More ATI Missing Vertex Texture Fetch Info

Status
Not open for further replies.
P

PRIME1

2[H]4U
Joined
Feb 4, 2004
Messages
3,942
http://www.ati.com/developer/gdc/D3DTutorial01_3_0_Shaders.pdf

It seems back in March ATI did a presentation at the Game Developers Conference. While talking up the virtues of Vertex Shaders there is no mention of them not including the Vertex Texture Fetch, nor are they instructing anyone about a "work around"

Maybe the R580 or R600 will have this feature. The R520 may have had problems with it so they dropped it.
 
ATi's next architecture should take a unified approach so I don't think you'll ever see a vertex texture fetch in any of their cards - but that's not a bad thing. I don't even think nVidia will have it if G80 is unified. I don't even know why people like you are harping on this issue when it's not a big deal. Besides PF, no game on the market utilizes displacement mapping so why make a big deal about it?
 
Again, if we are moving towards uni arch, there wont even BE these specialised units in many cases. It will be handled in software... Just like ATis workaround for their current lacking units.
 
eno-on said:
Again, if we are moving towards uni arch, there wont even BE these specialised units in many cases. It will be handled in software... Just like ATis workaround for their current lacking units.
Click to expand...

Yep. :)
 
For the next few years SM3.0 will be the standard. A unified architecute is the future but not the present. It will be a while before you see hardware, let alone software that supports it.
 
PRIME1 said:
For the next few years SM3.0 will be the standard. A unified architecute is the future but not the present. It will be a while before you see hardware, let alone software that supports it.
Click to expand...
Im kinda guessing a real-time (or otherwise) compiler will be used in future drivers for unified architecture. I'm thinking ATi could write into their drivers a way for this texture fetch thingy thats catching so much flack, to work properly on their hardware. Meaning it would be executed BY hardware. Thus meaning it does the 3.0 spec, in hardware.
So seemingly, if a completely unified arch was capable of doing all the sm3.0 stuff in hardware, with a compiler, it would be in spec for sm3.0. Not to mention, I thought ATis stuff after the r580 was going to be completely uni arch anyway? Meaning, it wouldn't meet any spec unless drivers made it do what it was supposed to on hardware.
 
PRIME1 said:
For the next few years SM3.0 will be the standard. A unified architecute is the future but not the present. It will be a while before you see hardware, let alone software that supports it.
Click to expand...

So what? Any developer that wants to use vertex texture fetches will just add in R2VB through fourcc. Again I ask you, what's the big deal?
 
5150Joker said:
So what? Any developer that wants to use vertex texture fetches will just add in R2VB through fourcc. Again I ask you, what's the big deal?
Click to expand...
No big deal, I found the ATI article interesting. Obviously without a number of SM3.0 games out there it's tough to measure the impact of not including a texture fetch. However if you read the article linked it sounds like it could be very important.

But I see no reason why it should not be discussed (as this is a forum for discussing such things).
 
PRIME1 said:
No big deal, I found the ATI article interesting. Obviously without a number of SM3.0 games out there it's tough to measure the impact of not including a texture fetch. However if you read the article linked it sounds like it could be very important.

But I see no reason why it should not be discussed (as this is a forum for discussing such things).
Click to expand...


In your original post you didn't mention anything about discussing the impacts of a vertex texture fetch, you made an accusation that somehow ATi was being deceitful - at least that's what I got from it. BTW ATi didn't have problems with adding it in R520, they simply felt it wasn't used enough to warrant spending the die space on it and they were right.
 
5150Joker said:
In your original post you didn't mention anything about discussing the impacts of a vertex texture fetch, you made an accusation that some how ATi was being deceitful - at least that's what I got from it. BTW R520 didn't have problems with adding it in, they simply felt it wasn't used enough to warrant spending the die space on it and they were right.
Click to expand...
The R520 did have problems with manufacturing and had a few re-writes, however I have read no details as to what changes were made during those re-writes. Back in March they clearly had a better view of the SM3.0 implementation of vertex shaders. I won't overtly accuse ATI of anything since I have no corroborating evidence to make any such claims.

However, while it's just as easy for you to say ATI is right in not including it, it's equally easy to say that NVIDIA has been doing this a while themselves and found that it was needed.
 
PRIME1 said:
No big deal, I found the ATI article interesting. Obviously without a number of SM3.0 games out there it's tough to measure the impact of not including a texture fetch. However if you read the article linked it sounds like it could be very important.

But I see no reason why it should not be discussed (as this is a forum for discussing such things).
Click to expand...

Except to be discussed in a useful way, it needs to be discussed by people with the technical knowledge to fully understand the issues present in architecture. As its clear you don't understand them, Kyle, Brent, and almost everyone else on this forum doesn't understand them maybe you should head on over to Beyond3D and ask these questions if you're so "interested"
 
OldBoy said:
Except to be discussed in a useful way, it needs to be discussed by people with the technical knowledge to fully understand the issues present in architecture. As its clear you don't understand them, Kyle, Brent, and almost everyone else on this forum doesn't understand them maybe you should head on over to Beyond3D and ask these questions if you're so "interested"
Click to expand...

That's where I've been going and reading up on the issue. Lots of good information there.
 
PRIME1 said:
The R520 did have problems with manufacturing and had a few re-writes, however I have read no details as to what changes were made during those re-writes.
Click to expand...

I thought everyone already knew why? It was because of some soft ground issue they were having during manufacturing.
 
OldBoy said:
Except to be discussed in a useful way, it needs to be discussed by people with the technical knowledge to fully understand the issues present in architecture. As its clear you don't understand them, Kyle, Brent, and almost everyone else on this forum doesn't understand them maybe you should head on over to Beyond3D and ask these questions if you're so "interested"
Click to expand...
I am glad you know so much about my programming ability and my knowledge of DirectX implementation. It's good to see that only a handful of people at beyond3d are the experts on this and no one else could possibly have the literacy level to read up on it and do research for themselves. I am not a member of that forum, I am a member of this forum and I chose to discuss it here. If you don’t have anything valuable to add, you can always PM me your complaints.
 
PRIME1 said:
If you don’t have anything valuable to add...
Click to expand...

Take your own advice. You're not technically competent to even understand the issue and you asking a bunch of other people who aren't either adds nothing of value to anything around here.

If you were in fact as "interested" as you "claim" to be, you would take the five seconds to head on over to Beyond3D and either read the informative posts about it in the forums by people who do have the technical knowledge to properly discuss the issue or sign up and post your question in the forums to satiate your "interest".
 
PRIME1 said:
I am glad you know so much about my programming ability and my knowledge of DirectX implementation. It's good to see that only a handful of people at beyond3d are the experts on this and no one else could possibly have the literacy level to read up on it and do research for themselves. I am not a member of that forum, I am a member of this forum and I chose to discuss it here. If you don’t have anything valuable to add, you can always PM me your complaints.
Click to expand...

Well..if you DID go over to beyond 3d forums you would see that it ISN'T a big deal

Hell I'll even link you to a relevant topic: http://www.beyond3d.com/forum/showthread.php?t=24352
 
PRIME1 said:
People are just as split on the issue there as they are here. So that cleared up nothing.
Click to expand...

So asking a bunch of people who are even LESS qualified is going to clear up the issue. Come on :rolleyes:
 
OldBoy said:
So asking a bunch of people who are even LESS qualified is going to clear up the issue. Come on :rolleyes:
Click to expand...
Clearly YOU don't want to discuss the issue. But I do have a programming back ground and I do believe in the full implementation of standards. As stated by someone in everyones favorite beyond3d forum....
DemoCoder has disabled reputation

Default
Even if the loophole has been "closed", the fact remains, the spec has been written as if VT was a requirement. If MS patches the spec now, or "reinterprets" the spec to allow it to pass, it just means the lobbying succeeded. But what about the other IHVs who were trying to build SM3.0 prior to this (NV isn't the only one), who put in extra effort to meet all of, what was then assumed, *required* SM3.0 features? That seems a little unfair that the spirit of the original spec is being changed just to satify one IHV's marketing requirements to slap a "SM3" moniker.

Now the market is fragmented yet again. Devs need yet another code path, and the "SM3.0" moniker in fact, means nothing, since almost everything that SM3.0 was supposed to have required is now optional.

They may as well called SM3.0, SM2.d or something. I thought the whole point of major version number jumps in shader models was to set some minimum level of required features.

SM3.0 seems to be a minefield as far coding is concerned, and it seems developers may as well just stick with 2.0
Click to expand...
 
PRIME1 said:
Clearly YOU don't want to discuss the issue. But I do have a programming back ground and I do believe in the full implementation of standards. As stated by someone in everyones favorite beyond3d forum....
Click to expand...

So, since you're so knowledgeable why aren't you heading over to the Beyond3D.com forums and pitching in your 2 cents? Wanting to discuss it on a forum where hardly anyone (if anyone) understands the issues present seems like a recipe for something else altogether ;)
 
My question still stands:
If you have a completely unified architecture, and use drivers to tell each pipeline what to do, and it meets all requirements of an API in hardware due to software telling that hardware what to do, is it compliant or not?
 
I'm sorry but at least one of demon coder's points is flawed and that is IHV's putting in extra effort for a 'required' feature for sm3.0 spec. I was brought up thinking asked questions generally get answered if asking the right person, if the sm3.0 spec was 'unclear' then why not (as an IHV) just ask MS about it?
 
This thread should be locked as well. The same thing will happen here as what happened in the original one. Why just start a new thread about the same thing? How can this one add anything that wasnt added in the first?

Only time will tell if this feature is ever becomes widely used or not and whether this is the big deal Nvidia fans ar making it out to be.
 
tornadotsunamilife said:
I'm sorry but at least one of demon coder's points is flawed and that is IHV's putting in extra effort for a 'required' feature for sm3.0 spec. I was brought up thinking asked questions generally get answered if asking the right person, if the sm3.0 spec was 'unclear' then why not (as an IHV) just ask MS about it?
Click to expand...

Notice also that he even contradicts himself by saying "the spec has been written AS IF..." as though the spec was written a certain way. He's guessing which means he has no idea what he's talking about.
 
BoogerBomb said:
This thread should be locked as well. The same thing will happen here as what happened in the original one. Why just start a new thread about the same thing? How can this one add anything that wasnt added in the first?
Click to expand...

Then again, maybe that was never the point of this thread ;)
 
PRIME1 said:
The R520 may have had problems with it so they dropped it.
Click to expand...
Yep, page 4 pretty much shows ATI has changed their minds or had problems with it.
 
I ask again, if VTF is such an amazing, integral and earth-shattering 3.0 feature, why haven't more developers decided to use it? Why haven't more TWIMTBP partners added support for it into their games?

Someone really needs to get some feedback from the likes of Carmack and Sweeney on this topic. Maybe that would put it to rest for the masses (not the !!!!!!s, obviously).
 
John Reynolds said:
I ask again, if VTF is such an amazing, integral and earth-shattering 3.0 feature, why haven't more developers decided to use it? Why haven't more TWIMTBP partners added support for it into their games?

Someone really needs to get some feedback from the likes of Carmack and Sweeney on this topic. Maybe that would put it to rest for the masses (not the !!!!!!s, obviously).
Click to expand...

Even if they do include it, is it more beneficial than ati's workaround? (I'm going to presume that it will be included tranparantly in drivers sets in the future)
 
tornadotsunamilife said:
Even if they do include it, is it more beneficial than ati's workaround? (I'm going to presume that it will be included tranparantly in drivers sets in the future)
Click to expand...

The devil is in the details. :p
 
John Reynolds said:
I ask again, if VTF is such an amazing, integral and earth-shattering 3.0 feature
Click to expand...
ATI was the one hyping it up at GDC and included it in the R500/Xenos, and MS thought enough of it to include in the SM3.0 specs. Why don't you ask them? :D

Cue Nirvana's "All Apologies"
 
John Reynolds said:
I ask again, if VTF is such an amazing, integral and earth-shattering 3.0 feature, why haven't more developers decided to use it? Why haven't more TWIMTBP partners added support for it into their games?

Someone really needs to get some feedback from the likes of Carmack and Sweeney on this topic. Maybe that would put it to rest for the masses (not the !!!!!!s, obviously).
Click to expand...
I don't think Carmack is a big DirectX fan. (Although I am sure his knowledge on the subject is vast. Sweeney would no for sure since he has already commented on the UT3 engine using SM3.0 from the ground up. Hopefully we will hear something soon.
 
pxc said:
ATI was the one hyping it up at GDC and included it in the R500/Xenos, and MS thought enough of it to include in the SM3.0 specs. Why don't you ask them? :D

Cue Nirvana's "All Apologies"
Click to expand...

Don't expect replies from me if you're going to start a discussion with insults or childish, snarky comments.
 
OldBoy said:
Take your own advice. You're not technically competent to even understand the issue and you asking a bunch of other people who aren't either adds nothing of value to anything around here.

If you were in fact as "interested" as you "claim" to be, you would take the five seconds to head on over to Beyond3D and either read the informative posts about it in the forums by people who do have the technical knowledge to properly discuss the issue or sign up and post your question in the forums to satiate your "interest".
Click to expand...

Oops you are right, I also had no business in this thread - as outside of reading b3d - I understand little.
:eek:
 
John Reynolds said:
Don't expect replies from me if you're going to start a discussion with insults or childish, snarky comments.
Click to expand...
Hey, when is asking the primary parties childish? That's the way to get the most direct answer(s).

I would ask ATI:
1. Why did you present vertex texture fetch as a major VS3.0 feature at GDC in March this year, but say it doesn't matter now?
2. Why did the R500 get this feature, but not the R520?
3. Will the R580 get this missing feature?

I would ask MS:
1. Is partial compliance still compliance, since that would be a major change that started with DX8?
2. Will the VS3.0/SM3 spec get amended to VS3.0a/SM3a or similar to reflect this reduction in features?
 
pxc said:
Hey, when is asking the primary parties childish? That's the way to get the most direct answer(s).

I would ask ATI:
1. Why did you present vertex texture fetch as a major VS3.0 feature at GDC in March this year, but say it doesn't matter now?
2. Why did the R500 get this feature, but not the R520?
3. Will the R580 get this missing feature?

I would ask MS:
1. Is partial compliance still compliance, since that would be a major change that started with DX8?
2. Will the VS3.0/SM3 spec get amended to VS3.0a/SM3a or similar to reflect this reduction in features?
Click to expand...

Nice try. The best question would be to ask MS what exactly is entailed in satisfying the SM3.0 specification. Your assumptions as to "partial" compliance are quite weak.
 
OldBoy said:
Nice try. The best question would be to ask MS what exactly is entailed in satisfying the SM3.0 specification. Your assumptions as to "partial" compliance are quite weak.
Click to expand...

No need to ask MS. The X1800 XL/XT shows all VS3.0 caps as exposed, which means full compliance in their world.
 
John Reynolds said:
No need to ask MS. The X1800 XL/XT shows all VS3.0 caps as exposed, which means full compliance in their world.
Click to expand...

Thanks. There you have it. Discussion over. Case closed.
 
OldBoy said:
Nice try. The best question would be to ask MS what exactly is entailed in satisfying the SM3.0 specification. Your assumptions as to "partial" compliance are quite weak.
Click to expand...
At least you realize that asking MS and ATI is not childish. We can disagree on the questions. :D

Continue your spamathon.
 
Status
Not open for further replies.
Back
Top