Is anyone able to submit a proper bug report on this? I'd be happy
to try, but I think my report would be lacking in details "QC on
NVIDIA chips is bad, mkay...uh...pls fix it"?
I'd hope that a proper bug report, one that suggest the right fix(es)
could get some attention?
On Feb 9, 2009, at 3:15 PM, vade wrote:
A little birdy mentioned this has to do with the image channel
ordering, and that the ATI drivers automatically swap the channels
to the necessary ordering to make things faster, using a shader on
the GPU. Apparently, something is sub optimal with ATI for this.
Maybe its specific to QCs opaque handing of images (QCImage
datatype), but ive head that this holds true for other apps.
your tests seem to confirm this, and my testing on a colleagues new
MBP is that while its faster than the older NV 8600M hardware, my
x1600 still stomps it for CI/Video pipeline processing.
The Raw GL calls though (vertex drawing/filling etc), are much much
faster than my x1600, so its a give and take.
On Feb 9, 2009, at 4:36 PM, Michael Diehr wrote:
Your benchmarks corelate with vram also. Have you tried 8600M with
the same vram as the ati cards?
We are just testing the ones we have, and don't have that
particular combo.
The tests we are doing should not be taxing VRAM, i.e. we are using
720x480 internal pictures which works out to under 2MB per image,
and only a dozen or so images max. So I don't think we are
running out of VRAM.
I just ran another test comparing the ATI 1...@256mb vs. the NVIDIA
8600 @ 128MB, this time doing a very simple test (3 video inputs
overlayed). Results: ATI = 30fps, NVIDIA = 18fps. Again, 3 x
2MB per frame images should not be getting anywhere near to 128MB
VRAM total.
Now, perhaps the 256MB version of the 8600 has not only larger VRAM
but faster VRAM as well, in which case it'd perform better?
Still, my gut feel is that the NVIDIA chips or drivers are not well
suited for QC based on what I've seen.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartzcomposer-dev mailing list ([email protected]
)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/quartzcomposer-dev/doktorp%40mac.com
This email sent to [email protected]
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Quartzcomposer-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com
This email sent to [email protected]
