Quote from DANIEL-CRO :

When just entering 3D mode in dialog right click IPD and set value outside range, LFS crashes

Looks like the same goes for every slider in that dialog box.

Quote from Flame CZE :

When on track using one of the cockpit views (in car or custom), the interface/text mip filter seems to be set to 0.0 no matter what I set in the Graphics options.

I tested this a bit and it's a little bit different for me. My own interface/text mip filter defaults to -1.0 (not the setting itself only the interface appearance, the setting remains on whatever i set it) only in cockpit/custom/follow views, but not in top down where car is also shown. In force view where the car isn't shown there is no issue in either of the views.

After some further testing, following a 'hunch', i discovered it is related to distance of the car from the viewer camera. Will attach two images to show this and for this purpose I have set bias to 0.0 (blurry one) and it defaulted to -1.0 (sharp one). This is why the top down camera didn't exhibit this issue - the car was too far away. Anyways i hope it helps in finding the cause rather then sidetracking again
Distance at which it 'flips' isn't set, it can be from 7 - 15m, it varies. It might also not be distance only, might be number of car polygons currently drawn (less with distance), scenery details in the scene, I don't know!

Hello

I've done a few performance tests and 3D mode (red-cyan glasses).

As for performance it had 52 fps on a version 0.6F and 61 fps for version 0.6F2 (lfsbench test) so 17% more fps.

I compared the 3D mode with an available mode for nVidia drivers and it seems to me that there's a little more depth. I miss it 16:10 HUD mode.

I throw to compare 3D Nvidia (Left) and 3D Scaven (Right)

Regards

Quote from Scawen :

The people who have such low frame rate in red-cyan mode, could you try with the Anaglyph.psh from the attached zip?

Have tried it just now, no change.
Apart from low fps, even mouse pointer reacts more slowly (lags).
Also, more FPS in windowed mode. The less the window size, the more FPS

i think wed need a colour adjustment mode to finetune things
my monitor is reasonably close to srgb and i still get a fair bit of ghosting (mind you i have very cheap red/cyan glasses but still)

Benchmark with Fraps

Version 0.6F
Min / Max / Avg
69 / 137 / 105

Version 0.6F2
Min / Max / Avg
88 / 164 / 128

>20%

Version 0.6F2 in 3D (red/cyan)
Min / Max / Avg
56 / 113 / 85

Field of 20 AI cars / cockpit view in the last car (20.) while 2 minutes

I seem to have the same shader issue in Linux. Mine throws an error message though, which might help (see attached).

OS: 3.13.0-32-generic, Xubuntu 14.04 x64
Wine: 1.6.2
GPU: Kaveri (APU) R7 200 Series, fglrx (proprietary driver)

I've yet to try the open source driver.

Quote from Scawen :

The people who have such low frame rate in red-cyan mode, could you try with the Anaglyph.psh from the attached zip?

Well, it didn't help very much. Now I get 17 FPS instead of 15 FPS.

As sermilan mentioned: FPS get higher decreasing the screen size in window mode. Mouse pointer lags no matter of the screen size.
This applies for both Anaglyph.psh files.

Windows 7
Pentium E6300
Geforce 9300
(yeah, I know I need to upgrade my system)

Quote from Major Pain :

FPS get higher decreasing the screen size in window mode.

lower resolution - less pixels - more FPS

Resolution set in LFS doesn't have any effect while LFS is windowed. Resolution is then affected by window size.

OK I just tested this FPS thing also. It drops, from 260 FPS to about 130 FPS in my case so indeed ~50%.

However... Nvidea Inspector reports to me zero GPU load increase? My GPU load sticks around 50-55% . I experience no increased mouse lag. Isnt it just a false FPS report?

Fraps may tell more.. Brb

// Fraps reporting the same. OK weird.

Quote from Scawen :

Did you try (with width set to 442mm) a lower IPD, like 30mm?

Well, with the setting you described above, it worked perfectly, like with high screen width and higher IPD. Looks like while the screen width was the real value, the problem was the IPD.

Quote from Scawen :

One other thing, have you tried a low FOV? I'm thinking that maybe the correct IPD and screen width only work if the FOV is real too. Maybe when your FOV setting is higher than the true FOV offered by your screen, then you need to use smaller IPD or greater screen width.

Tried FOV 63° it was great, but with 86°, still is. My only and main problem is ghosting (well, don't know really if ghosting means this, but I can see some of the cyan colors with my "red eye", and vice versa).
Btw I don't have extremely low FPS with the red-cyan 3D mode, because it's still playable (enough for me). 3D - 39,2, while with 2D - 58 FPS on South City.
The mouse goes slower (=lags) a bit while in the 3D mode, as sermilan pointed out.
As pasibrzuch mentioned, I join him - the interior looks awesome in 3D!
And guys... "FPS get higher decreasing the screen size in window mode." is totally normal, as Daniel-Cro said it.

Is there a specific reason why RC anaglyph does not use full height for interface/insim/... like 2D ?

Harder to read ...

Quote from sermilan :

Have tried it just now, no change.
Apart from low fps, even mouse pointer reacts more slowly (lags).
Also, more FPS in windowed mode. The less the window size, the more FPS

Well... I've set no AA in 2D mode, assuming it would stay off in 3D too, but it was on. After setting "Antialiasing" to "no" FPS went to 60+ and mouse lag was gone... so that one is solved!

A bit of off topic: you guys that are trying red-cyan glasses, after you've finished testing and put the glasses off, try looking through one eye only, than other one too

Framrate increased around 10% with 6F2 on my old rig.

Quote from Nilex :

Thanks so MUCH! I am eying an Oculus from now

It´s funny how ppl react only to the 3D mode
I don´t want to push the hype, but as a Rift owner I ´m dare to say following awesomeness factor can be expected in VR:

Quote from sermilan :

Have tried it just now, no change.
Apart from low fps, even mouse pointer reacts more slowly (lags).
Also, more FPS in windowed mode. The less the window size, the more FPS

OK, I read your later post and you solved it by switching off antialiasing. Anaglyph 3D renders two images to a render target, which, if antialiasing is enabled, must then be copied to a render target texture, before the two images can finally be combined, processing each pixel, into the main backbuffer, to be presented to the screen. So not only does it render two full images of the world, there is a lot of copying around after that which is significantly reduced by switching off AA. I guess your in-game mirrors might cause some slowdown if Mirror AA is enabled...

Quote from Shotglass :

i think wed need a colour adjustment mode to finetune things
my monitor is reasonably close to srgb and i still get a fair bit of ghosting (mind you i have very cheap red/cyan glasses but still)

There are some other options in the Anaglyph.psh file. You can comment out the active lines and uncomment some of the other options.

By the way, if you edit any of the shaders while LFS is in a window, you can cause a reload by resizing your window a little (by pulling the edge of your window a little).

Quote from Degats :

I seem to have the same shader issue in Linux. Mine throws an error message though, which might help (see attached).

OS: 3.13.0-32-generic, Xubuntu 14.04 x64
Wine: 1.6.2
GPU: Kaveri (APU) R7 200 Series, fglrx (proprietary driver)

I've yet to try the open source driver.

I remember people needing to update their D3DX9_43.dll for version F. These shaders depend on a good version of D3DCompiler_43.dll so maybe that could help?

Quote from cargame.nl :

OK I just tested this FPS thing also. It drops, from 260 FPS to about 130 FPS in my case so indeed ~50%.

However... Nvidea Inspector reports to me zero GPU load increase? My GPU load sticks around 50-55% . I experience no increased mouse lag. Isnt it just a false FPS report?

Fraps may tell more.. Brb

// Fraps reporting the same. OK weird.

Which FPS thing? Are you talking about the Anaglyph 3D mode causing a large slowdown? If so, that is explained above.

Quote from DANIEL-CRO :

Is there a specific reason why RC anaglyph does not use full height for interface/insim/... like 2D ?

Harder to read ...

All the 3D modes use a projected 3D interface, so that's the reason. But you can select 4:3 or 1:1 options.

Quote from Scawen :

Which FPS thing? Are you talking about the Anaglyph 3D mode causing a large slowdown? If so, that is explained above.

Yes, the massive drop in FPS number. I already thought it might not be coincidental that its around 50%.

Still amazing to see that the GPU load doesn't increase in my case. But I don't have much knowledge about how GPU accelerators work, only observing.

Thank you.
But a "Minor" Update is in my eyes a other Update as this little thing.

Quote from cargame.nl :

Still amazing to see that the GPU load doesn't increase in my case.

That is a bit freakish since I'd (like you?) have assumed that the FPS drop would be basically entirely due to the GPU. Perhaps the load reporting is bogus.

Well, sending commands from the CPU to GPU also takes time. Rendering the cars in 1 pass instead 3 passes will save a bit.

I've noticed there are minor graphical differences between 0.6F and 0.6F2: the dashboard surround on the XRG/XRT is slightly lighter in 0.6F2, for example.

Quote from PeterN :

I've noticed there are minor graphical differences between 0.6F and 0.6F2: the dashboard surround on the XRG/XRT is slightly lighter in 0.6F2, for example.

That is the only difference I've noticed as well. It seems the old version of that was wrong, and the new version works as expected.

I can't explain why it was wrong... it seems that the glass in front of the clocks became solid black when it was not in the light. That is, solid black if you tried to look through it to the dashboard surround. But somehow the actual clocks could still be seen clearly through it.

But with light on the glass, it became more transparent, so you could see through it as intended. Strange. I made these observations by looking from in-car view and also by using close-ups in SHIFT+U mode with the help of the near clipping plane to see the dark glass.

After a while I gave up trying to understand it, the new version works as expected so that's ok.

Scawen, it's possible to use Shader Model 3.0?

Anything's possible but I don't yet know why I would or wouldn't.

I chose shader model 2 as it seemed to offer what I needed. I have not researched properly, what shader model 3 could offer or what negative effects that may have on compatibility with a certain age of supported graphics hardware.

I can already imagine the response "Scawen is insisting that LFS needs to run on a casio wristwatch from the early eighties". No, I am not. However there is a balance between features and compatibility, and what I have said here is "I don't yet know where that level is".

Tried to set up some more DX9 related things in winetricks (but I already had whole DX9 package installed), and it didn't help in my case.
But I noticed the error messages this time, the first two messages are:
"Could not compile pixel shader (Car1) D3DERR_INVALIDCALL"
"Could not compile vertex shader (Car1) D3DERR_INVALIDCALL"

Which doesn't make sense to me... anyway, I really want to upgrade my OS ... in next 6 months for sure... So I will retest after that again.

Quote from Scawen :

Anything's possible but I don't yet know why I would or wouldn't.

Shader Model 3.0 appeared in 2004. It is 10 years ago!

- reduces resource consumption GPU in complex calculations thereby increasing the FPS

per pass can be applied to a object maximum four light sources. For comparison, the cards that support only 2.0 shaders, each light source requires one pass rendering.

it reduces the requirements for memory bandwidth and performance vertex shaders. And the number of processed triangles can be reduced up to 40%

You can compile TEST PATCH with SM 3.0 like TP with DX9

Quote from denis-takumi :

Shader Model 3.0 appeared in 2004. It is 10 years ago!

- reduces resource consumption GPU in complex calculations thereby increasing the FPS

per pass can be applied to a object maximum four light sources. For comparison, the cards that support only 2.0 shaders, each light source requires one pass rendering.

it reduces the requirements for memory bandwidth and performance vertex shaders. And the number of processed triangles can be reduced up to 40%

You can compile TEST PATCH with SM 3.0 like TP with DX9

This is actually very accurate.

A SM 4.0 videocard will get more FPS in a 3.0 enviroment than it does in a 2.0 using same graphical test. FutureMark apps can be used for reference.