Anand Hops on the R420 FP32 Bandwagon

G

Geo

Mostly Harmless
Legend
Just as it appears the 5200 finessed the min spec, why couldn't R420?

... But does the 5200 still finesse min spec by reducing FP32 to FP16 in games? B/c several ppl have mentioned that it's capable of using FP32 in "synthetic" tests.
 
I think the lower to middle range nvidia cards are a perfectly good example of why you won't see a minimum spec of fp32 until longhorn.
 
Just out of curiosity. Does Microsoft have information about the minimum requirements for DirectX 9 hardware publically available? I have never managed to find any.
 
Minimum precision is 24 bits floating point, unless _pp (partial precision) is specified (in which case it can be 16 bit floating point).
And there is no (and won't be) DirectX 9.1 period!
 
MDolenc said:
Minimum precision is 24 bits floating point, unless _pp (partial precision) is specified (in which case it can be 16 bit floating point).
And there is no (and won't be) DirectX 9.1 period!
Click to expand...

Based on the way he stated it, I'm starting to wonder if some of these folks are using "DX 9.1" and PS/VS 3.0 synonymously.
 
geo said:
From http://www.anandtech.com/weblog/index.html

"I've read the latest rumors on FP32 support in R420, but I don't see how that's any major news. As far as I know, the DirectX 9.1 spec (with pixel shader 3.0 support) calls for full precision to be 32-bit floating point, so if ATI wants to be DX9.1 compliant they will have fp32 support."

Uh oh, there goes Digi's vein again. . .

Props here for pointing it out: http://www.rage3d.com/board/showthread.php?s=&threadid=33736479
Click to expand...
AIIIIIEEEEEEE!!!! MY BRAIN, IT SCREAMS ON THE INSIDE!!!!!
 
MDolenc said:
...which is wrong becouse PS/VS 3.0 is already in DirectX 9.0 which has been out for over a year now.
Click to expand...

There is still some mileage to be had with marketting an "update" to DX9 (even if it may not necessarily be much of an update).
 
MDolenc said:
...which is wrong becouse PS/VS 3.0 is already in DirectX 9.0 which has been out for over a year now.
Click to expand...

I hear you. I'm just trying to figure out where the root of this repeating rumor (DX9.1) is coming from. I'm wondering if some are using it as a place-holder for PS/VS 3.0 compatibility because they don't understand that both PS/VS 2.0 and PS/VS 3.0 are in the DX9 spec. So, to them, maybe cards supporting PS/VS 3.0 are what they are pointing at when they say "DX9.1"? Then again, maybe I'm trying too hard to give them something sort-of-logical looking (even tho wrong, as you point out) to explain it with.
 
To me, it makes some sense for Microsoft to update the API for new hardware functionality that is certain to not be exposed through PS/VS 3.0. For example, there may be other useful instructions that Microsoft hasn't thought of that become hardware accelerated, or there may be support for more robust HOS, etc.

To be frank, I would be rather disappointed if the only major improvement (besides, hopefully, AA, Aniso, and other normal rendering improvements....) was a move to PS/VS 3.0.

Edit: Gah, that was a major typo...
 
Chalnoth said:
To me, it makes some sense for nVidia to update the API for new hardware functionality that is certain to not be exposed through PS/VS 3.0.
Click to expand...

When did NVIDIA start updating DX?
 
Dave: But wouldn't it be better (from PR perspective) to claim "unlimited instructions in pixel and vertex shaders" (since you have branching), etc. Why market something "designed for DirectX 9.1", when DX 9.1 does not and very VERY probably wont ever exist?

geo: That's what I'd like to know too. Decision not to have another runtime update (i.e. DX 9.1) was made about a year and a half ago, when shader model 3.0 came into DX 9. The only thing that is going to change is the D3DX part of SDK, when new HLSL targets will be added. This doesn't change end user situation even a bit. They can still run games with DX 9.0 (or 9.0b if you worry about that MIDI playback overflow) runtime that was released in December 2002! And this SDK update won't be called "DX 9.1", but something like "DirectX 9.0 SDK Update (Spring 2004)".
 
MDolenc said:
Dave: But wouldn't it be better (from PR perspective) to claim "unlimited instructions in pixel and vertex shaders" (since you have branching), etc. Why market something "designed for DirectX 9.1", when DX 9.1 does not and very VERY probably wont ever exist?
Click to expand...

No, I'm talking about marketting from MS's point of view. Everyone knows Shader 2.0 is part of DX9, the knowledge on 3.0 is much more limited - you may as well get some marketting out of saying "look, we don't sit idle, we make new features available with a new DX update" and you get the potential for driving a lot of hits to the MS website for the download.

However, won't the addition of the flag for support of NV3x's float support require a runtime update? As well as filtering and blending of float16 support?
 
Chalnoth said:
To me, it makes some sense for nVidia to update the API for new hardware functionality that is certain to not be exposed through PS/VS 3.0.
Click to expand...

Does anyone else see anything in this thread that pertains to nVidia..... :?:
 
MDolenc said:
Dave: But wouldn't it be better (from PR perspective) to claim "unlimited instructions in pixel and vertex shaders" (since you have branching), etc. Why market something "designed for DirectX 9.1", when DX 9.1 does not and very VERY probably wont ever exist?

geo: That's what I'd like to know too. Decision not to have another runtime update (i.e. DX 9.1) was made about a year and a half ago, when shader model 3.0 came into DX 9. The only thing that is going to change is the D3DX part of SDK, when new HLSL targets will be added. This doesn't change end user situation even a bit. They can still run games with DX 9.0 (or 9.0b if you worry about that MIDI playback overflow) runtime that was released in December 2002! And this SDK update won't be called "DX 9.1", but something like "DirectX 9.0 SDK Update (Spring 2004)".
Click to expand...

Would one of these new HLSL targets be for the R350 to expose the f-buffer or is a change like that not possible without a more major update?
 
You must log in or register to reply here.
Back
Top