More ATI Driver News from Derek Smart

[OpenGL guy]

]Sure it is possible to start from scratch, but to me, starting from scratch and from a different code base means just that : starting from scratch and from a different code base. Starting from scratch and from a different code base doesn't mean that you use, e.g. the same code and DLL from control panel or you execute the same branch of code to an op check (e.g. ZBIAS) in the same module at the same location....and from a module of the same name as the previous subset (i.e. the 8500).

Maybe you aren't so clever after all. The D3D driver for the Radeon 9700 is not ati3d2ag.dll

Example. I went in and disabled the ZBIAS checks (mentioned in my previous post above) I made for ATI boards (which I didn't have to do for Matrox, nVidia or ANY other board btw) and the problem - again - showed up on the 9700 Pro. You'd think that this is on thing that they'd actually fix in a new revision board. They didn't.

Zbias is an ill defined feature of D3D. For one thing, the refrast doesn't support it so you have no idea how much to bias the Z values for each bias level. How can you claim that ATi's implementation is broken when there's no reference to compare it to? Oh wait, you were probably comparing it to another IHV's implementation. The initial Radeon 9700 code used a similar calculation for Z bias as the Radeon 8500 because it worked well in most apps. I found one app where it didn't work well but I couldn't get a fix in that would work for this particular app and the other apps we were looking at before the release of the card. The 7.76 driver should have a better Z bias calculation.

Is it board architecture or driver problem? Its a driver problem.

Is it a board or driver problem? Please make up your mind.

I have experience in designing PCBs back from my old days. As such, I know enough to know that this particular ZBIAS problem is at a driver level.

What does designing PCBs have to do with D3D drivers?

[*]They removed W buffer support from the 9000/9007!! And they don't ever plan on supporting it because, according to them this will de-stabilize the driver and the card would take a significant performance hit

You know, if you really want W buffer, you can do it in the pixel shader. You're clever, you can figure it out.

 

[jb]

Derek/Dave,

thanks for the info and I did not ment he did influence them. Just trying to see if the old noodle was remembering things close to reality

 

[Derek Smart [3000AD]]

All I know is, like MT a LOT of functionality is broken in the 9700 drivers. I don't give a toss if PacMan runs fine or not. Obviously I was RIGHT or ATI would not have been able to reproduce the problem nor inform me that they were going to fix it asap. If there was nothing wrong and it was all on me, WHAT would they have to fix.

We are not saying that its not broken in your case, we are asking if this is because of the way that you are using it because there is no evidence in current games that indicates MT is broken.

Quite frankly I don't KNOW where or what got broken. If I did, I'd tell them specifically (as I always do). I simply sent them the info, some shots and some code notes.

Until I even got the response back (that they'd found it), I seriously thought that I had to do something differently for those cards. I didn't have a problem with that because, after all, it was a new card with some new architecture. I treated the instance as I did when going from DX7 to DX8. I didn't go crying to MS each time I found something that worked differently.

So, it wasn't until I heard from them that I knew for a fact that the driver was broken and that it wasn't on my end. Well, couple that with fog, specular, TnL etc etc - ALL things that used to work and just got broken at some point or another, I was just a bit peeved. And my response to them when I heard the urgency with which they were going to fix it (obviously it WAS a big deal to them - as it should be and has always been in my case)

arghhh!!! uhm, cool. OK, I take back every bad thing I've
said about you folks (well, not you specifically) all week!!!

Quite frankly, broken hardware TnL is more important to me than broken MT. Yet, you don't see me kicking up a storm about it, do you?

 

[mboeller]

can someone do the obvious and close this thread please !

 

[Joe DeFuria]

Here, I'll post it again for you because you make it seem like I'm here bitching about MT. Which goes to what I was saying before - ALL that you and your friends have been doing is TWISTING the very premise of the discussion and trying to make it about ONE THING. It seems that If you can't dazzle them with brilliance, baffle them with bullshit seems to be the mantra for you folks posting outside the realms of this discussion.

The truth is, Derek, you haven't made any complete and coherent arguments. You are basically "bitching at ATI for bad drivers."

We know you're doing that. We either don't know, or don't agree with, the basis for your rants.

ONE of the things you've very vocally bitched about is multitexturing. Another is w-buffer support. What else?

Ati dropped w-buffer support. They don't support 32 bit z buffer. End of story. (For all the power and influence you purport to have on this industry, ATI dared to drop w-buffer support or 32 bit z? Gah!)

Then there's the multitexturing issue, which is not evident on other games, you reported it, and it is being fixed.

Again, what are you bitching about again?

 

[Derek Smart [3000AD]]

Derek,

There is some very good technical people on this forum (from game developers .ie Daniel Vogel Epic Games, to ATI, PowerVR and Nvidia engineers), there is representatives from every major IHV here..so if want to debate this is the place to be

I know. I've heard from some devs via email already.

I will debate and discuss amidst all the personal attacks, smoke and mirrors etc, I suppose.

Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know.

 

[jb]

Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know

Yea but every once in awhile its kind of fun to do it

j/k we could use some sense of hummor around here...and no I was not that funny so I wont quit my day job

 

[OpenGL guy]

]Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know.

Yeah and it sure shows a lot of maturity.

 

[BRiT]

]Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know.

It's a lot easier when you're drunk...

--|BRiT|

 

[Joe DeFuria]

It should most definitely be possible to write a relatively simple program that tests nearly every possible configuration of various rendering states, and checks for proper rendering.

Please, tell me you are kidding, right?

 

[Mize]

]
Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know.

No, it's not, but it's real easy to crap all over a normally high quality forum like you've done today...no balls in that at all.

 

[Onde Pik]

Hehe looks like Derek scored.

Maybe Derek's games don't pull as much weight with the ATI driver team, as John Carmack does and maybe that is pretty logical. But I would just like to say that I couldn't care less about what Carmack says, Quake is the only ID game that I consider a good game. Actually I would like to say that we need more people like Derek Smart and less people like John Carmack. Now before everyone start waveing the "Carmack is God" flag hear me out.

I have been a gamer for more than fifteen years, and I have witnessed the evolution of games in that time. Until about 5-8 years ago, both presentation (graphics and sound) and gameplay increased steadily. Presentation more so than gameplay but oh well. Back then it was not the average joe who played on their C64, it was rather a certain type of people ... people like me. About 5-8 years ago it started to become more accepted to play computer, stuff like first person shooters, RTS games, 3d rendering, computer cafés and the Internet helped bring gaming to the masses.

The game industry boomed, game presentation boomed with it but sadly gameplay didnt follow, instead it decreased in quality. The masses didn't want their games to be too complicated but instead wanted better graphics. And they got their wish. For years all the best selling games have been either FPS, RTS or wannabe RPG games that aren't too complicated for the masses. Mainstream gamers naturally outnumber the 'old school' gamers by far, and thus the 'old school' gamers have become niche gamers.


What is the moral? Well there are not much money in niche gamers, and thus very few developers in developing niche games. Derek Smart is one of those developers and John Carmack is not. Some of the VERY best games I have played the last couple of years have been relativly low budget games. And the high profile games I have played have pretty much sucked. Diablo 2, WarCraft3, Max Payne etc. etc.

Id like to say that I respect what Derek is doing. And sadly I could never get the demo for 3000AD to work on my Radeon So I never tried it, and sadly I am too poor to buy it now since I am a student.

 

[Derek Smart [3000AD]]

Thanks for showing up. I am going to, quite literally, rip you to shreds for all the aggravation you bastards in driver development have caused me.

THAT and due to the context and tone of your post. Trust me, methinks you've bitten off more than you can quite possibly render.

Here, pucker up....

Maybe you aren't so clever after all. The D3D driver for the Radeon 9700 is not ati3d2ag.dll

Yeah. You're so clever Mr ATI guy that not only did you foolishly (a) misquote me (b) the misquote you pasted bore NO relevance to the attack you were making.

Show ME where I posted that that driver had anything to do with D3D

Good job. You're doing just as good as your driver development.

Zbias is an ill defined feature of D3D. For one thing, the refrast doesn't support it so you have no idea how much to bias the Z values for each bias level. How can you claim that ATi's implementation is broken when there's no reference to compare it to? Oh wait, you were probably comparing it to another IHV's implementation. The initial Radeon 9700 code used a similar calculation for Z bias as the Radeon 8500 because it worked well in most apps. I found one app where it didn't work well but I couldn't get a fix in that would work for this particular app and the other apps we were looking at before the release of the card. The 7.76 driver should have a better Z bias calculation.

Again. Your arguments have NO basis in reality and you're doing a BAD job of it. Even the trolls are doing a better job at posting nonsensical RUBBISH than you are.

Here, let me break it down for you....

REGARDLESS OF D3D DRIVER IMPLEMENTATION, I NEVER, EVER, EVER, EVER HAD TO MAKE ANY ZBIAS CHANGES IN MY CODE FOR ANY BOARD TO GET IT TO WORK. FOR THE ATI BOARDS, I HAD TO

There, get it now? Good. Lets move along with this farce, shall we?

AND, if your implementation of this was ok, why'd you have to go back and piss around with it in the 776 driver? Which, of course, is RUBBISH because when I was investigating this MT issue, I first thought that my ZBIAS exlcusion code for the ATI was no longer needed, hence the visual artifacts. Disabling them didn't help. And THATS when I passed it along to dev support.

So, from where I'm parked, the same crap thats in ALL the previous drivers which caused me to make those code exclusions to begin with, STILL exists in the 776 drivers. Good job!!!

Is it a board or driver problem? Please make up your mind.

What does designing PCBs have to do with D3D drivers?

You're kidding me, right? Doesn't a knowledge of hardware architecture lend credence to board design and hence software development for saidh hardware? Jesus H. Christ, how'd you get your job? Are their any vacancies? I know a real tool who's looking for a job. He'd fit right in.

You know, if you really want W buffer, you can do it in the pixel shader. You're clever, you can figure it out.

If you look at my post to Mike, thats what my first thought was, re my inference that I'd figure something out. Doh!!

The point is what about legacy games e.g. my 2001 game and so many before it, that do NOT have pixel shader support? huh?

 

[Elysian]

mr. smart perhaps if u know so much about their drivers and u can bitch so much(and are supposedly so smart) u should write ur own drivers and quit ur bitching

 

[Derek Smart [3000AD]]

]Its not very easy having to stand up in a crowded room and telling everyone to go fuck 'emselves you know.

Yeah and it sure shows a lot of maturity.

You might want to grow a sense of humour.

 

[Derek Smart [3000AD]]

can someone do the obvious and close this thread please !

No, siddown, fool.

 

[andypski]

Finally, why the hell would JC want W buffer support in drivers? The fact that you even use this example, just goes to show how clueless you are. He's NOT doing ANYTHING that even REMOTELY requires the benefits of a W buffer. *sheesh*. And do you think that even HE has the power to influence the architecture of a card? Even if he wanted to (which I know for a fact that he's not even likely to get involved in such)?

Hmmm... that's a bit of a puzzle...

You see, I know for a fact that JC has plenty of power to influence how cards are architected. And while he might not get directly involved, I think you will find that IHVs make some effort to get his opinions while they are architecting their cards.

Believe it.

 

[alexsok]

As far as I can say, this thread should really be closed, as it's obvious that no reasonable discussion will come out of this, every side has it's own complaints and opinions, u guys are not going to EVER agree on anything!

 

[andypski]

]

REGARDLESS OF D3D DRIVER IMPLEMENTATION, I NEVER, EVER, EVER, EVER HAD TO MAKE ANY ZBIAS CHANGES IN MY CODE FOR ANY BOARD TO GET IT WORK. FOR THE ATI BOARDS, I HAD TO

As a matter of interest, which other boards did you run it on using ZBias?

 

[Derek Smart [3000AD]]

]

REGARDLESS OF D3D DRIVER IMPLEMENTATION, I NEVER, EVER, EVER, EVER HAD TO MAKE ANY ZBIAS CHANGES IN MY CODE FOR ANY BOARD TO GET IT WORK. FOR THE ATI BOARDS, I HAD TO

As a matter of interest, which other boards did you run it on using ZBias?

Every board I have here. I have eleven dev stations here, running every manner of hardware. The problem showed up in the legacy Radeon drivers and got worse from there.

I never had the problem with any other board at all.

And as I said before, KNOWING that it was an UNFIXABLE DEFECT, I had to fashion a workaround. It was a messy proposition, but I had no choice.