My testflight with DX10 and PMDGs 777 looked great, but I noticed that performance in VC view drops around 15 FPS compared to the outside view in the same situation. Its flyable, but stuttering. Switching to outside view almost always gives me 25- 30 FPS (which is external set limit anyways). It makes no difference where I am when observing this, its the same over the sea or over a big city.

 

My system is a Phenom 2 X4 @ 4,0 Ghz with an OC'd 560Ti.

 

I am using 4xCSSA in Steves Fixer and 4 x SGSS set in Nvidia inspector.

 

Poolsize set to 0 (best results by experience on my rig)

 

Has anyone a hint what to look at regarding PMDG's VCs?

 

I am not posting this in PMDG's support forum because I can figure out their answer anyways : "please use DX9 for our planes!"

It's completely normal, and you'll find the same behavior with all of PMDG's aircraft. Other aircraft with detailed VCs are affected too, like Coolsky's DC9 for example.

Are you talking about DX10 in particular or do you say this is generally normal in complex airplanes? Because to the latter I can tell you no, its not. In DX9, my FPS are slightly worse in general but better with PMDG planes. Especially in cruise there is a huge performance hit. Switching AA or clouds off does not make any difference so I guess this has nothing to do with graphics settings. FPS seem to be stuck at around 10 - 15 FPS in VC view and sonetimes even drop below 5.

From my point of knowledge the issue seems to be CPU management related.

 

Considering how many ppl are there on the internet telling that PMDG planes work fine in DX10, I refuse to see this as normal. would be a shame otherwise because those planes are the reason I gave DX10 a try in the first place. (OOM's)

I found the cause: 777 immersion effects. Its a shame, but they dont seem to be working properly in DX10 so far.

 

After uninstallation, FPS in cruise are much closer to what I am used to with this plane.

I found the cause: 777 immersion effects. Its a shame, but they dont seem to be working properly in DX10 so far.

 

After uninstallation, FPS in cruise are much closer to what I am used to with this plane.

Works great for me, no significant frame rate loss with Immersion.

Yes maybe they either were left unconverted or there was something strange going on. I will try a reinstall. Other than thax DX10 works great and alongside with a few fixes (sun) it feels like a new simulator.

Just an idea...did you convert particle effects with the Fixer?

 

Converted particle effects are really BIG frame killers! Steam from power plants or contrails, smoke from braking wheels...all a no go. Interestingly, these effects seem to be more CPU dependend, cause the GPU utilization is reduced the same time the frames drop.

 

For me, adding some particle effects to the [exclude effect] section in the DX fixer product.ini is the best way to avoid unwanted conversion.

 

Matthias

Interestingly, these effects seem to be more CPU dependend, cause the GPU utilization is reduced the same time the frames drop.

 

When your CPU is overloaded GPU usage will drop because the CPU can't feed it enough data. It doesn't take all CPU cores being loaded 100% to cause this it just takes the primary thread being bogged down.

 

That's why a lot of these "low gpu usage" threads are erroneous

Try tweaking some things in the performance tab of the CDU. I managed to gain a few FPS there. May only work on older systems like mine though.

I know :wink:

 

The interesting thing for me is, that in DX10 more work is shifted to the GPU. But not so with particle effects. I thought, that modern graphic cards would be better in rendering smoke or spray effects and all that stuff than a CPU. But even in DX10 the rendering is done by the CPU, and so, the CPU is the bottle neck and the GPU is bored when dealing with particle effects.

 

Matthias

I know :wink:

 

The interesting thing for me is, that in DX10 more work is shifted to the GPU. But not so with particle effects. I thought, that modern graphic cards would be better in rendering smoke or spray effects and all that stuff than a CPU. But even in DX10 the rendering is done by the CPU, and so, the CPU is the bottle neck and the GPU is bored when dealing with particle effects.

 

Matthias

 

GPUs aren't faster than CPUs though. For some tasks, but not everything. Things like draw calls are handled by the CPU. Those are serial tasks which the CPU excels at better than the GPU. Can't really expect high GPU usage when you're issuing all those draw calls on the CPU. Now if you had a lighter aircraft, such as the default 737, you would see higher GPU usage in the VC due to less draw calls being issued. Try it, you'll see. Not the case here.

Thanks for explanation :smile:

 

At the end i only wanted to assist soulflight to choose the correct settings for a beautiful DX10 experience.

 

But as we just talk about these details... why is the unconverted effect less demanding on the CPU than the converted? Where is the difference in terms of CPU workload between a DX9 / unconverted file and the converted DX10 effect?

 

Matthias

I'm not sure. Most of FSX's effects are 2D textures, and not 3D volumetric like P3D's fog. I haven't really noticed anything there with the effects in terms of performance.

I'm not sure. Most of FSX's effects are 2D textures, and not 3D volumetric like P3D's fog. I haven't really noticed anything there with the effects in terms of performance.

 

You are correct concerning FSX' dafault effects. However, FSFX Packages addon effects or shockwave lights add volumetric effects, they seem to take a performance hit then.