New Release D3D9Client Development

[jarmonik]

Performance on my i3 HD3000 laptop is 30-40fps after terrain generation (takes about 4-5s to do that).

How much the frame-rate will increase if you disable atmospheric rendering from an atmosphere control panel. (the one with a lot of sliders) ?

---------- Post added at 05:08 ---------- Previous post was at 04:59 ----------

Does the exhaust flames use BLEND_Alpha?

Yes.

Is there a way to set that in transparent meshes/groups/textures so that it blends more clearly?

I have no idea what you mean. A semi-transparent textures should work in a meshes but a group having a transparent texture must have the highest group index number. Vessels aren't currently z-ordered, so, a vessels may not appear correctly behind a transparent textures.

 

[Weirdo Earthtorch]

How much the frame-rate will increase if you disable atmospheric rendering from an atmosphere control panel. (the one with a lot of sliders) ?

From a very quick test using a scenario with a lot of terrain (atop Everest) I would say atmospheric rendering costs me between 10-15fps. With no AA and atmospheric rendering, I get around 70-75fps; with no AA and no atmospheric rendering, its c. 80 - 90 fps.

I will try to post some screen-grabs with framerate later.

 

[SolarLiner]

I have no idea what you mean. A semi-transparent textures should work in a meshes but a group having a transparent texture must have the highest group index number. Vessels aren't currently z-ordered, so, a vessels may not appear correctly behind a transparent textures.

Ohh, I meant, the ADD flag on transparent "things". This makes sense for exhausts because of the way they look, but it would be cool if you could add those BLEND_Additive tags to transparent parts.

 

[jarmonik]

Originally Posted by Phoenix:
Would it be possible to add a "Tile Resolution Bias" control, as exists in DX7? Reducing the vertex count this way would improve basic frame rate for lesser computers, although it still might not fix the stutter.

Possible, yes, but not easy. I haven't had time to study the implementation Martin made and I was hoping to get some documentations for it. It's difficult to change it without fully understanding it in detail.

Looks like Martin has already implemented the Resolution Bias control. But there was a minor configuration problem. Will be operational in the next version of the client.

 

[jarmonik]

Beta 4 for Rev 8

Here is a Beta 4 release.

- Surface LOD bias is now operational.
- Atmospheric rendering is improved a bit. Current implementation includes only a simplified low weight model. Experimentation model SurfaceEx.fx can be activated from Advanced setup panel.
- Several bugs and CTDs are fixed.

Known issues:

- Cloud shadows are not working. (Difficult to implement with the new atmospheric model)
- Vessel lighting (sun light) not working if the vessel is below planet's mean radius.
- Vessel cockpit lights are not yet working.
- DG main panel is not working.
- Surface bases are floating/shadows are in incorrect place.

If there are any other problems it's good to report them.

 

[jarmonik]

Any thoughts about the atmospheric model ? Would it be good default/base model ?

Is the upper part of the horizon too bright (white) ?

 

[Phoenix]

Any thoughts about the atmospheric model ? Would it be good default/base model ?

Thanks for the LOD bias control. I can improve frame rate by 50% by going one notch lower, and it almost removes the periodic stutter when rolling.

Atmosphere: Switching dynamic exposure on or off has no noticeable effect on frame rate on my computer, and neither does switching the Atmosphere shader to different models. I can't really tell the visual difference between each shader, but maybe that was because I was only flying at low altitude in the Alps. All the atmosphere shaders look great.

Thanks.

 

[SolarLiner]

Still a solid 700 fps with the experimental atmosphere and the LOD bias full on quality. No stutter whatsoever.

I also made (for fun) a Venus atmosphere:

Red = 0.4
Green = 0.4444
Blue = 0.7
RWaveDep = 3.92
ScaleHeight = 38.2
OutScatter = 0.705
InScatter = 4.7
RayleighPhase = 3.276
Balance = 1.362
DepthClamp = 50
SrfColor = 1.592
SrfIntensity = 1
MiePower = 0.7695
MiePhase = 0.907067
Exposure = 0.7251
Aux1 = 0.584
Aux2 = 0
Mode = 0
OverSaturation = TRUE

And I changed the Mars one:

Red = 0.7
Green = 0.6013
Blue = 0.4765
RWaveDep = -3.056
ScaleHeight = 4
OutScatter = 0.2865
InScatter = 2.985
RayleighPhase = 0
Balance = 0.98
DepthClamp = 19.571
SrfColor = 1.154
SrfIntensity = 1
MiePower = 0.825
MiePhase = 0.883525
Exposure = 1.0367
Aux1 = 0
Aux2 = 0
Mode = 0
OverSaturation = TRUE

BTW, Mars' "sky shpere" is all glitchy: when you rotate around, the atmosphere goes all crazy and jumps everywhere.

 

[jarmonik]

Atmosphere: Switching dynamic exposure on or off has no noticeable effect on frame rate on my computer, and neither does switching the Atmosphere shader to different models.

Sorry, I didn't explain it well enough. The "Dynamic exposure" is reserved for future use. Currently all three shader models are equal. The "Experiment" is reserved for users own experimentation in a cases if someone wishes to test his own ideas. I'll start working on an "Advanced" implementation when we have a stable basic renderer operational. The basic shader "Default" is using a simplified model and some sliders are disabled from a control panel.

---------- Post added at 22:17 ---------- Previous post was at 22:08 ----------

BTW, Mars' "sky shpere" is all glitchy: when you rotate around, the atmosphere goes all crazy and jumps everywhere.

I didn't notice that before. But it's jumping pretty bad. I'll try to figure out what's going on..

Thanks, for the model parameters. Isn't the scale height a little too low for mars. I thought it was somewhere around 8k.

Scale height for a planet should be somewhere around:

h0 = Pressure / ( Density * Gravity )

For the Earth:
h0 = 101e3 / (1.29 * 9.81) = 7.9k

For the Mars:
h0 = 0.61e3 / (0.02 * 3.7) = 8.2k

 

[SolarLiner]

Isn't the scale height a little too low for mars. I thought it was somewhere around 8k.

I eyeballed the atmospheres, using what we have as photo references (not much for Venus) and trying to make the atmosphere somewhat eye candy as well.

 

[DaveS]

I think the in-atmosphere color is too yellow. Is there a reason for this change? It's almost like things are not white-balanced anymore.

 

[jarmonik]

I think the in-atmosphere color is too yellow. Is there a reason for this change? It's almost like things are not white-balanced anymore.

Sorry, the change was not intentional. The current plan is to create two different shading models for the atmosphere a light weight model and more advanced model. Currently we only have the light weight model and it's a simplified model from the one from Beta 3, to increase a frame-rates.

There is a minor problem. If the sunlight is considered to be white in orbit then the light passed through an atmosphere is shifted towards a yellow. If we configure the model in a way that the shift doesn't occur then it will also effect in a sky-color and a sunset/sunrise colors.

I made a few tests and it would seem that we can implement a camera altitude depended white balance control, that should fix the problem.

Currently I don't know much about the advanced model. It will be an experiment to see how it works. The problem is that more accurate computations doesn't necessarily mean a better results. But at-least we can add some low level noise to reduce a visible color gradients and maybe add some turbulence in a fog. Also, we could use a different scale height for a fog and haze.

 

[jarmonik]

D3D9Client Beta 5 for rev #8

- Small amount of noise should make a dark sky colors smoother
- White balance control implemented. (Altitude dependency not yet implemented)
- Notice anything new about the moon ?
- Saturn ring issue is fixed.
- Phobos is now visible again :lol:
- Debug Controls related color issues fixed.
- Martian "jumpy" horizon issue fixed.
- High altitude (>64km) cloud shadows on-line.
- oapiCreateSurfaceEx() and oapiLoadSurfaceEx() are now fully supported.
- Surface interface cleaned up and partially re-implemented. Pay attention to 2D surface issues. (Check the log)

I also had an opportunity to test the client with low-medium prized laptop with Intel HD Graphics 4000 and I got about 50-70fps with anti-aliasing and about 200-240fps without AA. So, I think we are doing pretty well.

 

[DaveS]

Ran beta 5 a bit and it seems pretty solid except for some weired flickering/"shaking" of the cloud shadows.

---------- Post added at 06:52 PM ---------- Previous post was at 06:18 PM ----------

Another item of interest:

This doesn't quite look right.

 

[Phoenix]

Flying over Innsbruck this evening looks fantastic.

Thanks.

 

[Weirdo Earthtorch]

Thanks for the latest update. Atmospheric rendering from ground level at sunset (in particular) looks stunning.

A couple of issues observed on my i3 Intel HD3000:

Outer planet moon rendering:

Seems that way with all outer planet moons. They render correctly as the view distance is increased. (Tested using the camera dialog box CTRL+F1)

This video shows some flickering of terrain, only seems to occur when the camera is close (<100m) to the vessel. Apologies for the bad flying/camera angle - it does become apparent. (occurs at about 0:40. Observed in Builds 4 and 5).


Finally another vid showing the same effect on the moon; and a terrain/distance issue (occurs also in the inline client, so not a D3D9 issue).


Also, D3D9 planet night lighting seems to render a lot later than the inline client (as in, the lights 'come on' 2-3 sim hours after the inline client renders them).

Overall thought the effect and performance is fantastic - thanks.

 

[jarmonik]

Another item of interest:
This doesn't quite look right.

Yeah, it´s a bit odd. The terminator is a little problematic. Could you explain a bit what would you change ? What´s wrong in it ?

---------- Post added at 08:09 ---------- Previous post was at 07:52 ----------

Outer planet moon rendering:

Thanks, I haven´t noticed that. Looks like CW, CCW culling issue.

This video shows some flickering of terrain, only seems to occur when the camera is close (<100m) to the vessel. Apologies for the bad flying/camera angle - it does become apparent. (occurs at about 0:40. Observed in Builds 4 and 5).

It´s a bit difficult to explain why it happens. But it´s problematic to render a high resolution terrain on distance while something else is close to the camera, like a vessel or a particle effects. It may take some time to fix that but when the time is right I´ll try to split the vessel rendering into a different rendering pass.

Martin is probably working on the other terrain issue on the Moon and Mars.

---------- Post added at 08:55 ---------- Previous post was at 08:09 ----------

D3D9Client R13.1 Is everything o.k. with this. Should we make it an official release for Orbiter 2010-P1

http://d3d9client.codeplex.com/releases/view/118435

 

[Marg]

I just noticed this client for Orbiter BETA potential is huge.
Sun is very good, when in atmosphere, I just was longing for such "glare"...
Maybe in space it could be made some "remnant" of this effect? Colouring near the horizon (when in atmosphere), at last, is not so gradual as it is in standart client (~8-bit colours).
But some notes:
1) I think sunset/rise colours ar mostly yellowish. No pink, light blue sky.
2) I see anomalous shadows on terrain, when sun is low, they are blocky (pixelated) black, when on ground level. If I go up 1-2 km they disappear turning to correct, anisotropically "filtered" appearance. It`s observable also on the Moon...

 

[DaveS]

Yeah, it´s a bit odd. The terminator is a little problematic. Could you explain a bit what would you change ? What´s wrong in it ?

It's just that the ground looks a bit weird and off with all that yellow. Based on what I have seen in the ISS HDEV camera views, the current RC sunset/sunrise is more realistic.

Also, the actual colors on vessels needs to be tweaked. I came up with very close ones here: http://www.orbiter-forum.com/showthread.php?p=450584&postcount=2653

---------- Post added at 11:27 AM ---------- Previous post was at 11:16 AM ----------

Sun rise video shot by one of the SD External TV Camera Group (ETVCG) cameras on ISS:

 

[Donamy]

This is for Orbiter 2010 correct ?