ATI DR Presentation ( Slides )

[Uttar]

www.notforidiots.com/GPURW.php

The following ATI for-developer presentation, made of 16 slides, has always been given to me in its integrality. Slide 1 and 11 have already been leaked, and I've been the person between the poster and the original source too. My decision to only leak those two slides at that time was due to the wish to leak the other 14 slides a few days later by myself.
Slide 11 is put partly out of context without the rest of the presentation, but this was not done on purpose. The general idea, in my opinion at least, remains unchanged with the other slides, although the statement is less "evil" in context.

The most shocking in this presentation, in my eyes, is not Slide 11. It's the wish, through the full presentation, to destroy special code paths, in an effort to "focus on gameplay", according to ATI's PR statement. Not anywhere as "evil" as some people would like you to believe, but still! No reason to be proud of it.

It has also come to my attention that several persons related to this affair are suspecting NV PR, as well as some of its employees in particuliar, of being the source of this presentation. Such a rumor, or perhaps speculation, while being founded on some correct information, is not accurate. The source of this presentation is part of NVIDIA, but not in their PR department.
And it ain't just NV spying on ATI - otherwise, how do you explain that numerous sources claim ATI canned the R400 after looking at the NV40 specs?

If you quote me on this, please at least quote the paragraph you're quoting in its integrity.

Also, a full-blown editorial will be released in the next 1-2 weeks.

Link: www.notforidiots.com/ATIDR


Uttar

[poursoul]

wierd...

The most shocking in this presentation, in my eyes, is not Slide 11. It's the wish, through the full presentation, to destroy special code paths, in an effort to "focus on gameplay", according to ATI's PR statement. Not anywhere as "evil" as some people would like you to believe, but still! No reason to be proud of it.

IMO that's not bad at all. It's good.

[Uttar]

Not really IMO.
To me, that's like saying "Okay, if you're in our marketing program, try not to optimize too much for nV." - remember that presentation is headed to developers, regarding their "Get In The Game" program AFAIK.

But as I said, yeah, it's not particularly big. Maybe I shouldn't have used the word shocking...


Uttar

[poursoul]

i guess i'm just taking it with less emphasis than you are. but i see your point.

Sorry to see this blow up in your face uttar. gl.

EDIT: DEAR GOD!!! i just went over to rage3d just to see the flames that you were putting up with over there, and DEAR GOD hellbinder really ripped into you didn't he! some people. anyway, chin-up.

[mrsabidji]

And I thought ATI was all-love for us gamers... Oh, the disillusion. :D
Anyway, I think it's actually shocking.

mrsabidji

edit : on second thought, that's not really shocking, you were right Uttar :D .

[reever2]

Originally posted by Uttar
Not really IMO.
To me, that's like saying "Okay, if you're in our marketing program, try not to optimize too much for nV." - remember that presentation is headed to developers, regarding their "Get In The Game" program AFAIK.

But as I said, yeah, it's not particularly big. Maybe I shouldn't have used the word shocking...


Uttar

How is this different than CG or the TWIMTBP program?

[Skuzzy]

I am glad they are pushing this. It's about time.

A standard code path is what these standard API's are all about. It does not rob a hardware company of doing things to make theirs better and faster, it just means games may actually come out on time and potentially underbudget, while giving ALL gamers a great product.
It also does not remove the possibility of having unique features, but most features can be done with shaders now, so the need to have something specific that works outside of an API standard is becoming less of an issue.

If you chose to think this is a bad thing, then you are not seeing the whole picture very well. From a developer standpoint it is a dream come true, hopefully.

[mrsabidji]

Originally posted by reever2
How is this different than CG or the TWIMTBP program?

I guess it's the "try not to optimize too much for (other IHV)" part. I may be wrong, but I think there's a small difference between asking devs to optimize for your hardware and the above proposition.

mrsabidji

[Skuzzy]

Originally posted by reever2
How is this different than CG or the TWIMTBP program?

Huge difference. Cg was not a standard. Never was. The TWIMTBP was designed to get proprietary support for non-standard features of the NVidia product line. The program did not push developers to use the standard, but to code around the standard.

[Skuzzy]

Originally posted by mrsabidji
I guess it's the "try not to optimize too much for (other IHV)" part. I may be wrong, but I think there's a small difference between asking devs to optimize for your hardware and the above proposition.

mrsabidji

The above says to write to the standard API. NVidia has been very successful at convincing people that code should be written specifically for the hardware it is to be run on. In this case, their own.
That is so far away from what DX and OpenGL are about as to be quite silly. That is where we were before DX and OpenGL came about. Anyone remember those days? It was a nightmare to write a game for a system and support all the various graphics cards out there. And those were simple cards! Today..I cannot imagine going through that again with the level of complexity the graphics cards are at.

[mrsabidji]

kinda off topic : Was Cg an nv-hardware specific language ? I'm just beng curious...

mrsabidji

[mrsabidji]

Originally posted by Skuzzy
The above says to write to the standard API. NVidia has been very successful at convincing people that code should be written specifically for the hardware it is to be run on. In this case, their own.
That is so far away from what DX and OpenGL are about as to be quite silly. That is where we were before DX and OpenGL came about. Anyone remember those days? It was a nightmare to write a game for a system and support all the various graphics cards out there. And those were simple cards! Today..I cannot imagine going through that again with the level of complexity the graphics cards are at.

Yeah, I remember those day (has a gamer, though). But I just think that just because there's a standard doesn't mean that every not-100%-standard hardware should be left aside. Especially when a huge number of people actually own that hardware. That's not an excuse for nvidia's (intentional) mistakes for that matter, but it's certainly not ATI's job to 'push' devs to do so either, if you know what I mean.

mrsabidji

[reever2]

Originally posted by Skuzzy
Huge difference. Cg was not a standard. Never was. The TWIMTBP was designed to get proprietary support for non-standard features of the NVidia product line. The program did not push developers to use the standard, but to code around the standard.

Really? How come every program using Cg except for tomb raider doesnt have any option to NOT use cg?

[Skuzzy]

Maybe because I had nothing to do with them? :D

Money and ignorance are a bad thing when used in combination. Some devs were convinced by NVidia that Cg would be the holy grail for shader development. Those devs are coming around now.
Like I said, NVidia did a great job of marketing, but it ultimately had no substance.

[The Baron]

Originally posted by mrsabidji
kinda off topic : Was Cg an nv-hardware specific language ? I'm just beng curious...

mrsabidji
No. Anyone could write a Cg compiler--the specs were all open. The problem was that yes, NVIDIA had final say over the language.

[Toaster]

I liked how the some of the slides where rotated 90degrees :p my neck still hurts..

I still think the Cg/Glide comparison is false, it's like compairing apples to oranges...

kinda off topic : Was Cg an nv-hardware specific language ? I'm just beng curious... mrsabidji

Kinda :
-Cg allows to use float, half and fixed
-the Cg compiler also does it's best to optimize for the GFFX ( like minimizing the num of registers )

For the rest, hey it's a non public presentation, what ya expect :)

(makes me curious about what some of the current Nv slides look like)

[StealthHawk]

Originally posted by Toaster

-the Cg compiler also does it's best to optimize for the GFFX ( like minimizing the num of registers )

It doesn't do a very good job of that. It actually hurts ATI performance in Tomb Raider than it helps NVIDIA performance :mad:

[demonic]

Originally posted by poursoul
EDIT: DEAR GOD!!! i just went over to rage3d just to see the flames that you were putting up with over there, and DEAR GOD hellbinder really ripped into you didn't he! some people. anyway, chin-up.

edit: nm found what i needed.

[Sazar]

Originally posted by Uttar
www.notforidiots.com/GPURW.php



Link: www.notforidiots.com/ATIDR


Uttar

why were only 2 slides released in the first place... I still fail to see the point of just releasing a couple of slides which do not deal with the point that you are taking issue with uttar...

?

btw... I thought a fair number of us were for the removal of special code paths as this would make it easier for everyone.. especially when it comes to costs for devs..

think... we could be playing half life 2 and doom III now or in a more recent future than it is likely we will if they were developed using standardized paths only...

its not a bad idea to support other paths for different architecture but it is ludicrous for US... the consumers... to have to wait perenially for products...

anyways... I was curious about the first question I raised... :)

ta for link btw...

[AthlonXP1800]

Here is a problem with pag16.jpg, the image is about 20% incompleted. I tried refreshed the page and cleared internet temporary files, nothing help, can Uttar upload the complete file?

[Sazar]

Originally posted by AthlonXP1800
Here is a problem with pag16.jpg, the image is about 20% incompleted. I tried refreshed the page and cleared internet temporary files, nothing help, can Uttar upload the complete file?

I am gonna attach image... hopefully it will show... it is showing up for me

http://www.notforidiots.com/ATIDR/pag16.jpg

[vandersl]

Ummm - are people reading the same slides as I am?

I didn't see a SINGLE bullet in there suggesting that developers shouldn't write separate code paths when needed. The primary focus of the slides was DX9 HLSL vs Cg, and the advantages HLSL enjoys.

All they want to emphasize is the use of a standard API and compiler, rather than Cg. I don't see what the problem is, or how this could possibly be construed as harmful to the consumer (especially since HLSL produces better code for the FX than Cg, by all reports).

Can anyone point to any slide or bullet where ATI says 'don't optimize for the FX'? Do they say 'don't support _PP'? I can't find it. All I see is their emphasizing that developers don't NEED to do that for ATI.

Much ado about nothing.

[astroguy]

Originally posted by AthlonXP1800
Here is a problem with pag16.jpg, the image is about 20% incompleted. I tried refreshed the page and cleared internet temporary files, nothing help, can Uttar upload the complete file?
when that happens.. hold onto ctrl when refreshing.. which will bypass the temp internet files..

[theultimo]

Well myself wants a standard, just so I don't have to find a specific patch, or wait longer for games to come out....


Remember the days of NO API? PowerVr with METLGL, 3dfx with Glide, ATI with 3dsomething..., Rendition with theyre own, and..........nVidia coding to the Direct3d "standard" which made a lot more sence, IMO. Instead of WipeoutXL with 5 different coding engines, there is only 1 which each card can use.....
Makes things a bit faster but longer development times, specific features not working on one card and the next (ATi 3d Rage Pro w/Bilinninar(sp?) Filtering, next to Matrox Mystque and S3 Virge with none)...........................................ha ve we gone BACK INTO THE STONE AGE OF GRAPHICS again?????????????

[rth]

wtf? ATI say they want developers to use standard codepaths so we don't end up with ridiculous HL2-style-5x-longer-development-because-nvidia-****ed-up-their-cards and you think this is a bad thing?

imo ANY kind of non-standard path is a bad idea. it extends development time just to make up for other people's mistakes and ends up being useless once new cards are out.

ATI telling developers to focus on gameplay is a GOOD THING.

this also has obvious benefits for ATI. if the FX's weren't quite so appalling at DX9 and HL2 was slow but still playable, then the game would be out by now. we'd all be enjoying it and the performance differences would be great advertising for ATI.

imo they should have released the game anyway and just made the recommended specs an ATI card. the special path could then be released as a patch.

it's a shame they didn't do this and make FX's use DX8, since in the end all their work was a waste of time and FX's use DX8 now anyway.