22 replies to this topic

[Kernel]

After building a VP cabinet, I was never satisfied by flipper latency compared to my real pinball. On a agame based primarly on timing, flipper latency for me is absolutely the most important thing to consider for having fun.

I've made "super slow motion" videos at 960 fps (aligned them and slowed them down further by 75%) where you can clearly see the differences between VPX (VSYNC and FAST VSYNC), Pinball FX3 and a real SS pinball.

Pinball FX3 - Fast VSYNC (240 FPS) - Pre rendered frames "1".
Visual Pinball X 10.6.0 rev. 3660 - V-Sync (60FPS) - Pre rendered Frames "1" - Fast flips ON - VPINMame 3.2 rev 4723.
Visual Pinball X 10.6.0 rev. 3660 - Fast V-Sync (? FPS) - Pre rendered Frames "1" - Fast flips ON - VPINMame 3.2 rev 4723.

Test made on the same ROM controlled table.

HARDWARE SPECS:
i7 4790k @ 4.4GHz - 16GB DDR3 RAM 1866 - GTX 1080.
Corsair Gaming keyboard set on 1000Hz polling rate.
Dell monitor: Measured average total display lag "4.0ms" on TFT Central (CLASS 1).

 

HERE THE VIDEO: https://youtu.be/gQ77fBA521o (Watch at 1080p @ 60FPS)

 

 

The worst situation is the VPX 60FPS VSync on option, very notieceble flippers delay when playing.

The FAST VSync is way better (not sure if on VPX works or it's off), less than half of the LAG. The winner is PinballFX 3, very close to a real machine.

 

PS: I will never say it enough: A big thank you to the Visual Pinball/VPM devs and table makers for sharing all their amazing work.

Edited by Kernel, 17 March 2019 - 05:14 PM.

[gtxjoe]

Fastflips was created to eliminate the VPX to Vpinmame lag that existed. have you tried that setting?

Also EM VP Tables should be more responsive than ROM based tables without fastflips. Have you observed that difference in your testing?

[The Loafer]

Thanks for the test.  Are you aware if your VPM is using fastflips?  doh!  GTXJoe beat me to it!   

 

Anyway ensure fastflips is enabled, ensure you are using the latest VP/VPM and rerun your tests.

 

Also I was under the impression for GPU Fastvsync to work that you had to turn OFF the VP VSync?  I have to admit, I'm confused about those options...

 

Visual Pinball X 10.6.0 rev. 3660 - V-Sync (60FPS) - Pre rendered Frames "1".  so for this one, you have VSYNC set to 1
Visual Pinball X 10.6.0 rev. 3660 - Fast V-Sync (? FPS) - Pre rendered Frames "1".  So for this one, you have VSYNC set to o?

Edited by The Loafer, 17 March 2019 - 03:42 PM.

[Kernel]

Fast flips enabled (modified op).

I've compared rom controlled FX3 (Attack from mars) and VPX.

I'll make a video with EM tables.

Edited by Kernel, 17 March 2019 - 03:40 PM.

[Schlabber34]

When building tables i use my desktop pc with a 144hz 21:9 gaming monitor. It's really hard to go back to the cab and enjoy playing a table on the 60hz standard monitor. 4k plus 120hz is the only satisfying way to play!

Gesendet von meinem SM-G955F mit Tapatalk

[Kernel]

When building tables i use my desktop pc with a 144hz 21:9 gaming monitor. It's really hard to go back to the cab and enjoy playing a table on the 60hz standard monitor. 4k plus 120hz is the only satisfying way to play!

Gesendet von meinem SM-G955F mit Tapatalk

I felt the same 20 years ago when I've built a mame cab. After using a real 15KHz arcade monitor I wasn't able to use mame on a SVGA CRT/LCD monitor anymore.

I have no 144Hz monitor, but I can imagine how much better it is. But also running 240 FPS on my 60Hz, as you can see, it makes a huge difference in response times and it's really a game changer.

[Schlabber34]

Yeah, but any framerate number above or below 60 is causes microstutter on my cab and that's even worse!

Gesendet von meinem SM-G955F mit Tapatalk

[Kernel]

Anyway ensure fastflips is enabled, ensure you are using the latest VP/VPM and rerun your tests.

 

Also I was under the impression for GPU Fastvsync to work that you had to turn OFF the VP VSync?  I have to admit, I'm confused about those options...

 

Visual Pinball X 10.6.0 rev. 3660 - V-Sync (60FPS) - Pre rendered Frames "1".  so for this one, you have VSYNC set to 1
Visual Pinball X 10.6.0 rev. 3660 - Fast V-Sync (? FPS) - Pre rendered Frames "1".  So for this one, you have VSYNC set to o?

Fast flips enabled, using latest VPM.

With VPX VSync ON (60FPS) the setting on nvidia drivers is "application controlled" and VPX VSync set to 1.

With VPX FAST vsync (? FPS) the setting on nvidia drivers is "FAST Vsync" and VPX VSync set to 0.

Edited by Kernel, 17 March 2019 - 04:49 PM.

[Kernel]

Yeah, but any framerate number above or below 60 is causes microstutter on my cab and that's even worse!

Gesendet von meinem SM-G955F mit Tapatalk

Fast VSync is designed to avoid that behaviour if it works right. How may frames can you reach without VSync? If you can achieve at least 120FPS you should get the stutter/tearing free image with way more less LAG as shown in the video. I know that maybe the micro stutters can also be introduced by B2S backglasses. There are so many variables to consider...

Edited by Kernel, 17 March 2019 - 05:37 PM.

[Schlabber34]

I will check that on a simple table with brute force and reflections off. Never looked at F11 fps count when using fast vsync.

Gesendet von meinem SM-G955F mit Tapatalk

[ludovids]

I just now got around to watching the video. Nice work, Kernel!

 

What are the time units for the clocks shown on the video?

 

It seems to me that what the video is showing us with regard to the flippers is a combination of the response time of the machine (real or virtual) to the button press, plus the response time of the video system.

 

When considering the response time of only the video system, I think the biggest benefit of increased frame rate to the person playing a pinball table, is having a more accurate view of where the ball is on the table. With a pinball moving fast at 10 mph (16 kph), in 1/120 of a second it would travel 1.47" (3.73 cm). Even at 1/4 that speed, it's still 0.37" (0.93 cm). I think that is a very significant difference for deciding when to press a flipper button.

 

I agree, getting fast and consistent flipper response is key to being able to make the shots in pinball. And quick video frame rate is key to knowing when to press a flipper button. Fortunately, FastFlips is available to satisfy the first requirement, and fast enough hardware is available to satisfy the second.

[Kernel]

I've put the clock on the 960 fps videos just to have a reference to see the differences between softwares/configurations. After that I've slowed the video down by 75%.

You see "seconds.hundredths of a second" on a 960 FPS video, so if I'm not wrong, one second of the clock should correspond to 0.0625 seconds.

 

Having more FPS allows you to see the position of the ball "in advance" (more precise/real location) and you can also have a faster system reaction to a key press. Double vantage.

Edited by Kernel, 25 March 2019 - 11:05 AM.

[ludovids]

Is  that actually "seconds:sixtieths of a second"? I noticed the number after the ':' goes from 0 to 59.

 

If so, then the VPX-FAST-VS case of "01:38" to visible triggered state takes about 0.1 sec. Whatever the time really is, I don't think I can perceive the delay, even though it's quite a bit longer than the 0.001 to 0.002 sec it takes VPX with FastFlips to respond to a flipper button push.

 

Maybe this is part of the reason why it seems so much harder to make a good shot with a fast moving ball than with a slow moving ball, such as one that was held by a flipper then released before taking a shot. The displayed location of the slow moving ball will be more accurate.

[DJRobX]

Great video, but you are showing a sample size of 1.   I would be interested to see the times for maybe 10 attempts.   There's still delay loops in play like the USB polling interval and the display's actual refresh rate, so we need more data to calculate an average.

 

You also should test with VPX's coil ramp up value on the flippers to 0.   Coil ramp up is intentional lag.  The idea being we want some time to measure shorter presses to allow short taps to fire the ball at lower velocity, to enable flipper tricks like post passes.   I've wondered if we can eliminate that particular lag by "retroactively" dampening ball speed instead of delaying the flipper movement, but I'm not enough of an expert on VP physics to want to touch any of that.    But would be great to see what the real impact of that is. 

Edited by DJRobX, 27 March 2019 - 02:40 PM.

[blindpeser]

Never recognized any lag in VR. Is it possible to get better numbers here? Is it possible that there is a difference between keyboard and controller?

[DJRobX]

Also I see something strange in the VP-Vsync video - the flipper goes straight to the fully up position, never going through its intermediate stages as it does in the non-vsync version.   It's almost as if your system dropped frames at a very inappropriate time.  

Using your math (1 second = .0625 seconds = 62.5ms) the vsync difference is 92.5ms, but vsync should never impart more than 16.6ms of delay at 60hz. 

 

Not trying to dispute your findings, I'm just looking to make sense of the results, as this has amazing potential for researching and improving issues, just as the addition of the F11 measurement lead to fast flip improvements.

[Kernel]

Is  that actually "seconds:sixtieths of a second"? I noticed the number after the ':' goes from 0 to 59.

You're right, I'm so poor at math

 

Great video, but you are showing a sample size of 1.   I would be interested to see the times for maybe 10 attempts.   There's still delay loops in play like the USB polling interval and the display's actual refresh rate, so we need more data to calculate an average.

 

You also should test with VPX's coil ramp up value on the flippers to 0.   Coil ramp up is intentional lag.  The idea being we want some time to measure shorter presses to allow short taps to fire the ball at lower velocity, to enable flipper tricks like post passes.   I've wondered if we can eliminate that particular lag by "retroactively" dampening ball speed instead of delaying the flipper movement, but I'm not enough of an expert on VP physics to want to touch any of that.    But would be great to see what the real impact of that is. 

Good point, I have to find the time to do more tests.

 

Also I see something strange in the VP-Vsync video - the flipper goes straight to the fully up position, never going through its intermediate stages as it does in the non-vsync version.   It's almost as if your system dropped frames at a very inappropriate time.  

Using your math (1 second = .0625 seconds = 62.5ms) the vsync difference is 92.5ms, but vsync should never impart more than 16.6ms of delay at 60hz. 

 

Not trying to dispute your findings, I'm just looking to make sense of the results, as this has amazing potential for researching and improving issues, just as the addition of the F11 measurement lead to fast flip improvements.

Thanks for your thoughts and interest in this. I'll do more slow motion videos with your suggestions.

[schwabinger]

Hi all,

 

I treated myself to a GFX 1050TI recently (my 32" TV playfield only can do 60hz).

 

So what would be the the best configuration for bringing flipper latency down?

 

Nvidia settings: Vertical Sync: Fast?

 

What do I have to provide in VPX`s video options:

 

VPS/Limiter/VSync?

Maximum pre-rendered Frames?

 

More configuration items that could reduce flipper latency?

 

Thanks to everyone here that produces those greate pinball tables and all the software that is necessary to run them!

 

Greetings,

 

Schwabinger

[Outhere]

Default settings in your video card

Try this for starters

 VP-10 Settings-1.jpg   144.34KB   19 downloads

[BoltBait]

I would like to see a test done with a real pin cab controller instead of keyboard:

 

https://virtuapin.ne...products_id=105

 

and

 

https://virtuapin.ne...&products_id=38

 

 

 

Also, could you show us the settings screens for the video card and VPX that you're using for each setup?

 

Thanks!

Edited by BoltBait, 11 May 2019 - 05:16 PM.