I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Here you can post bug reports, comments, suggestions and other feedback regarding GRIP Early Access
User avatar
codemonkey
Grip Developer
Grip Developer
Posts: 97
Joined: Sat Oct 10, 2015 5:02 am

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby codemonkey » Mon Apr 11, 2016 6:56 am

kerikeron wrote:I'm not a coder or anything but wouldn't higher FPS behave differently because of it detecting what you're impacting at a different point? I remember playing many games where you'd have to have at least 60+ FPS in order to clip onto certain pixel edges because having anything below it wouldn't allow your character to land after the first collision. (Character would fall below the pixel edge before the next physics tick)

Again, no idea how the insides of games work all that well but I've played them long enough to know that manipulating FPS causes physics anomalies. Try using FRAPS or something to force your frame rate down at different FPS's and I bet your car jumps farther and farther at higher FPS's with diminishing returns on the distance per FPS.

No, the physics is ticked at 60Hz no matter what is going on with the rendering - we've tested this extensively rendering at 30Hz and are pretty sure it's behaving correctly, if a little more difficult to play due to the input lag. That said, that some of you seem to be experiencing some difficulties has prompted us to double-check all the physics-related code for the vehicle and ensure it's all being done on that physics tick.
Rob
Technical Honcho
Caged Element

User avatar
Skid
Posts: 41
Joined: Wed Feb 03, 2016 8:29 am
Location: Lincoln, UK
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Skid » Mon Apr 11, 2016 7:32 am

codemonkey wrote:
kerikeron wrote:I'm not a coder or anything but wouldn't higher FPS behave differently because of it detecting what you're impacting at a different point? I remember playing many games where you'd have to have at least 60+ FPS in order to clip onto certain pixel edges because having anything below it wouldn't allow your character to land after the first collision. (Character would fall below the pixel edge before the next physics tick)

Again, no idea how the insides of games work all that well but I've played them long enough to know that manipulating FPS causes physics anomalies. Try using FRAPS or something to force your frame rate down at different FPS's and I bet your car jumps farther and farther at higher FPS's with diminishing returns on the distance per FPS.

No, the physics is ticked at 60Hz no matter what is going on with the rendering - we've tested this extensively rendering at 30Hz and are pretty sure it's behaving correctly, if a little more difficult to play due to the input lag. That said, that some of you seem to be experiencing some difficulties has prompted us to double-check all the physics-related code for the vehicle and ensure it's all being done on that physics tick.


Worth noting since it's relevant, the NVIDIA control panel allows the user to not only configure if PhyX his handled by the CPU or the GPU but it also allows the use to say which GPU to handle PhysX. So if the people with physic issues getting worse at lower frame rates have their rendering GPU set to their PhysX GPU as well, if the GPU is overloaded it could probably slow down the PhysX ticks. So it might be an idea to get people to confirm their PhysX settings. I'm not sure how PhysX is handled on AMD cards though, I suspect it's just offloaded to the CPU but that's a guess.

Feel free to correct me if I'm wrong, I'm making educated guesses.
: ) Skid : )
My Multi-Monitor Focused YouTube Channel: http://www.youtube.com/c/SkidIncGaming
My Twitch: http://www.twitch.tv/skid_inc
"Imagination is the only weapon in the war against reality."

User avatar
potterman28wxcv
Early Supporter 6
Early Supporter 6
Posts: 1285
Joined: Wed Jul 01, 2015 12:20 pm
Location: Grenoble - France
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby potterman28wxcv » Mon Apr 11, 2016 10:45 am

Skid wrote:
codemonkey wrote:
kerikeron wrote:I'm not a coder or anything but wouldn't higher FPS behave differently because of it detecting what you're impacting at a different point? I remember playing many games where you'd have to have at least 60+ FPS in order to clip onto certain pixel edges because having anything below it wouldn't allow your character to land after the first collision. (Character would fall below the pixel edge before the next physics tick)

Again, no idea how the insides of games work all that well but I've played them long enough to know that manipulating FPS causes physics anomalies. Try using FRAPS or something to force your frame rate down at different FPS's and I bet your car jumps farther and farther at higher FPS's with diminishing returns on the distance per FPS.

No, the physics is ticked at 60Hz no matter what is going on with the rendering - we've tested this extensively rendering at 30Hz and are pretty sure it's behaving correctly, if a little more difficult to play due to the input lag. That said, that some of you seem to be experiencing some difficulties has prompted us to double-check all the physics-related code for the vehicle and ensure it's all being done on that physics tick.


Worth noting since it's relevant, the NVIDIA control panel allows the user to not only configure if PhyX his handled by the CPU or the GPU but it also allows the use to say which GPU to handle PhysX. So if the people with physic issues getting worse at lower frame rates have their rendering GPU set to their PhysX GPU as well, if the GPU is overloaded it could probably slow down the PhysX ticks. So it might be an idea to get people to confirm their PhysX settings. I'm not sure how PhysX is handled on AMD cards though, I suspect it's just offloaded to the CPU but that's a guess.

I guess it depends on how much resource this PhysX is demanding. If it's enough for the CPU, well of course you'd be better off using your CPU for that, but if the CPU is not enough, it would run much slower on the CPU than on the GPU, because then the CPU would have too much pressure and would start context switching a lot.
Because I'm even more boring than Broscar -> Rules of the GRIP forum
Check out the sign-ups for the Rollcage Neo-mode Tournament 2017, starting 15th April !

And with a rusty wrench
God created the french

Admin of the Image

User avatar
codemonkey
Grip Developer
Grip Developer
Posts: 97
Joined: Sat Oct 10, 2015 5:02 am

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby codemonkey » Mon Apr 11, 2016 12:34 pm

Skid wrote:Worth noting since it's relevant, the NVIDIA control panel allows the user to not only configure if PhyX his handled by the CPU or the GPU but it also allows the use to say which GPU to handle PhysX. So if the people with physic issues getting worse at lower frame rates have their rendering GPU set to their PhysX GPU as well, if the GPU is overloaded it could probably slow down the PhysX ticks. So it might be an idea to get people to confirm their PhysX settings. I'm not sure how PhysX is handled on AMD cards though, I suspect it's just offloaded to the CPU but that's a guess.

Feel free to correct me if I'm wrong, I'm making educated guesses.

With UE4, PhysX is always run on the CPU regardless of what the control panel says. It's worth noting the physics is run in parallel to the main game tick, and takes up much less time than that tick, so it's not a performance issue.
Rob
Technical Honcho
Caged Element

User avatar
Skid
Posts: 41
Joined: Wed Feb 03, 2016 8:29 am
Location: Lincoln, UK
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Skid » Mon Apr 11, 2016 1:22 pm

codemonkey wrote:With UE4, PhysX is always run on the CPU regardless of what the control panel says. It's worth noting the physics is run in parallel to the main game tick, and takes up much less time than that tick, so it's not a performance issue.


Fair enough, but that's a little silly of UE4, what's the point of having a card I can dedicate to PhysX if UE4 is going to override the driver settings, I wonder if there is a way to override it. But I did always assume that the physic system was on the same thread(s) as the graphics system.
: ) Skid : )
My Multi-Monitor Focused YouTube Channel: http://www.youtube.com/c/SkidIncGaming
My Twitch: http://www.twitch.tv/skid_inc
"Imagination is the only weapon in the war against reality."

User avatar
codemonkey
Grip Developer
Grip Developer
Posts: 97
Joined: Sat Oct 10, 2015 5:02 am

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby codemonkey » Tue Apr 12, 2016 5:27 am

Skid wrote:
codemonkey wrote:With UE4, PhysX is always run on the CPU regardless of what the control panel says. It's worth noting the physics is run in parallel to the main game tick, and takes up much less time than that tick, so it's not a performance issue.


Fair enough, but that's a little silly of UE4, what's the point of having a card I can dedicate to PhysX if UE4 is going to override the driver settings, I wonder if there is a way to override it. But I did always assume that the physic system was on the same thread(s) as the graphics system.

I think their rationale was purely compatibility across a wide-range of diverse projects, whether the user has AMD or NVidia hardware. And like I said, it's parallelised with the main update tick so it's not a big performance issue. It made sense to me, esp. when UE4 is actually GPU heavy, so it's already being driven hard with rendering - physics isn't free just because it's on a GPU.
Rob
Technical Honcho
Caged Element

User avatar
Ryu Makkuro
Posts: 390
Joined: Thu Feb 04, 2016 4:46 pm
Location: Poland
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Ryu Makkuro » Tue Apr 12, 2016 6:11 am

codemonkey wrote:
Skid wrote:
codemonkey wrote:With UE4, PhysX is always run on the CPU regardless of what the control panel says. It's worth noting the physics is run in parallel to the main game tick, and takes up much less time than that tick, so it's not a performance issue.


Fair enough, but that's a little silly of UE4, what's the point of having a card I can dedicate to PhysX if UE4 is going to override the driver settings, I wonder if there is a way to override it. But I did always assume that the physic system was on the same thread(s) as the graphics system.

I think their rationale was purely compatibility across a wide-range of diverse projects, whether the user has AMD or NVidia hardware. And like I said, it's parallelised with the main update tick so it's not a big performance issue. It made sense to me, esp. when UE4 is actually GPU heavy, so it's already being driven hard with rendering - physics isn't free just because it's on a GPU.

Especially when CPU's become an actual bottleneck only in a 3-Way SLI configs at 1080p. That is unless someone is running some old game at lowest settings on a Titan X xD
Image

User avatar
codemonkey
Grip Developer
Grip Developer
Posts: 97
Joined: Sat Oct 10, 2015 5:02 am

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby codemonkey » Tue May 10, 2016 1:33 am

I just thought I should mention here that we're making some real, if slow progress with vehicle physics improvements. After a lot of experimentation, testing and coordination with the PhysX team at NVidia, we've been making some significant forward moves with the vehicle collision responses.

It's now impossible to fall out of the world, extremely unlikely that you'll ever get jammed inside the scenery and the annoying kick-outs that you often received when riding hard against a curving wall or tunnel have all-but been eliminated.

Next we'll be tackling a custom collision response handler to give the vehicles a greater sense of weight during collisions and stop them getting unintentionally airborne quite so often. Then finally, we'll head into the vehicle setup again to more specifically configure the engine, tyre and suspension code / properties to really tighten up the experience and make GRIP a real joy to play.

These last two stages are likely to take far less time than the work we've done just recently on collision response, so expect some real improvement in the next build, with more to come in the build following that.

Over and out.
Rob
Technical Honcho
Caged Element

Muskylounger
Early Supporter 4
Early Supporter 4
Posts: 54
Joined: Sat Jul 25, 2015 7:40 pm

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Muskylounger » Tue May 10, 2016 2:24 am

Sounds great!

User avatar
potterman28wxcv
Early Supporter 6
Early Supporter 6
Posts: 1285
Joined: Wed Jul 01, 2015 12:20 pm
Location: Grenoble - France
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby potterman28wxcv » Tue May 10, 2016 7:21 am

codemonkey wrote:I just thought I should mention here that we're making some real, if slow progress with vehicle physics improvements. After a lot of experimentation, testing and coordination with the PhysX team at NVidia, we've been making some significant forward moves with the vehicle collision responses.

It's now impossible to fall out of the world, extremely unlikely that you'll ever get jammed inside the scenery and the annoying kick-outs that you often received when riding hard against a curving wall or tunnel have all-but been eliminated.

Next we'll be tackling a custom collision response handler to give the vehicles a greater sense of weight during collisions and stop them getting unintentionally airborne quite so often. Then finally, we'll head into the vehicle setup again to more specifically configure the engine, tyre and suspension code / properties to really tighten up the experience and make GRIP a real joy to play.

These last two stages are likely to take far less time than the work we've done just recently on collision response, so expect some real improvement in the next build, with more to come in the build following that.

Over and out.

That's really great news :D Can't wait to test it out, it will solve a damn lot of issues
Because I'm even more boring than Broscar -> Rules of the GRIP forum
Check out the sign-ups for the Rollcage Neo-mode Tournament 2017, starting 15th April !

And with a rusty wrench
God created the french

Admin of the Image

User avatar
Ryu Makkuro
Posts: 390
Joined: Thu Feb 04, 2016 4:46 pm
Location: Poland
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Ryu Makkuro » Tue May 10, 2016 1:55 pm

You guys know how to make someones day :D

Image
Image

User avatar
Queadah
Early Supporter 6
Early Supporter 6
Posts: 993
Joined: Sun Oct 04, 2015 2:22 pm

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Queadah » Tue May 10, 2016 2:34 pm

:D

User avatar
Cybruiser57Péter
Early Supporter 4
Early Supporter 4
Posts: 201
Joined: Fri Nov 27, 2015 5:59 am
Location: Hungary, Budapest

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Cybruiser57Péter » Fri May 13, 2016 6:08 am

Me love some good news :)
Hell yeah.png
Hell yeah.png (168.19 KiB) Viewed 6760 times
Science WINS again!!!
Then they opened a portal to another dimension and kidnapping the locals.
The pissed locals are head enough and strike back.
and then... ALL HELL BREAKS LOOSE :twisted:

Image

User avatar
Ryu Makkuro
Posts: 390
Joined: Thu Feb 04, 2016 4:46 pm
Location: Poland
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Ryu Makkuro » Fri May 13, 2016 2:41 pm

codemonkey wrote:After a lot of experimentation, testing and coordination with the PhysX team at NVidia, we've been making some significant forward moves with the vehicle collision responses..

Does all this coordination with PhysX team mean that the basic form of PhysX for the vehicles that nVidia provides will be improved? As in if someone will decide to use PhysX for physics, would they benefit from the improvements you did so far on it?

I always forgot to ask about this when I'm checking the forum while it keeps bothering me all the time I'm not xD
Image

User avatar
codemonkey
Grip Developer
Grip Developer
Posts: 97
Joined: Sat Oct 10, 2015 5:02 am

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby codemonkey » Tue May 17, 2016 1:25 pm

Ryu Makkuro wrote:
codemonkey wrote:After a lot of experimentation, testing and coordination with the PhysX team at NVidia, we've been making some significant forward moves with the vehicle collision responses..

Does all this coordination with PhysX team mean that the basic form of PhysX for the vehicles that nVidia provides will be improved? As in if someone will decide to use PhysX for physics, would they benefit from the improvements you did so far on it?

I always forgot to ask about this when I'm checking the forum while it keeps bothering me all the time I'm not xD

Not really. Most of our vehicle code is bespoke and outside of PhysX - we don't use the vehicle rigging offered by PhysX / UE as it just wasn't up to the job of handling GRIP / Rollcage style vehicles. In fact all of our vehicle dynamics code is custom-written.

The work we're doing with Nvidia is related to collision detection / response, establishing how to achieve what we want with the features PhysX already has. Take CCD for example. We wanted to use this on the entire vehicle for its collision with the world, but that brought with it many problems and often destroyed handling. It turns out what you need is something like a small sphere at the centre of the vehicle with CCD enabled that is just used to stop the vehicle falling out of the world and nothing more - the normal collision shape is still present and still performing the normal world collision / response but without CCD enabled. This is Nvidia's best advice on how to marry the best aspects of continuous and discrete collision detection. Unfortunately, this wasn't possible with UE however, as its interface to PhysX is a bit broken in this regard. It's not really designed very well for multiple collision shapes per physics object. So in the end, we had to rewrite some of the UE engine code to allow for our CCD sphere and stopping the vehicles from falling out of the world. That's a pretty long story for such a small feature.

Much of our work with PhysX has been like this though - trying to find the right balance of things within PhysX to produce the desired behaviour, and then twisting and bending UE to make it possible.

Luckily, I'm talking mostly to a guy at Nvidia who previously worked on Blur, so he knows his stuff when it comes to physics and when it comes to racing. Mostly he's just helping us with the technical details under the hood with PhysX, but it's great that he has an understanding of what we're talking about when it comes to vehicle behaviour too - it breeds a lot of confidence.
Rob
Technical Honcho
Caged Element

User avatar
Ryu Makkuro
Posts: 390
Joined: Thu Feb 04, 2016 4:46 pm
Location: Poland
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Ryu Makkuro » Tue May 17, 2016 5:50 pm

So it's basically adapting Unreal Engine to PhysX, not the other way. Good to know.

PS. Although I personally disliked Blur in its entirety, the experience in the genre is definitely something that helps in communication.
Image

User avatar
KazzyMac
Early Supporter 6
Early Supporter 6
Posts: 370
Joined: Thu Jul 02, 2015 7:03 pm
Location: North Lincolnshire, Ye Olde Englande
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby KazzyMac » Sun May 22, 2016 7:41 pm

Ryu Makkuro wrote:So it's basically adapting Unreal Engine to PhysX, not the other way. Good to know.

PS. Although I personally disliked Blur in its entirety, the experience in the genre is definitely something that helps in communication.

You disliked Blur? Lame, I loved that game. #2 in the world on xbox, baby. #brag

Didn't know that one of the devs was working for nVidia now. Oi, Rob, poke him and tell him Blur 2 needs to be made, damnit. I WON'T STAND FOR ACTIVISION'S SHENANIGANS. :evil:
Image

Look at the road-holding on these babies. Clarkson would cream himself. :ugeek:

Broscar, your car skin is called Tropical. ; )

User avatar
Ryu Makkuro
Posts: 390
Joined: Thu Feb 04, 2016 4:46 pm
Location: Poland
Contact:

Re: I do not enjoy anymore (evolved to discussion with codemonkey about physics details)

Postby Ryu Makkuro » Mon May 23, 2016 6:53 am

KazzyMac wrote:
Ryu Makkuro wrote:So it's basically adapting Unreal Engine to PhysX, not the other way. Good to know.

PS. Although I personally disliked Blur in its entirety, the experience in the genre is definitely something that helps in communication.

You disliked Blur? Lame, I loved that game. #2 in the world on xbox, baby. #brag

Didn't know that one of the devs was working for nVidia now. Oi, Rob, poke him and tell him Blur 2 needs to be made, damnit. I WON'T STAND FOR ACTIVISION'S SHENANIGANS. :evil:

At the time I was playing Split/Second, which felt a lot better (and explosions :D ). Blur always gave me that disconnected feeling, along with input lag and floatiness to the physics. Graphics were too "neony" as well for my taste. Horrid stuff in short. It quite honestly felt like Mario Kart with normal cars... in the bad sense of that phrase.

Too each their own though :P I applaud you for playing that thing for a long time. My brain would melt xD
Image


Return to “GRIP Early Access Feedback”

Who is online

Users browsing this forum: No registered users and 1 guest

cron