flying_w

Thank you for performing this analysis, it makes for interesting reading. Up to now I'd ruled out an upgrade from Sandybridge to Haswell, but it may be worth reconsidering. Two questions: 1) Are there any differences in CPU cache size for the processors tested so far? 2) Is everyone using the same AI package? I imagine both the workload of model rendering and schedules can affect CPU performance. Perhaps there are no models in sight in the test flight, but FSX still needs to drive the AI through their workflow. Simon

I may have imagined this, but is a fix for performance problems in clouds with high levels of SGSS on the radar? I thought I read it here somewhere that Steve understood something about the issue and may be able to fix it, but I can't find that now. Simon

Thanks - I'd changed it in Rex at some point, and now I've changed it back again I've whitecaps!

I'm hanging in there on DX10, as I think I've overcome many of my image quality issues. One of the enhancements I know is out there is white caps, however I'm unable to find them. Are there any particular settings I need to see them? Tried all varieties of water 2.x and wind. Simon

I came from the GTX470 to a 680. While I can't say I've noticed much in terms of FPS, I'm able to configure much more in terms of image quality processing. Recently I've moved to DX10, and that needs all the help it can get in that department. The only downside (in DX9) is that I can no longer use the bufferpools UsePools=0 nor RejectThreshold settings. They worked great on the 470, but on my 680 there are display artifacts in no time. Others however do not have this problem. Overall, I've no regrets. And as for other games, like Crysis, it's awesome. Simon

I discovered that regular supersampling in place of SGSS eliminated the autogen tree shimmer without the performance cost. However it did not help on airport buildings and AI aircraft. I'm pretty sure these are mipmapped, being MyTrafficX and default FSX airports. That's the last part of the puzzle to solve, and I'm not sure I can find a solution without a slide show.

I have dual screens, and had to switch one off in Windows in order to get Vsync. I also have a utility called TeamViewer for desktop sharing, and needed to close that down too. Simon

No, DX9 is most decidely not perfect. I can still find shimmers in DX9, though not to the same extent as DX10 at the same settings or higher within the bounds of GPU performance. I'll try to explain the DX10 challenge in the following pictures. Each is just a small part of a 1080p full screen at 8xS and 4xSGSS. The shimmer is seen in certain types of scenery object (i.e. airport buildings, AI aircraft and winter leafless autogen trees) that are some distance from the camera. Move up close to each object concered, and they are perfect. For comparison with DX9, I can do better using 8xS and 2xSGSS. To get DX10 to the same quality, I have to go 8xQ and 8xSGSS. In shimmer1.png below, Manchester Airport EGCC is a mile or two away, and the red circle is drawn around the airport buildings. In a stationary screenshot they look fine, but as soon as the aircraft moves they start to shimmer to the extent that it's hard to ignore across the entire length of the terminal building. http://www.supertraf...im/Shimmer1.png In shimmer2.png, the autogen trees are to the left. It's difficult to see, but the small woodland here has trees that are just branches and plenty of transparency between then. That seems to cause shimmer, summer trees with leaves are much less of a problem. Once again, they look fine until the aircraft is in motion. http://www.supertraf...im/Shimmer2.png Shimmer3.png does not look so good even when stationary. http://www.supertraf...im/Shimmer3.png Perhaps it's high time to just settle down and fly before I start over analysing the situation. DX10 still has lots of goodness. Simon

Thanks from me too. I gave it a try, and while it looks great when the simulator is paused as soon as we are on the move the shimmers are still there and still slightly worse than those I see in DX9. So I guess it's a trade off, loose a little in image quality to gain all the DX10 features you describe (at least it's that way on my PC). Simon

Thank you everyone for the help. After many trials I found the following to be a good combination: MultiSamplesPerPixel=4 MultiSampleQuality=8 4xSGSS (No FXAA) (No SweetFX) It's not DX9 image quality (at 2xSGSS), but I think it's something I can live with. Flying in low visibility or in summer also helps! Sometimes that's working my graphics cards pretty hard, we'll have to see if there are any cloud combinations that kill it. But for now all is good, at last I can do some "proper" flying in DX10.

Would you mind sharing your Nvidia Inspector settings please? I'm not sure if I have it working or not. Simon

I love almost everything about DX10, especially the elimination of graphical spikes that I used to see with BP=0 on DX9. However one challenge remains, shimmering of ground objects and in particular winter trees (branches without leaves have lots of transparency inbetween) and airports in the distance. To get near to the level of quality I had with DX9, the following configuration is necessary: MultiSamplesPerPixel=8 MultiSampleQuality=8 8xSGSS Of course, it's not long before this is a frame rate killer whenever significant clouds are involved. So my question is, is there some other way to do this or should I really be lowering my expectations of image quality a little for DX10? Simon

The biggest impact on performance is those AI which require a physical model to be rendered, because you can see them out of the window. Next is those out of view, typically at other airports or enroute etc. While they don't require as much effort to maintain, FSX has to keep driving them in terms of workflow and their situation in the world. SuperTrafficBoard does have some tools to help maintain your performance if the number of AI is a challenge. We can: Delete AI on the ground at airports other than the one you are at (i.e. other than the one you have selected on STB). The impact of this can be significant, for instance I've experienced FPS gains of 10-15% by doing so (your mileage may vary of course). Reduce all AI by a certain percentage. That may be an interesting alternative to reducing the slider as the choice of AI to eliminate is random so you will see a different set of AI each time you use it. One thing we can't do is to eliminate AI that are enroute, and in some cases UT2 traffic. The scheduling engine (either FSX or the UT2 runtime) tends to recreate them after a while, which does not always happen for AI on the ground. Some excellent suggestions are made in previous appends. If you find they don't lead you to a solution I think the idea of reducing your AI traffic at a specific airport to a specific number is an excellent idea, and I'd like to investigating including that in STB. Simon

The FSX version of MyTrafficX can be imported into P3D for the most part, certainly I have it working that way. However there is a native P3D version you can buy if you want all the tools to work. Simon

A "Tick+", but perhaps the "+" is more about the GPU. We'll see soon enough with more benchmark results. http://www.anandtech.com/show/4378/ivy-bridge-a-tick-with-configurable-tdp See the last "slide".

Don't overlook the architecture improvements. I've heard Ivy Bridge described as a tick+, meaning a little more than just a fabrication change. In particular, look at the hyperPi comparison here: http://www.tweaktown.com/articles/4621/intel_ivy_bridge_overclocking_with_the_core_i7_3770k_and_core_i5_3570k_cpus/index3.html 2600K@5.2Ghz: 11:50 3770K@4.8Ghz: 9:48 IvyBridge is considerably ahead of Sandbridge despite being 400Mhz slower in the test. SuperPi tests are a pretty good reflection of CPU performance, so it would be interesting to see how this works out in FSX. Otherwise I'd agree the overclock leaves something to be desired, let's hope it improves with respins. Simon

The only crash I've encountered is on application exit, which they understand and are fixing. Incidenentally I've not seen any further FSX g3d.dll crashes after installing the latest FSUIPC.Simon

SmartAssembly is something used by a number of add-ons to protect their content. I believe Ultimate Traffic 2 may produce this message also.

FSXMark11, by the book.

Migrating from an i7-920@4Ghz to i5-2500K, and wanted to get some object analysis of the difference. Still on the 920 and benchmarking the first flight, I see an average of 30-32 fps much in line with other 920 users. However the 2nd and subsequence flight (i.e. without restarting FSX in between) saw more like 40 fps ave. That didn't seem right compared to others, is it a common experience? Is it expected to restart FSX between each benchmark? I also tested an overclock of my GTX470, 607->707Mhz. This improved the ave fps by up to 7%, but more interestingly improved the lowest fps by 33% (15 to 20). Finally testing a Vertex 2 SSD in place of a 2TB Hitatchi yielded no appreciable difference in any measurements. Next stop is sandybridge.Simon

I've not seen any, the SDK and examples seem to be regular C.

This morning it has has decided to work again, and yes the switches do switch in the window!

It looks like my Saitek switch panel has stopped working. While the gear LEDs still show up, none of the switches have any effect. For those of you that have one that works, can you confirm something for me. In the "Saitek Pro Flight Switch Panel Test Page" application available from the Window start menu, do the buttons shown in the application change according to physical switchings made on the hardware panel itself?Mine just sit there, no matter which one I try to switch.ThanksSimon

Because I want to move the sliders more to the right, use aircraft more complex than the default 737, etc.

I've been following the Sandybridge story with a lot of interest. In contemplating the upgrade, one thing concerns me about my current situation as to how much benefit I will see from an upgrade. Almost always I see CPU running at less than 100%, and by that I mean individual cores. Shouldn't the main FSX thread that typically runs on core 0 consume everything it can? If it cannot, where's the bottleneck? It does not appear to be GPU, as that's running well below 50% according to GPU-Z.Now let me qualify it a little further. I typically get around 80% core utilisation running either FPS locked or unlimited, on a vanilla fsx.cfg or highly customized. In fact the only thing that can push the utilization beyond 80% is UsePools=0.Do others see this? If I can't use all of my current CPU core capability, will Sandybridge just result in the same FPS and a lower core utilization ceiling? It doesn't sound right, but I can't figure it out.Simon