Quote from Matrixi :

LFS physics run at 100Hz. Using vsync in racing games is horrendous as it adds a boatload of input lag to all controllers. Playing at a locked 100 FPS without vsync feels considerably better and more direct.

Valuable information, thanks.

It certainly adds an input lag when vsync is on. I couldn't get used to it when I had it on, so I have it off.

Remember to check if your monitor does support 75Hz refresh rate (75 FPS) With this frequency it is easier for me to read moving text .

unfortunately no monitor these days suppots 100hz output
oh the good old crt days

Quote from Shotglass :

unfortunately no monitor these days suppots 100hz output
oh the good old crt days

We have TN panels capable of doing 144 Hz, CRT monitors were good, but technology has catched up and now you don't have any real reason to miss them so much

which wont help one bit for eliminating update rate interferences on a sim that runs its physics at 100hz

Quote from Shotglass :

which wont help one bit for eliminating update rate interferences on a sim that runs its physics at 100hz

I'm pretty sure if they can run at 144Hz they can slow down to 120, 100, 75 and 60Hz, just like any other monitor do.

So to sum it all up, question for Scawen.
Is it possible to use VSYNC (e.g. 60FPS), without affecting to physics loop. Like physics run at 100Hz (or more?), but graphics 60FPS?

Quote from DANIEL-CRO :

So to sum it all up, question for Scawen.
Is it possible to use VSYNC (e.g. 60FPS), without affecting to physics loop. Like physics run at 100Hz (or more?), but graphics 60FPS?

I'd be surprised if it didn't already ! (Yup, I have read all it has been written). But I'll let the man himself answer .

Quote from Whiskey :

I'm pretty sure if they can run at 144Hz they can slow down to 120, 100, 75 and 60Hz, just like any other monitor do.

highly unlikely that they can actually operate their panel at 100hz
also all 120+hz screens except for one excessively expensive eizo are tn rubbish

So i get E13 working and when i went to change my car on a server it crashes hear is a screenshot

Quote from DANIEL-CRO :

So to sum it all up, question for Scawen.
Is it possible to use VSYNC (e.g. 60FPS), without affecting to physics loop. Like physics run at 100Hz (or more?), but graphics 60FPS?

The physics always runs at 100Hz, even with graphics at 60FPS but this means the number of physics frames per visual frame is variable, like 2, 1, 2, 1, 2, 2, 1, 2 etc.

Quote from cougar8 :

So i get E13 working and when i went to change my car on a server it crashes hear is a screenshot

Windows is looking for a solution! Obviously it will not find one! Would be more helpful if it told you the crash address. Seriously, MS is going nuts these days.

- Are you able to change your car while in single player?
- Does this happen every time?
- Is it possible in your version of Windows to get a crash address?

Quote from cougar8 :

So i get E13 working and when i went to change my car on a server it crashes hear is a screenshot

Do you see something like this?

Click "View problem details", and paste it here

Quote from Scawen :

The physics always runs at 100Hz, even with graphics at 60FPS but this means the number of physics frames per visual frame is variable, like 2, 1, 2, 1, 2, 2, 1, 2 etc.

I tested with how many times per second car physics data is updated, I would clearly say its affected by vsync/fps limit.

Quote from DANIEL-CRO :

I tested with how many times per second car physics data is updated, I would clearly say its affected by vsync/fps limit.

Well... either I'm not sure what you mean, or you are mistaken.

For example, try running a hotlap replay. If there is a difference in the number of physics calculations, the replay will go out of sync. Every physics update is run at a time step of 0.01 seconds, and LFS tries to keep this in sync with real time by doing enough physics calculations to keep up with real time.

Quote from Scawen :

If there is a difference in the number of physics calculations, the replay will go out of sync. Every physics update is run at a time step of 0.01 seconds, and LFS tries to keep this in sync with real time by doing enough physics calculations to keep up with real time.

Yes, this struck me the other day. So the problem (if it exists!) is likely to be connected with inputs from controllers (as someone else suggested) and/or visual feedback. The car would then be nominally doing exactly what it's told, but what the driver is telling it to do is messed up either because the driver isn't seeing the right things or the controllers aren't being read in a timely fashion...
I'm planning to do some experiments myself later. Might even do "blind" tests (i.e. I get someone else to set the fps limit and switch off the fps display )

Just to confirm what I think is well known... the inputs and outputs are only done at the frame rate. There's no point taking inputs between 2 physics updates done in a single visual frame because there is so little time between them. So the best solution is 100Hz frame rate, not possible with current monitors with vertical sync enabled.

Aha. Thanks, and apologies - I was unaware.

No need to apologise!

Seems it wasn't clearly known. It made me think a bit, what are we doing with that wasted time, waiting for the vertical sync... nothing? You could possibly introduce an artificial delay between 2 physics updates that happen in one frame, to read an extra input between the two physics updates (so you'd read two inputs with a tiny time gap between them instead of just one). But you might risk missing the vertical sync and cause a glitch and I'm not convinced it would help in any way. Or maybe a multithreaded system with the physics and graphics totally decoupled, but the physics system would need to save "consistent visual states" for any graphical update that comes along while doing the next physical update.

Complicated stuff with a lot of bug potential... maybe the solution is 100Hz monitors. Yes, that is the proper solution!

OT : Also in my opinion (and some others) TV makers and standards setters should be looking at higher frame rate displays, recording and transmission systems, instead of just higher and higher pixel counts that you need a magnifying glass to see. Higher frame rates don't even need much more bandwidth because less of the screen changes between frames...

Funny you say that - some time back, before I realised LFS was single-threaded, I had assumed that there was a physics thread which would regularly update, and a separate graphics thread which would display stuff from the most recent snapshot (bit like double-buffering) of the physics data
(Beneath this was the possibly-wrong assumption that the snapshots required by the graphics thread were pretty small.)

My own particular problem right now appears to be linked to CPU overload (open tracks). Not sure what would fix that. Is it really the case that the physics loop is still updating at 100 Hz even when the CPU is pegged and the graphics update rate falls to 30 fps or less?
(I confess I have yet to try the DX9 patch as I was assuming it would make things worse.)

Edit: no argument on the tiny pixels. Total waste!

Quote from Neilser :

Is it really the case that the physics loop is still updating at 100 Hz even when the CPU is pegged and the graphics update rate falls to 30 fps or less?

Yes, so the physics is always consistent and stable regardless of graphical overload.

Quote from Neilser :

(I confess I have yet to try the DX9 patch as I was assuming it would make things worse.)

No harm in trying it as you can just run whichever exe you prefer - there's no complicated roll-back!

Quote from Neilser :

Edit: no argument on the tiny pixels. Total waste!

Yes, though I understand the need for more more pixels on *really* big screens and also for the Oculus Rift.

On topic :

I look forward to releasing a minor update this week with antialiasing in mirrors and also on the Oculus Rift display. These must be options in both case - there is a copy involved which costs frame rate, specially on the Rift due the the massive render target. I guess you will use native resolution rather than 1920 on the Rift if you use antialiasing as you may find that downsampling is no longer required.

Scawen, instead of hoping for 100 hz monitors (which is unlikely), why not look towards a 120hz cycle for the physics as there are already displays that run at 120hz (plus divides evenly by a 60hz display, rather than having slight drift).

Not that it matters too much, but that's a more likely solution than hoping for 100hz monitors.

Perhaps I''m a bit daft, but doesn't most 120/144hz monitors run at 100hz just fine?

Bumping the physics engine up to 120 Hz would make it run smoother for both 120 Hz and 60 Hz displays, which would mean virtually everyone. Probably the only people who would find that change a negative would be people who play on a television that only runs at 50 Hz, which probably isn't very common at this point since, as far as I know, most modern European TVs (LCD, plasma, whatever) support both 50 and 60 Hz.

Quote from felplacerad :

Perhaps I''m a bit daft, but doesn't most 120/144hz monitors run at 100hz just fine?

How much did you pay for that monitor? Sorry but I am from rretarded country where everything is 5-30% more expensive because of taxes.

EDIT: meh, 380 €