PDA

View Full Version : ATI does not support 1.4?



bakery2k
02-06-2003, 10:14 AM
After owning 3 or 4 nvidia based cards, I purchased my first ATI card today, a radeon 9000 64MB.
I downloaded and installed the latest drivers from www.ati.com (http://www.ati.com) and was dissapointed to not find OpenGL 1.4 support. I also looked at the hardware registry at www.delphi3d.net/hardware (http://www.delphi3d.net/hardware) and could not see any radeon cards using GL1.4. BTW, I am using Win98. When will this feature be available (if it is not already)?

I also had a problem with ARB_vertex_program. My cel-shading demo (available from my site) ran incorrectly and the frames per second dropped from 120 to 15 when enabling ARB_vertex_program.
Also, my shadow volumes demo does not work correctly, although there is no speed drop.

Korval
02-06-2003, 10:21 AM
Why would you expect a Radeon 9000 (a cheap 8500) to support OpenGL 1.4? It can handle ARB_vertex_program, but not ARB_fragment_program.


My cel-shading demo (available from my site) ran incorrectly and the frames per second dropped from 120 to 15 when enabling ARB_vertex_program.

A clear sign of software use. I don't know why; the card has the vertex program hardware for it.

bakery2k
02-06-2003, 10:24 AM
OpenGL1.4 does not require the support for ARB_fragment_program. I had 1.4 support on my GF3.
Anyway, on delphi3d even the 9700 cards do not have version 1.4.

Ostsol
02-06-2003, 10:40 AM
What are you refering to when you say "OpenGL 1.4" support? Are you refering to the functionality that has been upgraded from merely being extensions to being an integral part of the GL spec?

bakery2k
02-06-2003, 10:41 AM
Yes.
My drivers, and all those which I can find, report "1.3.xxxx" in the GL_VERSION string.
It's not a huge problem, as I can use extensions, but it was a bit surprising.

OneSadCookie
02-06-2003, 11:16 AM
The Radeon 8500/9000 series don't support ARB_depth_texture or ARB_shadow, and therefore can't support OpenGL 1.4. The Radeon 9700 should be capable of it.

The GeForce 3/4Ti apparently can't support all of 1.4 either, but the NVidia drivers report 1.4. Unfortunately, I haven't been able to find out what functionality of 1.4 the 3/4Ti don't accelerate http://www.opengl.org/discussion_boards/ubb/frown.gif

Elixer
02-06-2003, 11:58 AM
I find it very strange you would get a 9000 if you had a GF3, that is a downgrade.

Perhaps next time, you should do a bit more research, and then find out that the 9000 is the same thing as a 8500?

I am still amazed that people that get GF4 MX cards think they they have a real GF4 card, and not a slightly updated GF2 MX card.

vincoof
02-06-2003, 12:39 PM
>> Unfortunately, I haven't been able to
>> find out what functionality of 1.4
>> the 3/4Ti don't accelerate

EXT_blend_func_separate


>> Perhaps next time, you should do a bit
>> more research, and then find out that the
>> 9000 is the same thing as a 8500?

Actually, 8500 is (meant to be) superior to a GF3. So, if 8500 == 9000, and if 8500 > GF3, then 9000 > GF3. Logical, isn't it ?


>> I am still amazed that people that get
>> GF4 MX cards think they they have a real
>> GF4 card, and not a slightly updated GF2
>> MX card.

GF4MX is really better than GF2MX, by far. There's a huge gain. Though, if you mean "GF4MX is lower than GF3" then I agree.

[This message has been edited by vincoof (edited 02-06-2003).]

Ostsol
02-06-2003, 12:49 PM
Originally posted by vincoof:
Actually, 8500 is (meant to be) superior to a GF3. So, if 8500 == 9000, and if 8500 > GF3, then 9000 > GF3. Logical, isn't it ?

The 8500 is superiour in speed to the 9000 and the Geforce 3 and it does have advantages in terms of some features. However, there are some OpenGL 1.4 features that Geforce 3 supports while the 8500 does not, such as ARB_shadow.

Elixer
02-06-2003, 01:02 PM
I meant to say in features in all cases, as in the GF4 MX doesn't really have anything new feature wise that the GF2 MX doesn't have, it may be a bit faster, but it doesn't even have the features that the GF3 has.

Looks like Ostsol wrote what I meant to say. http://www.opengl.org/discussion_boards/ubb/smile.gif

Damn marketing strikes again. I wonder if ATI will call their next card the 10K? http://www.opengl.org/discussion_boards/ubb/wink.gif

Ostsol
02-06-2003, 06:30 PM
Heheh. . . I'm wondering if NVidia will remake the Geforce 3 or 4 and call it the GeforceFX MX. http://www.opengl.org/discussion_boards/ubb/tongue.gif Hmm. . . or maybe they realize how stupid a name that sounds. . .

As for ATI. . . I really hope that don't decide to name their next chip the 10k. Best to simply come up with a new name. . .

bakery2k
02-06-2003, 11:55 PM
Originally posted by Elixer:
I find it very strange you would get a 9000 if you had a GF3, that is a downgrade.

Perhaps next time, you should do a bit more research, and then find out that the 9000 is the same thing as a 8500?

I am aware that the 9000 is basically an 8500. That's why I bought it. I now have access to ATI_fragment_shader etc on one card and NV features on another.
My current project can thus run well on both vendor's hardware.

Tom Nuydens
02-07-2003, 12:35 AM
Originally posted by OneSadCookie:
The Radeon 8500/9000 series don't support ARB_depth_texture or ARB_shadow, and therefore can't support OpenGL 1.4. The Radeon 9700 should be capable of it.

Not being able to accelerate all of it is no excuse for not supporting it. If it were, we'd all still be using OpenGL 1.0. Accumulation buffer, aux buffers, color index mode and more are not accelerated on the majority of current cards.

-- Tom

Ventura
02-07-2003, 01:19 AM
I too have both a GeForce3 and a Radeon 8500 and to be honest, i prefer the Radeon for programming.
yeah its raw fill-rate isnt as high, but 6 texture units, and proper dependent lookups are _far_ more usefull to me.
One thing that is bugging my at the minute is that ext_vertex_shader on the radeon doesnt seem to have position invariance like the nv_vertex_program does??
does anyone know if there is any way of doing this on the Radeon? i presume there is support for it in the arb_vertex_program which the card should support too.

M.

ehart
02-07-2003, 03:31 AM
bakery2K:

The 9000 cannot support shadow mapping as was mentioned in the thread. On another note the 9700 doesn't presently export a 1.4 string yet either. This is largely just because of a desire to do more testing.

Ventura:

EXT_vertex_shader on ATI parts should always be position invariant if you multiply the vertex by the builtin modelviewprojection matrix, unless you are exceeding HW resources and falling back to SW. If you are transforming it differetly, then it won't be invariant.

-Evan

Korval
02-07-2003, 09:01 AM
Not being able to accelerate all of it is no excuse for not supporting it. If it were, we'd all still be using OpenGL 1.0. Accumulation buffer, aux buffers, color index mode and more are not accelerated on the majority of current cards.

ATi could waste a lot of time writing software support for these few extensions. Or, they could not advertise full 1.4 and spend more time writing better drivers with fewer bugs.

I much prefer to have bug-free hardware accelerated card features than to have an implementation expose extensions that are just getting in the way of killing bugs.

Besides, the 9000 will be gone in 6 months or so, when ATi creates a low-end (<$100) DX9 capapble card. Why should ATi spend driver developer time on the 9000 when those people could, in theory, help with the unified R300/R350 drivers?

Theodor
02-07-2003, 11:28 AM
does anyone know if there is any way of doing this on the Radeon? i presume there is support for it in the arb_vertex_program which the card should support too.

Latest ATI drivers does provide support for ARB_vertex_program for Radeon 8500.
And it does work with position invariance option.

However I got a problem with that extension on Radeon 8500 - my demo seems to just hang up on VertexAttribPointer (while just this same code works for GF2).
Anybody here managed to use glVertexAttribPointer successfully on Radeon 8500?

Humus
02-07-2003, 12:02 PM
Yes, glVertexAttribPointer worked fine for me on my old 8500 last time I used it.

Ventura
02-07-2003, 04:31 PM
Thanks Evan,
I have found out with a quick test that Radeon cards are invariant with the basic transorm. I guess the card runs the fixed T&L path with vertex programs anyway to guarantee the invariance.
Its a bit out of date now, but it might have been woth while mentioning in the spec.
M.