not get a fps as the game will always crash when it starts unless I turn textures down until it looks like b bit graphics (so I could only play this in my cab, never this desktop)

Under your wonderous version VP9_DX9_test9, South Park actually runs, Smoothly, at 98.6fps 91.9avr.

 

You are right about my vid card its only a 512mb Nvidia Geforce 9800 GTX+

Intel Pentium D 3.40 Ghz

2 Gig memory

Windows 7 ultimate 32bit

 

Other tables I have tried have not had such a significant difference, but they all have ran better.

Man mukuste, I'm really glad you chose to use your powers for good and not Evil!

 

Thank you Thank you!

[mukuste]

Aaah ok, I didn't realize one of those was VP 9.2.1 and the other was the DX9 version, that clears things up

 

The way you wrote it I thought SweetFX made your VP ten times faster.

Edited by mukuste, 09 March 2014 - 01:27 AM.

[Shadowsclassic]

Aaah ok, I didn't realize one of those was VP 9.2.1 and the other was the DX9 version, that clears things up

 

The way you wrote it I thought SweetFX made your VP ten times faster.

Oh sorry, yeah I see that now.  No, just comparing the current version with the future.

I can now even play Black Rose with a B2S backglass on this older PC. For some reason BR would always crash if I enabled DirectB2S.  I could only run it with UVP backglasses since it was first released.

[vampirolatino2]

Tested version test9 and is getting better and better, still Elvyra and the Party Monters table have missing top 2 monsters and I found that the apron are more darker?

 

thanks everyone!

[xzotic]

Bring on Test9, If stutter is gone its gonna be my regular exe for daily use.  Weekend maybe?  Hate to ask its just so exciting I can't help myself. Sorry

 

I'll be looking forward to this also. No point me testing the current version until the stutters have been fixed as it's the #1 requirement for smooth and realistic gameplay. 9.2.1 is the benchmark to beat - super smooth. Dx7 build still rules!

[StevOz]

Test 9 does not work at all here test 8a was much better improving frame rate by up to 20%, whilst test9 takes forever to render the table and has the worse emreel problems and graphic decals will not render over lights on the playfield. Yeah just the most minimalist of original tables tested, The Gyruss.

Edited by StevOz, 09 March 2014 - 07:42 AM.

[mukuste]

 

Bring on Test9, If stutter is gone its gonna be my regular exe for daily use.  Weekend maybe?  Hate to ask its just so exciting I can't help myself. Sorry

 

I'll be looking forward to this also. No point me testing the current version until the stutters have been fixed as it's the #1 requirement for smooth and realistic gameplay. 9.2.1 is the benchmark to beat - super smooth. Dx7 build still rules!

 

 

Test9 specifically has changes to improve the stutter, didn't you read the post? Test it and tell me if you see any improvement. How do you expect the stutter to be fixed without providing test feedback.

Test 9 does not work at all here test 8a was much better improving frame rate by up to 20%, whilst test9 takes forever to render the table and has the worse emreel problems and graphic decals will not render over lights on the playfield. Yeah just the most minimalist of original tables tested, The Gyruss.

 

Please be more specific. What does it mean, "takes forever to render"? What EM reel problems?

 

How can I fix the problems if you don't tell me what they are?

[StevOz]

Test9 screenshot, lights obviously render over emreels, graphic decals over lights and kicker holes also incorrect. flippers also have that funky look again.

 

Actually the render time was not that bad, exit time was though.

Attached Files

•  test9.jpg   115.68KB   9 downloads

Edited by StevOz, 09 March 2014 - 08:16 AM.

[mukuste]

Test9 screenshot, lights obviously render over emreels, graphic decals over lights and kicker holes also incorrect. flippers also have that funky look again.

 

Actually the render time was not that bad, exit time was though.

 

Are you sure this is with Test9? I haven't seen the "kicker through objects" issue since much earlier versions. Same for the flipper glitches. I think you inadvertently ran an old version here.

[StevOz]

Yes that was test9 here, I know I'm running marginal hardware, though here is the test8a screenshot.

 

 test8a.jpg   118.89KB   16 downloads

Edited by StevOz, 09 March 2014 - 08:29 AM.

[teppotee]

 

 

No point me testing the current version until the stutters have been fixed as it's the #1 requirement for smooth and realistic gameplay. 9.2.1 is the benchmark to beat - super smooth. Dx7 build still rules!

Just wondering what purpose do comments like this serve in a DEVELOPMENT thread for dx9? No value / information / help and only de-motivating to developmers. Just my 2 cents...

 

Anyway quick observations about performance between test versions 7,8,9

 

Test9 is really close to test 7 now (which was the smoothest so far). 

 

With test8 I got max frame times of 16-25 (for a table which can not be mentioned here) and with test9 the max frame time remained around 7-8. So a big improvement there (and easy to notice that there is much less stutter). Also many drawing order issues are gone so going definitely to the right direction. These results on my laptop with i5, 6gb, GTX570M and win7 64.bit. In my cab with GTX580 I haven't noticed stutter even with test8a.

 

Only small percentage of tables (well... at least of the one's I'm playing) uses reels or decals so that's probably why I haven't noticed any of those issues.

 

I hope the ball collision bug in test9 can be fixed soon. (even if there aren't that many other changes to be released).

 

Thanks again for your work!

[StevOz]

Test9 screenshot, lights obviously render over emreels, graphic decals over lights and kicker holes also incorrect. flippers also have that funky look again.

 

Actually the render time was not that bad, exit time was though.

 

Actually if you flick between those 2 attached images you can see obvious differences in the rendering of the lights and shadows.

[mukuste]

 

I hope the ball collision bug in test9 can be fixed soon. (even if there aren't that many other changes to be released).

 

Thanks for your feedback on the stutter, it's valuable.

Toxie has provided a fix for the multiball issue, so I'll update the test version.

 

Everyone: if you want to get rid of the multiball problem, head to the Test9 post and download Test9a:

http://www.vpforums....=57#entry258041

[kiwi]

@ Test9a

 

The original plunger is messed up, this happens in my table not yet released.
If I change the order and put the plunger in front, the fault disappears.
I'll see if I find the same problem in another table released.

 

 

Thanks

 

Max