You should be doing much more complete analysis. For example trying to use some profiling tools to measure the actual GPU load.
Also MP itself should be limiting to v-sync, so the max fps will be the 60Hz in your case.
#header.image can cause high CPU (1 Viewer)
- Thread starter Swede
- Start date
You should be doing much more complete analysis. For example trying to use some profiling tools to measure the actual GPU load.
Also MP itself should be limiting to v-sync, so the max fps will be the 60Hz in your case.
Germany
I've attached a core.dll containing my fixes. It's built to be used with the last public SVN build. By default it won't pack the skin gfx anymore. Please share your results.
"Rivatuner" is an easy to use tool which works like the taskmanager to show how busy the GPU is.
- Thread starter
- #73
rtv
Tested your core with Rivatuner. Graph attached. Nothing new that we didn't allready know from MP log, what should I test that could confirm that packed textures are doing any good?
Test was: Startup, goto Moving Pictures, browse 15 movies, close
Only downside in both my fix and rtv's core is that the volumebar does not work.
Tested your core with Rivatuner. Graph attached. Nothing new that we didn't allready know from MP log, what should I test that could confirm that packed textures are doing any good?
Test was: Startup, goto Moving Pictures, browse 15 movies, close
Only downside in both my fix and rtv's core is that the volumebar does not work.
Attachments
Too bad that GPU usage is not shown, only CPU. So the test results aren't showing the benefits from the much less texture / rendering state swapping that GPU is doing.
- Thread starter
- #75
Well, there doesn't seem to be a way to measure that. I'm satisfied now that MP can be used in low spec PC's
Removing packed textures is only a hack that disables optimized GPU usage (please refer to the URLs I posted). The correct way to solve the issue on 128MB GPUs has been also posted (mainly targeted to skinners).
Many users have reported stuttering video playback while in EPG, most likely those users would be having much bigger performance issues in those situations without the packed textures.
We shouldn't remove the packed textures usage, unless someone will be willing to do some real profiling on GPU usage.
- Thread starter
- #77
The perfect solution is what rtv does: an option to use cache or not. Then everyone can test for them selves.
Perfect solution would be something that users don't have to fiddle with.
- January 7, 2008
- 167
- 58
How about using multiple texture atlases based on which MP screen we're on? We could still pack-and-cache all needed textures and do a (partial?) atlas change on transition from e.g. Home Screen to My Movies.
Commonly used textures (for a given definition of "commonly used") could be packed into the same packfile to ensure they're always available (or alternatively accept textures get put into more than one texture pack).
I guess this could add som lag to screen switching, but most of the time is spent inside a given screen, not on switching between screens, so maybe that's acceptable?
Commonly used textures (for a given definition of "commonly used") could be packed into the same packfile to ensure they're always available (or alternatively accept textures get put into more than one texture pack).
I guess this could add som lag to screen switching, but most of the time is spent inside a given screen, not on switching between screens, so maybe that's acceptable?