|
at that rate you could cache the entirety of new zealand in a pretty reasonable 1TB or so
|
# ? Nov 17, 2019 22:30 |
|
|
# ? Jun 1, 2024 15:58 |
|
Jabor posted:at that rate you could cache the entirety of new zealand in a pretty reasonable 1TB or so flight simulator 2020, hope you really like new zealand
|
# ? Nov 17, 2019 22:32 |
|
The question isn't whether you can cache it it's whether you can stream it in at a rate that isn't noticeable
|
# ? Nov 17, 2019 22:33 |
|
abigserve posted:The question isn't whether you can cache it it's whether you can stream it in at a rate that isn't noticeable well arithmetic gets us a number that works well enough to take a guess -- 16 kb/acre, give or take. i just don't know how many acres are going to be visible in any detail in a given day. i'm not a real life pilot. that's not a thing that i can ballpark
|
# ? Nov 17, 2019 22:34 |
|
actually who am i kidding, i am google strong
i think i can make this into a number of sorts so at any point in time, there is potentially 11 million acres visible https://www.wolframalpha.com/input/?i=area+of+a+circle+with+radius+235+miles%2C+in+acres so now let's look at an actual route
the strip has an area of 213.9 million acres the two hemi-circles add up to 111.0 million acres total: 324.9 million acres https://www.wolframalpha.com/input/?i=324.9+*+10**6+16+kilobytes+in+gigabytes that looks like approximately 5 TB in data to cache all of the data for a two hour flight from Chicago to New York. hope you have a 6 gbps connection at home to stream all that poo poo in 2 hours! https://www.wolframalpha.com/input/?i=5198+gb+%2F+7200+seconds+in+megabits+per+second hopefully i have way overshot on the amount of data loaded or else this is not gonna work worth poo poo.
|
# ? Nov 17, 2019 22:41 |
|
Notorious b.s.d. posted:https://www.wolframalpha.com/input/?i=2+petabytes+%2F+126.1+billion+acres+in+bytes+per+acre Notorious b.s.d. posted:that also means the scenery set for just the continental united states is about 30 TB your math is bad the wolfram alpha calculation assumes that the data is distributed evenly across the entire earth's surface. 70% is water -- i doubt they're streaming photos of every square meter of the pacific -- and only 3% of the land area is urbanized. rural areas are basically just a terrain heightmap, a photo texture, and a bunch of autogenerated trees. it's going to be much more than 16kb per acre in areas that people will be commonly flying, such as nyc or london, and much less than that over the sahara desert. also, you need to load terrain in a circle around the plane, not just in the area underneath. so the amount of data required will change with viewing distance. if i fly a cessna 152 at 100 knots on a clear day with 20-mile visibility, i need to have roughly 1 million acres in memory and i need to load/unload about 54,000 acres per minute. obviously there's some lod/mip-mapping optimizations but that's the general case. there's never been any suggestion that you'd be able to cache an "appreciable fraction of the globe." the local caching will likely be, like, one city at a time so that you can still land and take off and buzz your house without internet access. Sagebrush fucked around with this message at 22:55 on Nov 17, 2019 |
# ? Nov 17, 2019 22:47 |
|
you worked it out for yourself while i was working it out for you, but yes, you will never be able to cache all of the data at maximum resolution for the entirety of your flight. that realistically shouldn't matter because at 30,000 feet you can only barely make out individual houses by eye, let alone on a computer screen. as you fly at high altitudes you will stream a low-resolution version of the terrain that is appropriate to the viewing distance. you will need higher resolution scenery near your departure and arrival points. small aircraft flying low also generally fly slow, and the visible horizon shrinks, so the math above (50,000 acres per minute) makes it doable. if you want to fly an sr-71 over shanghai at 1000 feet and mach 3 it is plausible that you could exceed the capacity of a gigabit ethernet connection or even a local cache on an SSD. i guess we'll find out when it's released!
|
# ? Nov 17, 2019 22:53 |
|
Sagebrush posted:also, you need to load terrain in a circle around the plane, not just in the area underneath. so the amount of data required will change with viewing distance. if i fly a cessna 152 at 100 knots on a clear day with 20-mile visibility, i need to have roughly 1 million acres in memory and i need to load/unload about 54,000 acres per minute. obviously there's some lod/mip-mapping optimizations but that's the general case. yeah they probably have lower levels of details (both reduced textures and low poly models/terrain) for distant stuff that they can stream much faster and use that as place holder for the high detail scenery if it's not streamed yet. the caching probably work as a tiling system rather than per city, where the map is split into a grid. It's the easiest way to handle super large maps like those. That's what The Crew does, for instance (and asobo would be familiar with that btw since they ported it to the x360)
|
# ? Nov 17, 2019 22:59 |
|
also for any europeans who are unfamiliar with acres and think that loading 50,000 of them per minute sounds like a lot, an acre is somewhat less than half of a football pitch (soccer field). you really don't need a lot of detail in that area when flying over it at 1000+ feet (300+ meters) and 100+ knots (185+ km/h) e: actually that's a pretty good metric, because there is a soccer field under the traffic pattern at my local airport. i usually fly over it at about 85 knots and 800 feet. at that speed and altitude, i can see the markings, and i can see players on the field as slightly oblong white dots, but that's the limit of my eye resolution. Sagebrush fucked around with this message at 23:05 on Nov 17, 2019 |
# ? Nov 17, 2019 23:03 |
|
Sagebrush posted:if you want to fly an sr-71 over shanghai at 1000 feet and mach 3 it is plausible that you could exceed the capacity of a gigabit ethernet connection or even a local cache on an SSD. i guess we'll find out when it's released! Wait, do you not?
|
# ? Nov 17, 2019 23:24 |
|
Notorious b.s.d. posted:well arithmetic gets us a number that works well enough to take a guess -- 16 kb/acre, give or take. the math isn't that hard. you just need to know how far the horizon is at the elevation you're at, which you can then use as the radius of a circle to calculate the area at that time. then figure out how much of that area bounded by the horizon is different in the next second, and multiply by the flight time. for example, at 10k feet under ideal conditions you can see about 120 miles to the horizon. at one static point in time, with clear skies you will see about 45k square miles (the area of a circle, pi*radius^2) - nearly 29 million acres. the change in the area over time is half of the complement of the area of intersection between the two circles as described - the leading portion of the circle. at 500mph, a plane will travel about 700ft in a second. that math is a bit more complicated but there's a nice calculator here; over a second it's about 10k acres of new area that's visible. in an hour, it's 10k *3600. this is all assuming you're just travelling in a straight line in one direction in level flight based on that flight data, a plane travelling at 500mph, at elevation of 10k ft, over a one hour flight will sweep out a 2d cylindrical section that covers 36,000,000 acres in that hour. so roughly 36 million acres an hour. at 40k feet this would be significantly higher, as the horizon is about twice as far; you can see about 240 miles in ideal conditions, so the area is about triple. that'd give what, maybe 36 mil * 16 kb, slighly over a half terabyte of data or so an hour at 10k feet at 500mph? i double checked and i think that's right - 29million acres alone at 16kb is lot of data; at any given time it'd be nearly 500 gigs. if you were at 40k and they actually had the horizon at 240 miles it'd be even more. but yeah - that's roughly how much area you see during a flight in an hour - a lot! e: i revised this, because i hadn't considered that it was just the area of the leading edge, thus i was off by a factor of 2 Broken Machine fucked around with this message at 02:32 on Nov 18, 2019 |
# ? Nov 18, 2019 00:17 |
|
an sr-71 can't do mach 3 at 1000ft anyway, you'd melt the engines (and probably the skin as well at that altitude!) Farmer Crack-Ass fucked around with this message at 00:29 on Nov 18, 2019 |
# ? Nov 18, 2019 00:22 |
|
also, let's not underestimate the fanaticism of hardcore sim enthusiasts. "oh, i'd need a fatass multi-terabyte storage array to locally cache my favorite airline routes? i'd have an excuse to spend more money on my habit?? i'd have a reason to get gigabit internet that doesn't involve piracy??? sign me the gently caress up!! "
|
# ? Nov 18, 2019 00:28 |
|
Farmer Crack-rear end posted:an sr-71 can't do mach 3 at 1000ft anyway, you'd melt the engines (and probably the skin as well at that altitude!) i knew that and was debating putting a qualifier in my post or changing it to a more appropriate mach number but figured "oh, nobody will really notice or care"
|
# ? Nov 18, 2019 00:47 |
|
Broken Machine posted:the math isn't that hard. you just need to know how far the horizon is at the elevation you're at, which you can then use as the radius of a circle to calculate the area at that time. then figure out how much of that area bounded by the horizon is different in the next second, and multiply by the flight time. you missed the post immediately after in which i did that math lol
|
# ? Nov 18, 2019 04:36 |
|
Sagebrush posted:the wolfram alpha calculation assumes that the data is distributed evenly across the entire earth's surface. 70% is water -- i doubt they're streaming photos of every square meter of the pacific -- and only 3% of the land area is urbanized. rural areas are basically just a terrain heightmap, a photo texture, and a bunch of autogenerated trees. i deliberately used land area because i assumed a flight simulator wouldn't spend very much time on features underwater i hadn't stopped to consider urban/rural split or how much of the world is both flat and empty
|
# ? Nov 18, 2019 04:37 |
|
Notorious b.s.d. posted:i deliberately used land area because i assumed a flight simulator wouldn't spend very much time on features underwater well, you didn't, because your wolfram alpha calculation shows 126 billion acres which is the earth's entire surface area. about 36 billion acres is land. 1 billion acres is cities.
|
# ? Nov 18, 2019 05:02 |
|
Sagebrush posted:well, you didn't, because your wolfram alpha calculation shows 126 billion acres which is the earth's entire surface area. about 36 billion acres is land. 1 billion acres is cities. i blame google
|
# ? Nov 18, 2019 05:09 |
|
You are assuming the giant dataset actually has to be streamed to the clients and not that it's being preprocessed in the butt before it gets pooped out in a much more consumable way to the client. From the wording of some articles it sounds like the 2 petabytes number comes from the amount of bing imagery that was/is constantly ingested to generate the actual textures and geometry the game users - presumably the output of that will be dramatically less than the total size of the dataset.
|
# ? Nov 18, 2019 06:20 |
|
abigserve posted:From the wording of some articles it sounds like the 2 petabytes number comes from the amount of bing imagery that was/is constantly ingested to generate the actual textures and geometry the game users - presumably the output of that will be dramatically less than the total size of the dataset. that makes me feel better
|
# ? Nov 18, 2019 06:27 |
|
Broken Machine posted:the math isn't that hard. you just need to know how far the horizon is at the elevation you're at, which you can then use as the radius of a circle to calculate the area at that time. then figure out how much of that area bounded by the horizon is different in the next second, and multiply by the flight time. Another way of looking at this is they are definitely optimizing their dataset on a LOD basis much like any mapping platform. as stated 2PB is just their backend dataset. so if you look at just the angular resolution of the human eye (about 0.02 degrees) at 10kft you get linear resolution of 3.5ft and need a pixel for every 12.2sqft, or .0035 megapixels per acre. so you're instantaneous view is around 100,000 megapixels. Google tells me JPEG compression gives you around 10kB per megapixel, so 10GB fills your instantaneous view and your hourly data rate at 10kft and 500mph is only going to be 12GB
|
# ? Nov 18, 2019 17:52 |
|
Jenny Agutter posted:Another way of looking at this is they are definitely optimizing their dataset on a LOD basis much like any mapping platform. as stated 2PB is just their backend dataset. so if you look at just the angular resolution of the human eye (about 0.02 degrees) at 10kft you get linear resolution of 3.5ft and need a pixel for every 12.2sqft, or .0035 megapixels per acre. so you're instantaneous view is around 100,000 megapixels. Google tells me JPEG compression gives you around 10kB per megapixel, so 10GB fills your instantaneous view and your hourly data rate at 10kft and 500mph is only going to be 12GB that's cool they'll be able to get the bandwidth required way down, thanks for the explanation; i don't know much about computer graphics or mapping, aside from that one time i modeled an icosahedron (which can be described by three mutually intersecting golden rectangles). i was mostly just enjoying figuring out roughly how much area you'd traverse and see while flying the prospect of flying around and landing at well-modeled airstrips around the world sounds great; landing at courchevel is one of the first things i'll try
|
# ? Nov 18, 2019 18:57 |
|
Notorious b.s.d. posted:NAS chassis delivered to the user's home i once got a bunch of aerial surveillance footage sent via this method
|
# ? Nov 18, 2019 19:32 |
|
Broken Machine posted:the prospect of flying around and landing at well-modeled airstrips around the world sounds great; landing at courchevel is one of the first things i'll try you are in luck! they have shown courchevel specifically in the trailers already. the entirety of france will undoubtedly be very sharp and clean since the dev team is french (and like half the places they've shown so far are in that country) the less aviation-obsessed may recognize it as the airport with a sloped runway going off a cliff where james bond steals a plane carrying nuclear torpedoes https://www.youtube.com/watch?v=7tBPOH4aXq8
|
# ? Nov 18, 2019 19:43 |
|
So it looks pretty good obviously, but have they said anything about the actually important nerd poo poo like flight model, ATC, clickable buttons and what not? I'm perfectly fine with the graphics of Falcon 4.0 so this isn't my top criteria really.
|
# ? Nov 18, 2019 23:22 |
|
im sure theyre not loving up the table stakes part
|
# ? Nov 18, 2019 23:29 |
|
They've said that there's actually a flight model with aerodynamic modelling and not just a spherical-plane-that-goes-up-when-you-pull-the-stick sort of thing.
|
# ? Nov 18, 2019 23:56 |
|
mobby_6kl posted:So it looks pretty good obviously, but have they said anything about the actually important nerd poo poo like flight model, ATC, clickable buttons and what not? I'm perfectly fine with the graphics of Falcon 4.0 so this isn't my top criteria really. - flight model is based on a physics simulation of "hundreds" of elements of the airframe. they have said it's not quite the same as x-plane's blade element model, but that it is leagues beyond FSX's lookup table engine. the atmosphere model runs in small-scale around your plane so the interactions of flows around e.g. the landing gear are modeled correctly, instead of just saying if gear_deployed then drag+=50. a couple of previewers have said that it feels very accurate. - no specific word on how extensive the ATC will be but they have repeatedly said it will be in the game and that they know how important it is. - all cockpits are fully clickable. in game mode, rather than simplifying the controls, the switches will just flip themselves at the appropriate time. there will apparently be an assisted mode where the controls will be up to you but they will highlight themselves when you need to do something with them. if you are perfectly fine with the graphics of falcon 4 shoo grognard. get out
|
# ? Nov 19, 2019 02:53 |
|
you may not like it, but this is what peak simulation looks like
|
# ? Nov 19, 2019 03:23 |
|
What I really wanna know is.. how realistic are the crashes? I don't wanna see some cop-out text screen of "Oh, aww, you made an oopsie, we're not going to render an exploding plane just so you can get your chuckles. Close your eyes and pretend you see blossoms of explosive carnage"
|
# ? Nov 19, 2019 03:40 |
|
infernal machines posted:you may not like it, but this is what peak simulation looks like i wish xplane could hit that kind of framerate
|
# ? Nov 19, 2019 03:43 |
|
Binary Badger posted:What I really wanna know is.. how realistic are the crashes? I wouldn't get my hopes up for that, this is probably where they run into licensing issues with the airplane manufacturers who probably don't want them to depict damaged versions of their planes or graphic crashes
|
# ? Nov 19, 2019 08:30 |
|
So they'll just copy the X-Plane method which shows the plane remaining seemingly intact even after an impact at Mach speeds. Just clouds of smoke and the plane slowly sinking into the ground like it was being dissolved in acid, gotcha. maybe I'm not searching hard enough but nobody seems to have any crashing video from 2020? Edit: huh just tried Xplane on the stebephone and the Cessna breaks into three neat gibbets (that bounce like rubber balls) after a straight dive from 9000 ft. with fire and smoke at least. Binary Badger fucked around with this message at 09:58 on Nov 19, 2019 |
# ? Nov 19, 2019 09:30 |
|
otoh this thread is *probably* in general getting its hopes up for a product that'll turn out to be hopelessly flawed in a million ways, so just assume the crashes will be fully modeled, gruesome, and feature a complex combo system for injuries, deaths, and property damage.
|
# ? Nov 19, 2019 09:42 |
|
hope they simulate embarrassing phone calls you have to make to the tower after loving up a landing https://www.youtube.com/watch?v=roRhnd8kZ6A&t=53s
|
# ? Nov 19, 2019 13:11 |
|
If you can make a call, you didn't really gently caress up the landing.
|
# ? Nov 19, 2019 16:14 |
|
Sagebrush posted:- flight model is based on a physics simulation of "hundreds" of elements of the airframe. they have said it's not quite the same as x-plane's blade element model, but that it is leagues beyond FSX's lookup table engine. the atmosphere model runs in small-scale around your plane so the interactions of flows around e.g. the landing gear are modeled correctly, instead of just saying if gear_deployed then drag+=50. a couple of previewers have said that it feels very accurate. Good news: they actually have done this, down to taking into account temperature differences at each point, so if you fly just your right wing into a cloud it'll be affected realistically. They're also actually modeling atmospheric turbulence off not just terrain, but building geometry too. https://www.youtube.com/watch?v=Bw-opH4f8Qg
|
# ? Nov 24, 2019 04:22 |
|
Will you be able to import Cities Skylines cities into this?
|
# ? Nov 25, 2019 03:07 |
|
No.
|
# ? Nov 25, 2019 03:08 |
|
|
# ? Jun 1, 2024 15:58 |
|
Will there be aircraft carriers you can land and depart from on this one
|
# ? Nov 25, 2019 03:55 |