Benchmarked .3ds Models by File Size
Posted: 01.02.2006, 01:24
Benchmarked .3ds Models by File Size
The Venus 90x models were created today using three different mesh spacings. Celestia benchmarks for FPS, frames per second, were recorded. The Magellan Spacecraft database was used for topographical elevations, with gaps filled in by Venera and Pioneer data. All three models had 90x exaggeration of elevations. The Nvidia adapter has 64 Meg RAM. It is a GeForce 4 MX . It is used on my AMD Athelon at 2.21 GHz with 2 Gig of RAM.
The models were made by a Perl program to make .stl file for stereolithography. The .stl files were converted to .3ds files using Accutrans3D by MicroMouse Productions. Star magnitude was 15, no galaxies enabled. Venus_90x radius = 32km. Distance = 4 radius. Field of view 26 degrees, Venus_90x almost touches two edges.
Here are the results :
Not looking at Venus_90x : Frame rate = 59.991 FPS
1 degree grid of latitude and longitude for Venus_90x_1
File size 2,017 kbytes
Number of triangles in mesh 128,522
Frame rate = 14.997 FPS
1/2 degree grid of latitude and longitude for Venus_90x_2
File size 8,238 kbytes
Number of triangles in mesh 516,242
Frame rate = 5.183 FPS
1/3 degree grid of latitude and longitude for Venus_90x_3
File size 18,780 kbytes
Number of triangles in mesh 1,163,162
Frame rate = 2.400 FPS still or rotating
Two 1/3 degree grid models of 18,780 k each
File size sum = 37,560 kbytes
Frame rate = 1.224 FPS
Three 1/3 degree grid models of 18,780 k each
File size sum = 56,340 kbytes
Frame rate = 0.822 FPS
Four 1/3 degree grid models of 18,780 k each
File size sum = 75,120 kbytes
Frame rate 0.618 FPS
As soon as the 4 Venus_90x are off screen, the frame rate returns to 60 FPS. Looking at them from 61,000 km away, 60 FPS because they are 4 dots together. At 48,000km away 4 FPS as 4 dots separate.
At 41,000 km .619 FPS as 4 dots are clearer than any star. (Stars are scaled disks).
_________________________________________________________
Note 1
See last post in the thread, by rthorvald :
http://www.celestiaproject.net/forum/viewtopic ... s&start=15
where he proposed benchmarks. His proposed benchmarking was supported by Maxim and ElChristou, but models were not available.
Note 2
Higher resolution models were also created, but the file sizes were too big to be converted from .stl to .3ds using Accutrans3D. The Venus_90x_4 with 1/4 degree grid has a .stl file size of 101 megabytes. The Venus_90x_6 with 1/6 degree grid has a .stl file size of 227 megabytes. If they were converted to .3ds their sizes would have been 33 megabytes and 75 megabytes. The sinusoidal projection database from Magellan allows Venus grids a small as 1/22 degree. Earth goes down to 1/30 degree on ETOPO2. Mars MGS MOLA goes down to 1/128 degree grids. At that resolution, a Mars .stl file would be 100 Gigabytes.
The Moon topographical database, published in 1996, is severely flawed. The Mercury files are coming in 2011.
Note 3
An exaggeration factor of 90 was used for this work on Venus. It is not useful for Celestia to use a 3D model of a planet with unexaggerated elevations. A planet's 3D model would look like a smooth sphere if the heights were not exaggerated, when limited to a grid of 1/3 degree. Only small moons and asteroids have surface shapes that have enough relative variation to be modeled with no exaggeration, and still provide a noticeable terrain.
The Venus 90x models were created today using three different mesh spacings. Celestia benchmarks for FPS, frames per second, were recorded. The Magellan Spacecraft database was used for topographical elevations, with gaps filled in by Venera and Pioneer data. All three models had 90x exaggeration of elevations. The Nvidia adapter has 64 Meg RAM. It is a GeForce 4 MX . It is used on my AMD Athelon at 2.21 GHz with 2 Gig of RAM.
The models were made by a Perl program to make .stl file for stereolithography. The .stl files were converted to .3ds files using Accutrans3D by MicroMouse Productions. Star magnitude was 15, no galaxies enabled. Venus_90x radius = 32km. Distance = 4 radius. Field of view 26 degrees, Venus_90x almost touches two edges.
Here are the results :
Not looking at Venus_90x : Frame rate = 59.991 FPS
1 degree grid of latitude and longitude for Venus_90x_1
File size 2,017 kbytes
Number of triangles in mesh 128,522
Frame rate = 14.997 FPS
1/2 degree grid of latitude and longitude for Venus_90x_2
File size 8,238 kbytes
Number of triangles in mesh 516,242
Frame rate = 5.183 FPS
1/3 degree grid of latitude and longitude for Venus_90x_3
File size 18,780 kbytes
Number of triangles in mesh 1,163,162
Frame rate = 2.400 FPS still or rotating
Two 1/3 degree grid models of 18,780 k each
File size sum = 37,560 kbytes
Frame rate = 1.224 FPS
Three 1/3 degree grid models of 18,780 k each
File size sum = 56,340 kbytes
Frame rate = 0.822 FPS
Four 1/3 degree grid models of 18,780 k each
File size sum = 75,120 kbytes
Frame rate 0.618 FPS
As soon as the 4 Venus_90x are off screen, the frame rate returns to 60 FPS. Looking at them from 61,000 km away, 60 FPS because they are 4 dots together. At 48,000km away 4 FPS as 4 dots separate.
At 41,000 km .619 FPS as 4 dots are clearer than any star. (Stars are scaled disks).
_________________________________________________________
Note 1
See last post in the thread, by rthorvald :
http://www.celestiaproject.net/forum/viewtopic ... s&start=15
where he proposed benchmarks. His proposed benchmarking was supported by Maxim and ElChristou, but models were not available.
Note 2
Higher resolution models were also created, but the file sizes were too big to be converted from .stl to .3ds using Accutrans3D. The Venus_90x_4 with 1/4 degree grid has a .stl file size of 101 megabytes. The Venus_90x_6 with 1/6 degree grid has a .stl file size of 227 megabytes. If they were converted to .3ds their sizes would have been 33 megabytes and 75 megabytes. The sinusoidal projection database from Magellan allows Venus grids a small as 1/22 degree. Earth goes down to 1/30 degree on ETOPO2. Mars MGS MOLA goes down to 1/128 degree grids. At that resolution, a Mars .stl file would be 100 Gigabytes.
The Moon topographical database, published in 1996, is severely flawed. The Mercury files are coming in 2011.
Note 3
An exaggeration factor of 90 was used for this work on Venus. It is not useful for Celestia to use a 3D model of a planet with unexaggerated elevations. A planet's 3D model would look like a smooth sphere if the heights were not exaggerated, when limited to a grid of 1/3 degree. Only small moons and asteroids have surface shapes that have enough relative variation to be modeled with no exaggeration, and still provide a noticeable terrain.