1,746
edits
(1) Add 1 already determined high settings hardware/performance profile for 2018.3 LTS - as the new 1st rollout OSM2City worldbuild is harder to characterise. 2) add some quick notes for hardware recoomendations in 2021 telling people to wait until WS3 if possible and look at the 2018.3 section.) |
m (cleanup) |
||
| Line 16: | Line 16: | ||
* People buying new hardware with FlightGear in mind are encouraged to wait until LTS ''afte''r the current 2020.3 LTS if possible - if they imagine they won't be replacing the hardware for a while. | * People buying new hardware with FlightGear in mind are encouraged to wait until LTS ''afte''r the current 2020.3 LTS if possible - if they imagine they won't be replacing the hardware for a while. | ||
*# Future performance requirements will not be known until the LTS after the current LTS. Some of the big projects underway on the FlightGear next branch need to be finished before requirements will be known. These projects include moving to the new [[Compositor]] rendering framework, next generation scenery rendering and data format, higher version of OpenGL, in-sim UI replacement, and so on. | *# Future performance requirements will not be known until the LTS after the current LTS. Some of the big projects underway on the FlightGear next branch need to be finished before requirements will be known. These projects include moving to the new [[Compositor]] rendering framework, [[World Scenery 3.0 roadmap|next generation scenery]] rendering and data format, higher version of OpenGL, in-sim UI replacement, and so on. | ||
*# Hardware prices will be expected to drop once the situation with the pandemic eases. Hardware is getting steadily faster. Hardware with the same performance is getting cheaper. You should never buy hardware for a future version of FlightGear or any other program. It's always better to wait until the application releases and you have time to start using it properly, | *# Hardware prices will be expected to drop once the situation with the pandemic eases. Hardware is getting steadily faster. Hardware with the same performance is getting cheaper. You should never buy hardware for a ''future'' version of FlightGear or any other program. It's always better to wait until the application releases (and you have time to start using it properly to justify purchases), | ||
*# There is a 1st rollout of buildings, roads, and objects for the world based on OSM data. Please don't buy hardware based on this rollout. The performance in future rollouts will improve drastically as objects are changed to a newer format. See [[OSM2City 1st Worldbuild]] for details. | *# There is a 1st rollout of buildings, roads, and objects for the world based on OSM data. Please don't buy hardware based on this rollout. The performance in future rollouts will improve drastically as objects are changed to a newer format. See [[OSM2City 1st Worldbuild]] for details. | ||
* For current hardware performance (with the OSM2City world-build off): FlightGear 2020.3 is faster than the 2018.3 LTS. The 2018.3 hardware profile below is valid - but only one profile and hardware build is available in that section as of writing this. | * For current hardware performance (with the OSM2City world-build off): FlightGear 2020.3 is faster than the 2018.3 LTS. The 2018.3 hardware profile below is valid - but only one profile and hardware build is available in that section as of writing this. | ||
| Line 27: | Line 27: | ||
''Note on hardware requirements'' | ''Note on hardware requirements'' | ||
Requirements are only valid for a target FPS, monitor resolution, usage, and settings performances. People have a huge range of hardware. For flight simulators the usage can vary depending on the craft used, how the craft is used, and over what scenery the craft is used. The visibility range can change with altitude, how fast the scenery is loaded can change with speeds, how complex the visuals are depends on how much vegetation or man-made buildings are in the scenery. The following all place different demands: slow and low with a balloon/[[pterosaur]]/glider/wingsuit/ground effect | Requirements are only valid for a target FPS, monitor resolution, usage, and settings performances. People have a huge range of hardware. For flight simulators the usage can vary depending on the craft used, how the craft is used, and over what scenery the craft is used. The visibility range can change with altitude, how fast the scenery is loaded can change with speeds, how complex the visuals are depends on how much vegetation or man-made buildings are in the scenery. The following all place different demands: slow and low with a balloon/[[pterosaur]]/glider/wingsuit/ground effect vehicle, slightly faster and higher in a single prop [[Cessna 172P|C172P]]/[[Cessna 182S|C182]]/[[Piper J3 Cub|Cub]]/[[Piper PA28 Warrior II|PA]], fast and low in a supersonic [[Saab 37 Viggen|Viggen]]/[[Grumman F-14 Tomcat|F-14]] jet, higher altitude but slower in a [[Airbus A320 Family|A320]] airliner, extremely fast and high in suborbital craft like the [[X-15]] , or using a [[Space Shuttle|Shuttle]] or [[Vostok-1]] space craft that is fast enough to reach orbit. View distances can vary from [[:File:EC135 Heer.jpeg|individual grass blades]] in airport keep to seeing [[:Category:Screenshots of Earthview|entire continents]] from orbit using an orbital renderer. Flying over a dense forest can mean [[:File:Alouette-lll over the Carpathian Mountains in Romania in Autumn (Flightgear 2018.x).jpg|vast]] amounts of trees are in view, while flying over a desert or ocean can mean there are few objects. | ||
Without target FPS, monitor resolution, usage, and settings, requirements are meaningless other than for maximum settings (i.e. meaningless for what hardware people actually have) - even maximum settings are hard to predict as they depend on what scenery a flight can take place. It's not trivial to find exact specifications, even for commercial applications with lots of test systems. Some commercial 3d applications do give out specifications for less than max settings but these mostly do not give target performance/quality/usage at all, and a lot of 3d applications like games have limited usage patterns making it easier to check - since commercial screenshot galleries tend to be on the most powerful systems at time of release without stating hardware, not giving detailed performance/quality/usage targets is also a way of making people forget quality may not look anything like advertised on their hardware. The topic of quality and performance is not as simple as commercial applications make it out to be in their advertising. People should consider that a bit of time spent tweaking settings for the performance bottlenecks in your hardware, for your typical usages, can make a huge difference to visuals and FPS. | Without target FPS, monitor resolution, usage, and settings, requirements are meaningless other than for maximum settings (i.e. meaningless for what hardware people actually have) - even maximum settings are hard to predict as they depend on what scenery a flight can take place. It's not trivial to find exact specifications, even for commercial applications with lots of test systems. Some commercial 3d applications do give out specifications for less than max settings but these mostly do not give target performance/quality/usage at all, and a lot of 3d applications like games have limited usage patterns making it easier to check - since commercial screenshot galleries tend to be on the most powerful systems at time of release without stating hardware, not giving detailed performance/quality/usage targets is also a way of making people forget quality may not look anything like advertised on their hardware. The topic of quality and performance is not as simple as commercial applications make it out to be in their advertising. People should consider that a bit of time spent tweaking settings for the performance bottlenecks in your hardware, for your typical usages, can make a huge difference to visuals and FPS. | ||
| Line 59: | Line 59: | ||
* '''GPU''' (Graphics Processing Unit, or Graphics Card): A GTX 960 or 1050 TI - or equivalent performance. More than 3000 ratings in this list: as of [http://web.archive.org/web/20210412095421/https://benchmarks.ul.com/compare/best-gpus April 2021]. [https://benchmarks.ul.com/compare/best-gpus Direct link] (Note: ratings numbers and list may have changed by the time you see it). | * '''GPU''' (Graphics Processing Unit, or Graphics Card): A GTX 960 or 1050 TI - or equivalent performance. More than 3000 ratings in this list: as of [http://web.archive.org/web/20210412095421/https://benchmarks.ul.com/compare/best-gpus April 2021]. [https://benchmarks.ul.com/compare/best-gpus Direct link] (Note: ratings numbers and list may have changed by the time you see it). | ||
** Notes: It's overlays and tree density that ask | ** Notes: It's overlays and tree density that ask | ||
* '''CPU''': 4 core Intel Sandybridge i5 2500 (released 2012) or faster. This build will be '''limited by the CPU''' . Try to get a newer generation Intel CPU, or AMD Ryzen CPU, if possible - e.g. a | * '''CPU''': 4 core Intel Sandybridge i5 2500 (released 2012) or faster. This build will be '''limited by the CPU''' . Try to get a newer generation Intel CPU, or AMD Ryzen CPU, if possible - e.g. a Haswell (4xxx) or newer generation Intel i5 CPU. For a fairly approximate idea of CPU performance needed (i5 2500 has a rating of ~1700) see this archived copy of a single-core benchmark list: [http://web.archive.org/web/20210412100323/https://www.cpubenchmark.net/singleThread.html April 2021] . [https://www.cpubenchmark.net/singleThread.html Direct link] (Note: this direct list may have changed ratings numbers by the time you see it). This list only gives single-core performance - and it is approximate - for example, when using more than one core processor frequency will drop slightly.The only reason a 2012 Sandybridge 2500 can manage 20-30FPS is because core performance hasn't gone up that much - it has increased by about 2x. FlightGear uses multiple cores. Some people may be able to get FG to run fine with a fast high clock 2 core CPU from a newer generation of CPU. But a 4 core is recommended as future FG will use cores even more. | ||
* '''Memory''' (RAM): 8GB (12-16 GB preferred). If you lack RAM you need to reduce the level of detail ranges so less scenery is displayed. | * '''Memory''' (RAM): 8GB (12-16 GB preferred). If you lack RAM you need to reduce the level of detail ranges so less scenery is displayed. | ||
* '''Disk:''' to do! . | * '''Disk:''' to do! . | ||
| Line 65: | Line 65: | ||
=== FlightGear settings tips: === | === FlightGear settings tips: === | ||
* This build is '''limited by CPU'''. If your CPU is not faster (newer generation), at least you can try turning up graphics more to give the GPU something to do (try increasing Anti-aliasing, turning up | * This build is '''limited by CPU'''. If your CPU is not faster (newer generation), at least you can try turning up graphics more to give the GPU something to do (try increasing Anti-aliasing, turning up supersampling transparency anti-aliasing, turning up tree density). | ||
* GPU limited: Try turning down the resolution and [[anti-aliasing]] if you are GPU limited. The higher the resolution the more work the GPU must do. A 4k (UHD) resolution is like the pixel shader load of rendering on four 1080p screens at once! See [[Anti-aliasing#The%20trade%20off%20between%20graphics%20content.2C%20FPS.2C%20and%20graphics%20settings%20like%20anti-aliasing|The trade off between graphics content, FPS, and graphics settings like anti-aliasing]] | * GPU limited: Try turning down the resolution and [[anti-aliasing]] if you are GPU limited. The higher the resolution the more work the GPU must do. A 4k (UHD) resolution is like the pixel shader load of rendering on four 1080p screens at once! See [[Anti-aliasing#The%20trade%20off%20between%20graphics%20content.2C%20FPS.2C%20and%20graphics%20settings%20like%20anti-aliasing|The trade off between graphics content, FPS, and graphics settings like anti-aliasing]] | ||
* GPU: For '''4k monitors''': Try reducing resolution to 1080p. You should try the equivalent of a GTX 1080/2070 or a rating of about 7800 in this archived list, but you may have to tweak settings a bit: [http://web.archive.org/web/20210412095421/https://benchmarks.ul.com/compare/best-gpus April 2021]. [https://benchmarks.ul.com/compare/best-gpus Direct link] (Note: ratings numbers and list may have changed by the time you see it). See the advice on the anti-aliasing page. | * GPU: For '''4k monitors''': Try reducing resolution to 1080p. You should try the equivalent of a GTX 1080/2070 or a rating of about 7800 in this archived list, but you may have to tweak settings a bit: [http://web.archive.org/web/20210412095421/https://benchmarks.ul.com/compare/best-gpus April 2021]. [https://benchmarks.ul.com/compare/best-gpus Direct link] (Note: ratings numbers and list may have changed by the time you see it). See the advice on the anti-aliasing page. | ||
| Line 76: | Line 76: | ||
* GPU: ALS shaders sliders can be maxed in a lot dedicated GPUs, even quite old ones provided they are dedicated GPUs (and not integrated/mobile GPUs). It's mostly the scenery layer settings including tree density and LoD ranges that can be intensive. | * GPU: ALS shaders sliders can be maxed in a lot dedicated GPUs, even quite old ones provided they are dedicated GPUs (and not integrated/mobile GPUs). It's mostly the scenery layer settings including tree density and LoD ranges that can be intensive. | ||
* GPU / Disk space: turn on the <code>View > rendering > texture cache</code> for smoother FPS frame spacing and faster loading. This will take up extra space. | * GPU / Disk space: turn on the <code>View > rendering > texture cache</code> for smoother FPS frame spacing and faster loading. This will take up extra space. | ||
* ...what does this look like? Only the screenshots that have been uploaded to the wiki are available right now. There is a category with '''roughly''' [[:Category:Screenshots at high settings|high settings screenshots]] - a lot of these are below this setting, the overlays ones typically have transparency anit0aliasing higher than multisampling, some screenshots are with a much older Flightgear, a substantial amount of screenshots laptops or computers with less than 16 GB of RAM, some screenshots are at higher settings a bit than this profile or at 4k resolution.To get to the highest settings in the category at 1080p and trees at ultra, you would need a GTX 1060 or higher (about [http://web.archive.org/web/20210314111800/https://benchmarks.ul.com/compare/best-gpus 4100] in ratings), 12-16 GB of RAM - lowering LoD ranges will help (low ranges are fine for low altitude flights especially in mountain areas). Not turning up transparency AA will help. The CPU would be maybe a bit faster than i5 2500 to populate trees quickly at ultra e.g. a fast | * ...what does this look like? Only the screenshots that have been uploaded to the wiki are available right now. There is a category with '''roughly''' [[:Category:Screenshots at high settings|high settings screenshots]] - a lot of these are below this setting, the overlays ones typically have transparency anit0aliasing higher than multisampling, some screenshots are with a much older Flightgear, a substantial amount of screenshots laptops or computers with less than 16 GB of RAM, some screenshots are at higher settings a bit than this profile or at 4k resolution.To get to the highest settings in the category at 1080p and trees at ultra, you would need a GTX 1060 or higher (about [http://web.archive.org/web/20210314111800/https://benchmarks.ul.com/compare/best-gpus 4100] in ratings), 12-16 GB of RAM - lowering LoD ranges will help (low ranges are fine for low altitude flights especially in mountain areas). Not turning up transparency AA will help. The CPU would be maybe a bit faster than i5 2500 to populate trees quickly at ultra e.g. a fast Haswell i5, or an overclocked Sandybridge i5 2500k @ close to 5GhZ. | ||
== Recommended hardware for FlightGear 3.20+ == | == Recommended hardware for FlightGear 3.20+ == | ||
edits