Ok, Andy did some tests and here are his numbers:

GeForce 6200 AGP 256MB:
quake3: 1346 frames, 7.0 seconds, 192.8 fps
glxgears: 1360.8 fps

Sapphire Atlantis 7000 PCI 64MB DDR:
quake3: 1346 frames, 43.4 seconds, 31 fps
glxgears: 343.584 fps

Radeon 9000 PCI 64MB:
quake3: 1346 frames, 85.1 seconds, 15.8 fps
glxgears: 413.6 fps

The GeForce and 7000 numbers look usable for us, but there's something
wrong with the 9000.  Andy found all sorts of misrendering with the
9000.  If anyone would like to take a look at the glxinfo for the
9000, perhaps we can fix that.

We want to make a projection for OGD1.  To do that, we need to compare
the various limiting factors.  They are:

- Memory bandwidth
- PCI bandwidth
- Pixel pipes
- GPU clock

There are numerous other factors, but I think these are probably the
dominant ones.

To begin with, there's PCI.  That's going to be a major choke point,
and we have to determine its effects.  That means we either need to
try another PCI card (with faster GPU) or a Radeon 7000 AGP card.  Or
we can try another GPU entirely, one on PCI, the other on AGP/PCIe.

Here's how we stack up to the 7000 (us vs. them):

Pixel pipes:  2 vs 1
Usable memory throughput:  1350 vs 430
GPU clock:  150-200 vs 150

Depending on whether it's GPU choked or memory choked, we can move
anywhere between 2 to 3.5 times as many pixels per second.  If that's
all there is to it, we can say that OGA (on OGD1) would push at least
60fps for Quake III.

However, if the PCI bus is the problem, we'll only beat the 7000 if
our DMA packets require fewer words for the same number of triangles.

I don't know how small the glxgears triangles are, but if they're not
TOO small, we could double the 7000's performance, so we'd push about
680 fps.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to