Okay, this fascinated me, so I actually registered here (and bought PB) just to try it and comment.
I have two machines, with three video hardware choices. One machine has nVidia and Intel graphics, one has AMD. Using WHQL drivers for all three.
Using a near variant of this example, I get:
nVidia: ~570 circles per frame for about 4 seconds, ~400 for about 4 seconds, then 240 or so for at least 20 seconds.
ATI: ~640 circles per frame. No dropoff in at least 30 seconds.
Intel HD: ~2100 circles per frame. No dropoff.
If I change to OpenGL, I get:
nVidia: ~7000 circles per frame. No dropoff.
ATI: ~4k circles per frame. No dropoff.
Intel HD: ~6k circles per frame. No dropoff.
Specific hardware:
nVidia: GeForce 670 GTX, factory-overclocked board.
ATI: Radeon HD 6950, 2GB. Might be factory-overclocked but not by much.
Intel: Intel HD Graphics 4000. Not overclocked that I know of.
The Intel system is a recentish Core i7, the AMD system is a Bulldozer 6100 or so (6 core, 3.3GHz).
Note: I play video games on both of these systems (though not using the Intel HD graphics, that's for a second monitor running shells when I'm programming). On neither do I see results that make me think that I'm suffering from a factor of five performance penalty for running games under DirectX.
Strange deviations in effective framerate
Re: Strange deviations in effective framerate
DirectX9 commands are using 3D, even for 2D stuff. OpenGL, surely not. You could also try DirectX7 subsystem.
-
- Always Here
- Posts: 6426
- Joined: Fri Oct 23, 2009 2:33 am
- Location: Wales, UK
- Contact:
Re: Strange deviations in effective framerate
Hmm time is always 0.0? Number of Circles is consistently above 9000.
WinXP SP3
CPU: Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz
DirectX version: 9.0c
GPU: nVidia GeForce 7900 GT GDDR3 256MB
WinXP SP3
CPU: Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz
DirectX version: 9.0c
GPU: nVidia GeForce 7900 GT GDDR3 256MB
IdeasVacuum
If it sounds simple, you have not grasped the complexity.
If it sounds simple, you have not grasped the complexity.