Originally posted by: skace
Originally posted by: Zenoth
In MII: TW all you need to do is to populate a skirmish game with 80,000 troops on the field at maximum settings with AA and AF at high resolutions, and I guarantee you that Crysis in comparison runs very, very, very smoothly.
What are you testing there though? Sounds like a CPU test -- like most RTS max unit tests.
Believe me, you'll also test the GPU, if not that only. You don't even need to start the fight, right after the map loads, before the units placement phase ends, the A.I isn't active at all, just move the camera around with that many soldiers, or more (it ain't limited to 80,000, it can be higher). If anyone even with a Tri-SLi system manages to get 60FPS at say 1600x1200+ with a minimum of 4xAA and 16xAF then I will be very surprised. When the fight actually starts then the A.I kicks in and the frames drop is ten fold higher.
My own system can barely sustain 2 Vs 2 matches with only troops (no catapults, no cavalry, only men) with around 30,000 units, and that's without grass, no shadows, no AA, no AF, at 1280x1024, everything else at maximum. The skirmishes can have a 4 Vs 4 confrontation setup (and with mods you can go beyond the 8 total players limit), and the 80,000 units count or so can be reached with a 3 Vs 3 match. Technically speaking I don't know of any other games capable of testing a GPU like this one under these specific circumstances, but it's barely known or noticed since people usually play Campaigns, and even with Huge army sizes it is a rare sight to see more than 3,000 men battles (total, including enemy's numbers).
And to test properly the camera needs to be placed close enough to the ground, because the further away it gets from the action and the less detailed the troops become, to the point where they can become sprites, in which case the frames rate usually goes up noticeably (although it's still slow).