12 minutes ago, mSparks said:

If two people can do all of New York...

They mention a script. So what was their input data? OSM? Ripped off Google 3D maps?

In the article I linked, they say that for most areas of the world, this kind of data simply doesn’t exist. Just because New York might be the best documented spot on earth, it does not mean that for the other 99.999% of the earth surface the data exists too.

42 minutes ago, mrueedi said:

So what was their input data? OSM?

WED includes licenced ESRI orthophoto

also, the clue is in the name

autogen is short for "Automatically Generated"

  

4 hours ago, jarmstro said:

But that screen capture looks terrible? Like something I would have expected to see ten years ago?

Yeah, I'm not a fan of reshade, this is just fine for me, VR still knocks my socks off every time I put the headset on.

Edited June 30, 20214 yr by mSparks

49 minutes ago, mrueedi said:

They mention a script. So what was their input data? OSM? Ripped off Google 3D maps?

Although answered about the imagery, this raises another question if Xplane will introduce the ability to accept autogen placements on the gateway.

The output after all would be all Xplane content, but people will write scripts and use data as WED is already some sort of Xplane GIS tool and does accept that.

Edited June 30, 20214 yr by akita

On 6/28/2021 at 2:43 PM, Greazer said:

5. Make the engine optimized/multi-core.

With the news the other sim is getting a massive performance boost, this is critical for LR to fix.  The current FPS is just awful, after squandering 5 long years just to get onto Vulkan.

16 minutes ago, Greazer said:

With the news the other sim is getting a massive performance boost, this is critical for LR to fix.  The current FPS is just awful, after squandering 5 long years just to get onto Vulkan.

Couple of things on this.

The other sim isn't getting multithreaded systems akin to xtlua any time soon.

You can't particularly "backport" existing single threaded aircraft systems (aka SASL and Gizmo) into multithreaded, its a completely different development paradigm, akin to converting a terraced house into a royal palace - so you'll only see it maybe in entirely new aircraft developments, and none of the current developer community as far as I have seen really have the skills to do that successfully (raw multithreaded dev is orders of magnitude harder than single threaded, which was the #1 motivation for making xtlua.).

Edited July 1, 20214 yr by mSparks

The other sim will be getting about 60 FPS on Ultra with medium-high spec system.   It will not be bound to single core CPU performance anymore.  That's only the start and not even the DX12 overhaul and only optimizations.    They are doing everything to improve experience for customers of all systems including VR.  They are doing what LR could never do (until now at least). 

Edited July 1, 20214 yr by Greazer

10 hours ago, mSparks said:

WED includes licenced ESRI orthophoto

Thanks for proving my point that procedure generated scenery based on aerial images is far superior to any other data source.

And if 2 people have build New York with (guessed) 3000 buildings in (also guessed) 1 week, you can add 998 more people doing the same and the 1.5 trillion buildings globally still take more than 19000 years until the task will be accomplished. Claim disproved.

 

10 hours ago, mSparks said:

Protip, this quality of scenery

Nobody ever said, that ai gets everything right. Also this shadow detection surely is not 100% correct. But it is there, and the aerial images provide something simple as object height (just one of the many data points in my list) concistently with good accuracy and full global coverage. There exists no other way, to get that data from any other source. And the same accounts for all other data points I mentioned. Therefore the suggested gordian knots for XP wont cut it. Either global orthos will be built in or you can forget it.

 

10 hours ago, mSparks said:

Not coming to xplane ever.

Do you really think Xplane would not rock in the discipline "worst possible default scenery screenshot"?

 

 

Edited July 1, 20214 yr by mrueedi
clarification

1 hour ago, mrueedi said:

Do you really think Xplane would not rock in the discipline "worst possible default scenery screenshot"?

Worst than the competitor sim screenshot that mSparks showed? Oh God no!

  

1 hour ago, mrueedi said:

 

And if 2 people have build New York with (guessed) 3000 buildings in (also guessed) 1 week, you can add 998 more people doing the same and the 1.5 trillion buildings globally still take more than 19000 years until the task will be accomplished. Claim disproved.

 

Repeat with me mate, or at least stop assuming about a system you have no knowledge about: No sane person is placing individual trees or buildings.

Xplane agp agb ags (autogen formats) are not indivdual objects, they are complete flexible *scenes* that can take many forms and shapes based on *real data* (this includes *building heights* as well).

For cities like new york it's actually very close to perfect, it'll even perfectly place a hiking path in a park (I can upload some shots comparing to reality), for little towns with little data they will be as good as there algorithms in the render-farm (What LR uses to cut a global scenery).

WED is the official tool for Xplane world editing, it can have a *background* (with no actual content) of a ESRI or OSM.

You can place individual buildings or.. load actual data, use scripts, as you wish, as WED is designed as a GIS tool and accepts formats.

To sum up: A gateway system will help in refining a scenery placement if inaccurate, not to add a completely missing New York placing individual benches and trees.

1 hour ago, akita said:

2 hours ago, mrueedi said:

 

And if 2 people have build New York with (guessed) 3000 buildings in (also guessed) 1 week, you can add 998 more people doing the same and the 1.5 trillion buildings globally still take more than 19000 years until the task will be accomplished. Claim disproved.

 

Repeat with me mate, or at least stop assuming about a system you have no knowledge about: No sane person is placing individual trees or buildings.

I calculated the New York script case. I have not assumed manual placement. My numbers used that, so I am still right. It still is not feasible at all to get good coverage. I have studied WED, there is no trick to achieve global coverage before you and and I are long dead.

2 hours ago, Murmur said:

Worst than the competitor sim screenshot that mSparks showed? Oh God no!

Let's generate 100 random (but not on sea only) longitude/latitude pairs and put the camera there: I bet, xplane default scenery wins that discipline in the large majority, say in 98 of the 100 cases.

1 hour ago, mrueedi said:

I have studied WED, there is no trick to achieve global coverage before you and and I are long dead.

The new auto-gen and roads network features mentioned came with the last update, only about 2 months ago. They are not even available for gateway usage until whatever LR are changing is done.

It even has it's own 3d renderer now although not fully complete yet. You can load full tiles of base scenery from Xplane's own tiles or any different other source (autogen, roads, forests...)

So unless you recently checked it out, your knowledge is very much outdated.

Edited July 1, 20214 yr by akita

1 hour ago, mrueedi said:

I calculated the New York script case. I have not assumed manual placement. My numbers used that, so I am still right. It still is not feasible at all to get good coverage. I have studied WED, there is no trick to achieve global coverage before you and and I are long dead.

Just to add here.... there is no need to try to cover the world manually.. things have moved on since XP11 was released, e.g. It's now possible to find building data for most of the US, Canada, Australia, New Zealand and a lot of European countries. For other areas, it's more tricky but it is possible.

The tech now exists with enough computing power to do a lot of this automatically (detecting features etc)... depending on the availability of aerial imagery, however the biggest problem from LR or anyone else doing this venture would be trying to license worldwide aerial/satellite imagery and the resources to process it, something that MS have a huge advantage with.

5 hours ago, Greazer said:

The other sim will be getting about 60 FPS on Ultra with medium-high spec system.   It will not be bound to single core CPU performance anymore.  That's only the start and not even the DX12 overhaul and only optimizations.    They are doing everything to improve experience for customers of all systems including VR.  They are doing what LR could never do (until now at least). 

And every button in the cockpit of default aircraft will work just like the real thing: As if you are still believing the promises and marketing hype.

Last promise I saw was 30fps on XBOX series X hardware, not 60.

DX12/Vulkan and multithreading won't solve this problem anyway, when done well they solve the issue of minimums, the stutters and micro pauses. When done badly it takes 2 developers 9 weeks full time to find the cause of 1 of 200 "Not Responding" problems, and progressively longer as the list gets shorter.

MSFS now is bottlenecked on a single core for the same reason XP11 planes on SASL, Gizmo and XLUA are bottlenecked on a single core: .net WASM doesn't have any multi threading support*, and they would need to redesign everything to add it.

* Not quite true, but true enough the conclusion is unaffected.

5 hours ago, mrueedi said:

Thanks for proving my point that procedure generated scenery based on aerial images is far superior to any other data source.

Could you quote me where you were making that point earlier, I must have mis read it (Or you thought it but didn't say it?).

1 hour ago, tonywob said:

Just to add here.... there is no need to try to cover the world manually.. things have moved on since XP11 was released, e.g. It's now possible to find building data for most of the US, Canada, Australia, New Zealand and a lot of European countries. For other areas, it's more tricky but it is possible.

The tech now exists with enough computing power to do a lot of this automatically (detecting features etc)... depending on the availability of aerial imagery, however the biggest problem from LR or anyone else doing this venture would be trying to license worldwide aerial/satellite imagery and the resources to process it, something that MS have a huge advantage with.

But wouldn't this need to be streamed? How could anyone have enough personal storage for the whole world?