Hi. It's me again. Im not sure is this place to put such data :)
Additional report on running this game 800x600 on TC1100 Pentium M 1.1GHz and Elitebook Core 2 Duo 1.86GHz IN SHORT: Seems that 1GHz CPU is "baseline"for this game to be playable 60FPS drops to 20-30FPS And 500MHz may be limitation for this game to be playable (15-30FPS) And 500Mtex, 500Mpixel, 30mpoly decent GPU made > 2000, or low end > 2005
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Pentium M is combined to GeForce 4 Go 420 low end in 2004 more or less 500Mtexel, 500Mpixel per secon GPU
More or less https://www.techpowerup.com/gpu-specs/geforce4-go-4200.c1558
While Core 2 Duo to Intel GPU ~4000Mtex(huge), 500Mpixel(same) regions with only 125Mpolygons per second https://www.techpowerup.com/gpu-specs/gma-x3100.c1233
Performance of for example old Radeon 9550 AGP was 1500Mtex and 1500Mpix low end of 2005
So it seems 1GHz, 1Gtex, 1Gpix and 100milion polygons per second to be optimum for ClassiCube to run 60FPS On Pentium M maximum framerate was typically 50-55FPS, few drops to 19-22FPS depending on what chunk of world was visible. Averange 30FPS is achievable no problems. GeForce 4 Go 200MHz, 32MB machine is capable running for example IL2 Sturmovik, Neverwinter Nights type of games. In this case cube based game, number of polygons drawn is major limiting factor both for CPU and GPU. Not lighting or textures applied.
On Core 2 Duo, Intel 965 chipset GPU - game does not use 100% of CPU - CPU is 2x faster than Pentium M used only 35-50% nothing change whatever lighting i choose in settings. So GPU parameter of "milions polygons per second" was limitation. In this case 125Mpoly - im not sure what is clock of mine could be much lower on that laptop. GPU overal is not much better than older GeForce 4 Go FPS. Framerate 55-59FPS never drop below 33FPS but still dropped depending what chunk world was visible. As comparison OpenSpades (also voxel based game, no textures, just colors but much better lighting) Spades runs 9FPS on that Core 2 Duo, wont run on Pentium because of too old OpenGL driver to run. Limitation of OpenSpades protocol is world build of 64 x 512 x 512 cubes. Im impressed. I don't know what magic juice ClassiCube developer drink but it's well designed well balanced design.
EDIT EDIT EDIT: Mpoly per second matters not mtexures applied nor,mpixels rasterized but how many polygons can be created and rasterized. Obvious for game made of cubes. Each block 6x2 triangles = 12 triangles. In world of 100 high x 1000 wide x 1000 long you will have 12x 100k = 1.2milion polygons each frame. Perfectly in line to what i concluded based on specification of GPUs and CPU limitation. If game renders around 1mln polygons per scene to achieve 60FPS is demand on CPU or GPU side to render 60milion polygon per second. Voodoo 2 1998 180Mtexel to texturize polgons or something and for example NVidia TNT2 250Mtexel https://www.techpowerup.com/gpu-specs/riva-tnt2-pro.c1303 doesn not have T&L all is dependent on CPU. I don't know why i reed Voodoo is 15milion polygons per second whateve is. Guess such hardware manufactured prior to 2000 will run game at 10-15FPS, probably would be capable fill textures (mtexels / second), would be capable run at higher resolution (mpixels / second) but hardware slower than 1GHz CPU wont be able draw as much polygons. And early T&L capable GeForces wont be able create as many polygons.
Can you optimize game even further? Probably not. ClassiCube rules!
