Sigmoid

Well, still I think some fine-tuning might be in order, if we have pines on the coast of Pacific islands. And though the lack of beaches lined by palm trees might be quite painful to all, the lack of all 800 geographic zones may only tick off dendrologists... but having American suburbs in Central Africa is almost Monty Python-level of absurd. And I'm not dissing X-Plane, perfect scenery clearly wasn't a goal so far. So really my question is, given that Laminar seems to be (and is professing to be) concentrating more on scenery now... is there some kind of roadmap for remedying the most painful problems with autogen? Like no socioeconomic or geographical city categorization. I won't throw a tantrum for the lack of teutonic cottages in Germany or Switzerland, but it has to begin somewhere... Say: North America, Western Europe, Eastern Europe - Russia, Middle East, East Asia, Third World? Also I wonder if the system allows for dynamic extensibility... Like if someone models the typical architecture of New York City, with lofts, X-shaped brownbrick residential highrises and somesuch, and marks up New York City into proprietary autogen zones that correspond to his specific configuration...

Hey guys, how is this improvement coming along? Any chance of it getting into 2.25?

20fps is sorry performance indeed. (Let me guess, you just built a new computer with an i3... XD)

I wonder, is there a really good model of a small, trainer class single engine airplane around? A 152, or something comparable...

That is just plain sad... XD Thinking of the bleeding edge AMD processors of roughly ten years ago...

That logic is a bit off... A hyperthreaded core does NOT match the power of two non-hyperthreaded cores. What hyperthreading does is make context switching faster, which means you get as far as I know around 20% speed improvement when lots of threads are running concurrently. That is a lot, but not double. What might be confusing is that to the operating system, a hyperthreaded core "looks like" two cores. That doesn't mean it does the job of two cores, just that the OS won't have to do the scheduling between those two threads, instead the CPU itself will take over that task, which results in less overhead, and of course better performance as a result.

Btw... i3s have hyperthreading? Good for them. This is like saying that a Smart TDI also has turbo charging... Yes it does... That doesn't put it in the same league of a 4 liter turbocharged Ferrari engine.

I remember Austin talking in a video, saying X-Plane won't be too fast on the machine he's demonstrating on, because it's a single-CPU computer with ONLY 6 cores (sic!)... I don't know how the code looks, but the problem space has plenty of room for parallelization, and when the creator is saying that it's gonna be lousy because he doesn't have 12 (!) cores, I'm inclined to think at least some of that room was capitalized upon. And even if a single flight model isn't parallelized, I find it hard to believe that the autogen would run on the same thread as the flight model. Also what about weather, ATC, and most modern applications have a separate UI thread (or several UI threads), then there is stuff the OS does in the background... Really, I find it hard to believe that a modern computer wouldn't be better off with 4 cores. (Or 12 for that matter.)

Apparently you don't know much about multicore performance. The problem space of X-Plane definitely warrants an i7. The actual code is another thing, but I think it must be decently parallelized. Blender uses the GPU for most tasks, so it's no wonder that the CPU isn't loaded. i3 is dualcore, I wouldn't really consider it for anything even mildly CPU intensive. i5 and i7 are set apart by hyperthreading, which allows an i7 CPU to switch between 2 threads on a single core at hardware level, heavily reducing the OS overhead in heavily parallelized applications like servers - or flight simulators. So you're wrong, X-Plane can squeeze every bit of power out of an i7. The only case when I'd go below that is if I didn't have enough money for one. If you have the money, it's worth it.

Are there any good scenery packs that DO NOT use sattelite or aerial photo imagery for ground texture? It's something I'm personally averse to, I prefer green grass to blotchy sat images with rooftops and treetops stretched onto the globe. I found some packs that do exactly that, are there any that don't? Also, I'm wondering if anyone uses the HD mesh. Is it still maintained?

I think right now, one should definitely wait for the new, Haswell CPUs.

It's not the roads that are wrong, it's the shoreline. The islands are missing in XPlane. http://www.openstreetmap.org/?lat=47.216&lon=-122.157&zoom=13&layers=M

That, of course, is obvious. There should be some way to write a proper validation code in a way that the computer would understand... 99% of all bridges in the world are straight or uniformly curved, single roads with both ends on dry land. They rarely terminate, split or loop back into themselves over water.

In the demo, there is a lake south of KSEA with several roads in it.... And I don't mean bridges, they are suburban cul-de-sacs. You should add a functionality to the terrain preprocessing tools, so that if roads are in water without being marked a bridge or a tunnel, it would at flash a warning or something.

My 2.2Ghz quadcore Macbook Pro does around 13 fps with everything turned up to chock full. However, I need to bring it down significantly to get a stable 25-30 fps (everything between "default" / "a ton", grit and bump maps off). So in the future, turning detail down won't make roads and buildings disappear, only make them simpler? I guess it would make a lot of sense for VFR. Seattle looks pretty rural on my computer right now.