Imagine where MSFS will be in 3 years?

[akita 429]

1 hour ago, abrams_tank said:

 Why does MSFS yield better graphics with DX11 while the other simulators can't even do that with DX12 with Vulkan?

This question is actually a legitimate one not really in a matter of simulator x vs simulator y, it's rather about a misconception that a rendering engine=rendering pipeline.

a rendering engine, to you as an end user, should be transparent, does it peak an fps number and STAY THERE? stable? blurry? CTD? how efficient is memory handling? what tools it has for predictable performance?

A rendering pipeline, is responsible for the graphical outcome of your scene. i.e. the shader stages.

The first is ultra-hard even for the most advanced programmers, the second is "hard" but progress can be fast by even an individual because most of those steps, most of those fancy shaders can be copy-pasted, AMD/Nvidia has plenty of papers and examples, but it takes a bit more time to make it part of a coherent stable product.

A new rendering pipeline is something that can't be introduced in a middle of a product run because otherwise you are changing minimum specs which really not a great treatment for many customers, typically a new fancy shade comes with hardware that can run it then it is made to be compatible for previous generations (i.e., ray tracing shades can run on non RTX cards), this is why flight simulators in general are not early graphics tech adopters, most of the users do not have an RTX3090 and 16 cores, unlike a typical AAA gamer which will get the newest generation of cards on year cycle.

Edit: finally found it, a great example of how stuff look without the fancy pipeline shaders: https://www.fsdeveloper.com/forum/threads/bgldec-a-resample-bgl-decompressor.433789/page-6

comment#116 if not directed.

Edited April 13, 2021 by akita

[mrueedi 702]

 

12 hours ago, akita said:

The better the data, (and then how you walk it through the pipeline) the seamless it's going to be compared to ortho-imagery, i.e. will almost have ortho patterns but with 10x higher resolution world, in which you can define the smallest rocks.

What source defines where the small rock is located? There is no global database covering these details you mention. So another AI has to invent it. Which makes your approach inferior. You only get the world, how it "could" look like. But not how it really looks like. On an ortho image AI could extract the size, the shape, the location and the color of the rock on the other hand. So you not only get the detail, but these details also match real world. If you look at the blunt presentation of a city in your video, you can imagine that even if you generate details here and there (say a children playground, construction sites, hedges, bushes, fences) these details will not be at the correct location. If you speak of fine granular details, its impossible to ever have a worldwide database of all of these. Even getting just the shapes of all the houses and the locations of all trees globally (which MSFS already got) from any other source than images is an utopical illusion. For MSFS 99.8% of the details about the buildings, vegetation and infrsatructure had to be extracted from the images (for the lack of better data sources). Read here how the images are the best thinkable source to realisticly get the relevant data for the whole world.

Btw. the baked in details on ortho imagery can be removed rather easily by AI. MSFS is doing it. A lot of processing happened to get what we have. As the shadows of the buildings have been detected by Blacksharks AI in order to determine the building height, they are also able to remove the shadows out of the images. In MSFS they basically "debaked" everything. MSFS has the details at the correct locations and it has the orthos with hardly anything baked in and is populating the scenery with detailed 3D objects. MSFS ticks all the boxes. Your approach not.

Edited April 13, 2021 by mrueedi

[honanhal 1,420]

57 minutes ago, OzWhitey said:

They've got a good-looking engine, but performance is poor. Trying flying in VR with the clouds on normal settings some time, it's not pretty. Maybe DX12 will fix this, who knows.

I don't understand this. MSFS performance is much better than P3D v4.5 for me -- I mean, not even in the same ballpark. I liked a lot of things about P3D, but performance was absolutely not one of them.

To be clear, I believe you when you say you're seeing worse performance. I just can't work out why. My best guess is that your incredibly strong system (some might say "ludicrously overpowered" 🙂) achieves that elusive good performance in P3D by cracking a nut with a sledgehammer (5 GHz CPU! 2080 Ti!) but that in MSFS a lot of that is more or less wasted since it's much more GPU dependent. Personally, I'm thrilled that I no longer have to chase insanely high CPU speeds for simming, and that I don't have to constantly worry about cooling, but maybe that's just me.

As for the rest of this discussion, I always find the back and forth on ortho bizarre. No one is saying MSFS was the first to introduce the idea of satellite imagery with autogen, just that they took that idea and did it better than everyone else. There's no comparative advantage to the X-Plane system, and a lot of disadvantages. MSFS clearly doesn't do everything perfectly, but the combination of Azure and streaming imagery are genuine triumphs, and I'm frankly mystified that anyone would deny that in good faith. If I were still using P3D, I'd be poking LM to work on something similar.

James

[crimplene 631]

1 hour ago, OzWhitey said:

They've got a good-looking engine, but performance is poor. Trying flying in VR with the clouds on normal settings some time, it's not pretty. Maybe DX12 will fix this, who knows.

Strange, because X-Plane performance is really much worse. With ultra detail in 4K I always get smooth 25 to 40 fps with MSFS. I tried X-Plane with Vulcan in 4k a few days ago. Maxing out the detail gives fps in the low teens and it still looks like word not allowed 

[akita 429]

26 minutes ago, mrueedi said:

 

What source defines where the small rock is located? There is no global database covering these details you mention. So another AI has to invent it. Which makes your approach inferior. You only get the world, how it "could" look like. But not how it really looks like. On an ortho image AI could extract the size, the shape, the location and the color of the rock on the other hand.

You are right if we had a global ZL50 that can capture those rocks in a breakwater, while otherwise, those rocks can be generated using displacements and tessellation shaders from a much higher res base material.  Colors in natural resources in this world changes based on specific known rules (seasons and climate as an exapmle), you see a green field in bing, while in reality it is like that for only 2 months of the year. (and all the area might be already 5 years outdated).

So both has shortcomings but I prefer much sharper art work.

Also to say the least MSFS AI does not yet cover every ortho baked detail, there are plenty, it works in "dense" areas so otherwise buildings and grass may cover, but for the most part I still noticed a lot of them, including in photogrammetry (which looks good only if the building is mostly built from quads, anything more complex like antenna, chimney, electric strings and more it fails)

[Moria15 950]

3 minutes ago, crimplene said:

Strange, because X-Plane performance is really much worse. With ultra detail in 4K I always get smooth 25 to 40 fps with MSFS. I tried X-Plane with Vulcan in 4k a few days ago. Maxing out the detail gives fps in the low teens and it still looks like word not allowed 

I don't know how relevant this is but I have a solid system that runs both fine..  however.. I have to setup the machine differently for each..  For P3D, even with this system, I have best performance with overclocking the CPU/GPU and mem..  however overclocking significantly downgrades my MSFS performance and with zero overclocking in MSFS I get a much more stable and smooth performance with everything on ULTRA and running a 2k monitor.  If I have overclocking on in MSFS I struggle to get it to run at HIGH smoothly.

Now with the ways that PC's vary.. that could well be unique to my setup but for me it is significant.

Graham

[OzWhitey 3,353]

2 hours ago, virtuali said:

The one and only reason for the ACES closure was the 2008 financial crisis, when MS shareholders demanding jobs being cut, so the whole PC gaming division was dismissed.

We're getting off topic, but that's way oversimplified. You mention a factor in passing later in your post - Steve Ballmer. Gates was a flightsim fan and protected the franchise, Ballmer wanted to be rid of it once he was boss.

Now, the performance issue leading to ACES cancellation is hard to prove, but the vibe amongst simmers at the time was VERY hostile towards Microsoft (which is supported by your posts). The ACES guys used to come here occasionally, until the reaction got too toxic. My opinion back then was that FSX was selling a lot of copies - it got some pretty heavy marketing - but was also disappointing a lot of people because it was S L O W, and buggy to boot. A Windows PC running FS9 made MS look good. Early-era FSX, not so much. I personally think this was part of what pushed ACES studio over into the "liabilities" category.

Now, history has shown that ACES actually built a good foundation that despite a few stupid bugs has stood the test of time very well indeed. It's a shame they never lived to see its success!

[abrams_tank 7,387]

30 minutes ago, OzWhitey said:

We're getting off topic, but that's way oversimplified. You mention a factor in passing later in your post - Steve Ballmer. Gates was a flightsim fan and protected the franchise, Ballmer wanted to be rid of it once he was boss.

Now, the performance issue leading to ACES cancellation is hard to prove, but the vibe amongst simmers at the time was VERY hostile towards Microsoft (which is supported by your posts). The ACES guys used to come here occasionally, until the reaction got too toxic. My opinion back then was that FSX was selling a lot of copies - it got some pretty heavy marketing - but was also disappointing a lot of people because it was S L O W, and buggy to boot. A Windows PC running FS9 made MS look good. Early-era FSX, not so much. I personally think this was part of what pushed ACES studio over into the "liabilities" category.

Now, history has shown that ACES actually built a good foundation that despite a few stupid bugs has stood the test of time very well indeed. It's a shame they never lived to see its success!

My memory of what happened to FSX is similar to virtuali's memory.  That Aces was cut not because of poor sales of FSX, but because Microsoft was restructuring.  I believe it was Phil Taylor, who was pretty high up in the Aces studio an on the FSX project, who said this in his blog.  I can't seem to find that blog now - it seems to have been deleted.  But I do remember Phil Taylor running his own blog and commenting on the closure of Aces studio and why it was closed.  It's been such a long time that I read that comment from Phil Taylor, but I think Phil Taylor remarked that the sales of FSX was quite good, even though Aces Studios was cut.

Edited April 13, 2021 by abrams_tank

[OSM 683]

Microsoft Flight Simulator 2020 is the Best Flight Simulator ever, with the most advanced Graphics and Physics!!! 😃👍

Beatton River, Fort St John, British Columbia, Canada (56,2989° N, 120,7349° W)

 

[Nuno Pinto 1,394]

Right now, in 3 years either in the bin or it has actually become a simulator. Doesn't look good though...

[virtuali 4,419]

3 hours ago, OzWhitey said:

but that's way oversimplified. You mention a factor in passing later in your post - Steve Ballmer. Gates was a flightsim fan and protected the franchise, Ballmer wanted to be rid of it once he was boss.

The "quality" of FSX doesn't have anything to with the job cuts at Microsoft, which were of course caused entirely by the 2008 financial crisis, where shareholders demanded to cut some fat from Microsoft. ACES were less than 200 people out of 5000 which were laid off in January 2009.

So no, nothing to do with just FSX, it was caused by the generic slowing down of PC sales because of the crisis, other studios in additions to ACES were closed, like Ensemble studios, who made Age of Empires ( a very successfull game that sold about 25 milions copes in total ) and Halo Wars.

Edited April 13, 2021 by virtuali

[ADamiani 470]

14 hours ago, VFXSimmer said:

If that is the case then it just sounds like the sentient self driving car doesn't like you or feel you're worthy to be driven.  I'm sure your model-T still works tho.

Ok so the buyer is always the culprit. 

Easy.

A.

[ADamiani 470]

16 hours ago, mrueedi said:

You sound as if you were forced to buy it. Many can get tremendous value from it. But if it has no value for you, you should not buy it. Every buyer is reponsible to buy only stuff he likes. Especially in the area of hobbies.

If the covered functional scope does not meet your needs, you should pick another product, which suits you better. Sometimes people seem to think MS is obliged to build MSFS to suit their personal requirements.

I bought it after trying the XBOX pass version because I liked what I saw. THAT is what I bought. Then they started destroying it.

Isn't it funny? You test drive a new product, you like it, so you buy it. Then the producer starts adding bugs, for no apparent reason.

And please do not divert: I am not asking for something to my linking. I just want the product that I tested and was advertised by the producer.

That's all.

A.

[RobJC 2,069]

On 4/11/2021 at 6:49 AM, mrueedi said:

Absolutely, every complex software ships with hundreds if not thousands known bugs at any time. The bug tracking tools of big open source projects like Blender, Visual Studio Code or even the Linux Kernel are a glaring proof of that. A sea of bugs, often raised many years ago. Bugs, which will very likely never will be fixed. Dont expect anything else with MSFS.

Another distinction, which is not easily understood by many, is the difference between a bug and missing functional scope. People often are not happy with a product, because the delivered functional scope does not meet their needs. But the funtional scope per release is something, which is planned very early in each development cycle. E.g. whether the shipped functionality in the area of ATC is buggy or simply matches the somewhat limited yet planned functional scope for that release, you and I bascially are not able to tell. Only MS product management knows. Wrong ATC phrases could be a 100% compliant implementation of the requirement "Implement ATC on FSX level with some low effort enhancements". In many cases only a comparison with the requirement (which we dont have) can tell, whether a deficiency is a bug or simply lacking (but planned) functional scope. The former needs bug fixing, the later extensions. This distinction is crucial for devs.

What also needs to be considered is the life cycle of a complex software. I was deeply involved in a 200-man software development project and the first release only covered about 15% of the functional scope the system has today. Year by year another 15% were added, so today we are more or less feature complete.

Now apply the same to MSFS:

Imagine the perfect PC flight simulator having all the bells and whistles, you could ever wish or dream of. Now, with that near-utopical functional scope in mind, everybody can sketch a development plan, which a vendor would have to follow to accomplish that goal. Realisticly, the implementation of that plan and approaching that ultimate goal would take many years (10+) of dedicated work. If MSFS would have reached a state of wholeness right at launchtime or even after one year, where would be room for extension? Or, to put it more tangible, if ATC, the weather engine, the default planes, the system deepth, helicopters, the SDK, seasons, sailplanes, water masks and everything else already would be perfect, there would be no more work for Asobo or MS. No, it does not work that way. At some time, you release a software with considerable functional limitations all while the further development continues in parallel with ramping up a userbase.

 

Which sim isn't? Please unistall your mods on XPlane and repeat the comparison!

Well said. You obviously speak from experience. 

[OzWhitey 3,353]

3 hours ago, abrams_tank said:

My memory of what happened to FSX is similar to virtuali's memory.  That Aces was cut not because of poor sales of FSX, but because Microsoft was restructuring.  I believe it was Phil Taylor, who was pretty high up in the Aces studio an on the FSX project, who said this in his blog.  I can't seem to find that blog now - it seems to have been deleted.  But I do remember Phil Taylor running his own blog and commenting on the closure of Aces studio and why it was closed.  It's been such a long time that I read that comment from Phil Taylor, but I think Phil Taylor remarked that the sales of FSX was quite good, even though Aces Studios was cut.

No one has argued that it was due to poor sales, in any post in this thread. The sales were good, but community reaction was very negative (much more so than when fs9 was released). Cyberpunk 2077 has good sales, but the product has very much damaged cdpr’s reputation. I don’t know for sure, but i think a stuttering 8 fps sim with scenery menus that don;’t even work (seriously??) was not buying ms a lot pf good will, hence easy to cut when downsizing time arrived.