Now with the incredible news of this project:

http://hdrlab.org.nz/radeonhd-driver/

Using PCI Radeon X1000 series gfx cards for Os4 is possible I wonder if we could see some co-operation and porting for Morphos? That would be killer..

I wouldn't call 2D only driver "incredible news". Wake me up when they have working 3D that is at least as fast as the MorphOS 3D with Radeon 9000 or 8500/9100.

That being said, co-operation is not going to happen anyway.

Bigfoot and Frank Mariak can write great Radeon gfx drivers and I bet MorphOS Radeon driver will continue to be faster than OS4. BTW, PCI bus is not a good idea for 3D graphics.

[ Edited by Crumb on 2010/1/19 17:14 ]

Frankly, I'd just be happy if we were up to the 9500/9700 or 9600/9800. Most X1xx cards are PCIe (and thus, there is no MorphOS platform for them).
Besides, what's the advantage of using higher end video cards if the driver don't even use the abundance of shader hardware present on them?
Personally, I'd love a driver for the HD38 or HD48 series. I just bought an HD3850 with 512MB of memory on Ebay for $40.00 for my PC.
While the Mac 9000Pro I also recently bought cost a lot less, it's also a lot less powerful.

> Most X1xx cards are PCIe

I guess you lost one 'x' ;-)

> thus, there is no MorphOS platform for them

There's no problem in specifically purchasing a PCI X1xxx card. That most of these cards are PCIe is rather irrelevant. You'd just have to have your eyes opened while buying :-P

> what's the advantage of using higher end video cards if the driver don't even use
> the abundance of shader hardware present on them?

Availability, for instance. In that respect it's about recentness. That more recent cards are usually higher end is more of a by-product.

Yes, I missed one x (although I could have left off the 1 as well since I didn't bother addressing the x600-800).

And there is an AGP version of the 3850 (and the 4650 for that matter), so if we had the drivers we could use them (or the AGP X700, X800, or X1600).

And you're right, the fact that I could buy an HD3850 for less than an X1650 is just a current market phenomenon (lucky for me).

But I'd still like to see a driver for the Radeon 9500-9800 (and maybe the 9600 - although that's an unrelated GPU). I have a Radeon 9500 Pro on hand that can be modified to work in a Powermac (as could all the cards based on the GPUs I've just mentioned).

Since these cards are the next natural step after the 9000-9200, and they work in the platforms currently (or soon to be) supported, they make sense as the next to be supported (even if Radeon 9800s currently resell for more than an HD3850).

[ Edited by Jim on 2010/1/19 22:56 ]

[ Edited by Jim on 2010/1/20 3:49 ]

> there is an AGP version of the 3850 (and the 4650 for that matter), so if we had
> the drivers we could use them (or the AGP X700, X800, or X1600).

Right, for Pegasos AGP is to be prefered over PCI.

> the 9600 - although that's a unrelated GPU

In what way unrelated?

Unlike the GPU in the 9500-9800, the design of the 9600 uses fewer pipelines operated at a higher speed. Under Windows, the same driver will work for both the 9500 Pro and the 9700. Except for a few minor changes, the 9800 is an enhanced 9700.
Further, there are FireGL cards that are more closely related to the 9500-9700.
The 9600, while in name appearing to be related to these others, is no more closely related than is a 9000 or 9200.
While it would be nice to see a driver for the 9600, these other cards could all be implemented by a single driver (with minor tweaks for the differences in clock speed) or by a group of closely related drivers.
A driver for the 9600 would be distinctly different.
[ Edited by Jim on 2010/1/19 23:47 ]

[ Edited by Jim on 2010/1/19 23:51 ]

[ Edited by Jim on 2010/1/19 23:52 ]

> Unlike the GPU in the 9500-9800, the design of the 9600 uses fewer pipelines operated
> at a higher speed. Under Windows, the same driver will work for both the 9500 Pro and the
> 9700. Except for a few minor changes, the 9800 is an enhanced 9700. [...] The 9600,
> while in name appearing to be related to these others, is no more closely related than is a
> 9000 or 9200. While it would be nice to see a driver for the 9600, these other cards could
> all be implemented by a single driver (with minor tweaks for the differences in clock speed)
> or by a group of closely related drivers. A driver for the 9600 would be distinctly different.

Thanks. I didn't know that in such detail. I knew that 9500 and 9700 are almost the same, but not that the 9800 is closer related to 9700 than to 9600. So I'm completely with you to say that full support for 9500/9700/9800 should have priority over support for 9600 (or 9550, which is used in the last iBook G4 and is despite its name almost the same as the 9600).

For use in a Pegasos you should also be aware of the required voltage for newer AGP cards.

Good Point. What type of AGP port does the Pegasos have (and what voltage does it supply the video card)?
I'm already familiar with the leads that have to be taped over to use an 8X AGP card in a Powermac, but I'm not too familiar with the Pegasos or the Efika (which seems to have only a partial implementation of the AGP standard).
If we want to run more advanced video cards under MorphOS, there are a lot of considerations (other than the need for drivers) that will need to be addressed.

Piru
Thanks for the post. What about support for Radeon 9600+ for PEgasos? Since the Powerbook version of Morphos must have these drivers..

> What type of AGP port does the Pegasos have (and what voltage does it supply the
> video card)?

AGP 2.0 (1.5V).

> the Efika (which seems to have only a partial implementation of the AGP standard)

Efika 5200B itself has no AGP at all.

1.5V cool, but what speed 4X or 2X (it doesn't seem likely that it's 8X).

3.3V @ 1X on the Pegasos 2 according to the manual.

> 3.3V @ 1X on the Pegasos 2 according to the manual.

That I consider strange. The manual states AGP 2.0 and 3.3 V. I think to know that AGP 2.0 is 1.5 V and 3.3 V is AGP 1.0.

http://en.wikipedia.org/wiki/Accelerated_Graphics_Port#Versions supports my tune.

Can someone please clarify? Neko?

> it doesn't seem likely that it's 8X

Yes, since 8x would require AGP 3.0.

Andreas
You guys are aware that Peg is not REAL AGP right? Its more like a PCI AGP bridge hack... iirc

> You guys are aware that Peg is not REAL AGP right?

You mean it's neither AGP 2.0 nor AGP 1.0, and the manual is lying in claiming it's AGP 2.0?

> Its more like a PCI AGP bridge hack... iirc

If spec A is bridged to spec B, spec B is still spec B.

Quote:
Why is that ?