nVidias Dawn now works on R3x0

[breez]

So in short, if Carmack is going to use OpenGL, I have no problems with using vendor specific extensions...he pretty much has too. However, I would have rather there been a D3DQuake, than GLQauke.

What was the current version of DirectX when carmack did GLQuake?
Early D3D APIs sucked bigtime.

Must've been DX3 or DX5. DX6 was the first with 'good' D3D

 

[Joe DeFuria]

The wrapper implements nVidia proprietary extensions.
If that was legal to do ATI would do it too in their drivers.

I don't see it that way.

The 3rd party wrapper is just a piece of software. Are you saying any software writer must get nvidia's permission to use their extensions? ATI doing it in their drivers would be different, because the drivers are a direct interface to their hardware.

The wrapper just a "translator."

Interestingly, I recall (I think it was Creative?) at one point had a GLIDE wrapper for some of their nVidia products back in the day. It was even promoted as a feature. I think there was a law-suit over that...but don't know how it turned out.

 

[♪Anode]

]

Didn't he do this for Doom3? I remember reading his last .plan commenting that he dropped support for all vendor specific rendering paths other than the NV10 path.

This is a point that gets brought up all the time. Its totally Misinterpreting what he actually said. No where does he say that he is getting rid of all teh vendor paths. Do you really think he is going to Bail out on the R200 path for instance??? Not a chance. Further if he did take out the Nv30 path, the Nv30 and 35 would go down in Flames to the R300 series. So obviously its still in there and working quite well.

I for one Think that the Nv30 and 35 should be FORCED to use the ARB2 path. Then well see whos hardware is really faster in Doom-III. I am already tired of reading all the *Nv3x cards are made for doom* nonsense, and the Iquirerer articles about how Ati needs to get their act together etc etc. When the whole thing is becuase R300 cards run it the *RIGHT WAY* and Nvidia cards run it in Reduced Quality mode.

Its just not right. Nvidia are totally being allowed to get away with reduced Quality for speed yet again (empowered by Id themselves).

I bet you wouldnt be saying that if it were ATI which performed lower on ARB2.
And who said reduced quality. I havent seen any comparisons of Doom3 IQ except from Carmack's quotes which say that its really hard to note the differences.
And using lower precision != reduced quality if your eye cant detect the difference. In fact it is called intelligent programming since you are only using the precision you need and improving performance. It is not Carmack's (or NV's) fault if ATI decided on doing fp24 all along and giving him no choice in situations where fp16 or even fx12 precision would have been enough.

 

[DemoCoder]

Yeah, the procedural interface is alot cleaner, and the method naming is logical and clean.

With Microsoft you get #define D3DLONGCONSTANTFLAGYOUCANTREADVERYEASILY

 

[jjayb]

So which one of you coding geniuses is going to fix this wrapper to work with athlon thunderbirds? So far everyone I've seen getting the error when trying to launch this is using an athlon thunderbird. SSE extension problem? Who's going to volunteer to fix this?

 

[Joe DeFuria]

I disagree. D3D, in the state it was back in the Quake days, was a real pain to code for. OpenGL had some shortcomings too, but at least it was understandable.

Again, I'm not saying that Carmack didn't have very valid reasons for choosing GL. That's not my point.

I think gamers went through quite a bit of pain during the whole "general gl ICD suckage" era. What carmack did, IMO, was shift the burden away from himself for taking up the challenge to write a D3D engine....

...and passed the burden on to several IHVs to develop GL ICDs. I actually see what Carmack did as kinda selfish.

But there are certainly positives though. I'd bet that if Carmack coded D3D, the "consumer" cards today wouldn't have GL drivers at all. The differentiating factor between the "Quadros" and "FireGL"s vs. the GeForces and Radeons, would be the PRESENCE of GL drivers, not just "pro app optimized" drivers.

So what carmack did was open up GL programming and GL consumers to a much wider audience. I'm just not sure that has any pratical benefit to the gamer.

 

[WaltC]

You guys are being a bit one-sided with all this ARB2 talk. Yea, NV30/35 will be slower but they'll also run the game at a higher level of precision. Personally I like ATi's 24fp since it appears to be a nice sweet spot for performance and IQ for this generation, but Nvidia's higher precision should at least be mentioned during these discussions.

Agreed. I don't have a problem with a vendor path used to run the game at a higher precision (than ARB2), necessarily--it's just that I'm afraid the vendor path will be routinely mistaken for higher precision, when it will probably be running at < than ARB2 precision. My complaint with fp32 is in knowing when, if ever, it's being used. So I would be much more comfortable with it seeing actual game examples in which it is used... nVidia says often that fp32 is superior to fp24, which I'd like to see on occasion so that I could judge for myself. We certainly know it is *slower*--but the "better" part strikes me as 100% nVidia PR at present.

Extentions allow the chip companies to expose the functionality of their hardware as it comes out. That's what had allowed Carmack to do any shader development in the first place. OpenGL is created by a committee, and if everyone had to wait for them to agree on a standard and finalise it (like ARB2), they might as well have moved to Direct3D.

I believe that Carmack hasn't scrapped these code paths because he has already developed them, and they provide better performance. He would be betraying his customers if he would scrap performance paths. It'd be an altogether stupid thing to do.

Well, as "shader function" is a non-extensive part of DX9, I see no reason why the same couldn't apply to OpenGL--without any need for extensions. The API needs to mature--and my point is that extensions may be what's stopping or at least slowing that process down relative to the API.

At the time Carmack first pushed extensions, many people were hot under the collar about GLIDE and wanted to see it disappear--because it was more or less a 3dfx-hardware specific API. Lots of anti-GLIDE folks at the time felt 3dfx was slowing the industry down with things like Glide. Some of the debate got through and before its demise 3dfx had already officially announced it was abandoning Glide as it felt the D3d and OpenGL APIs were maturing enough to be able to take up the slack. My point is if developers start having to support multiple sets of extensions for their OpenGL programs the temptation to avoid all of that by moving to DX9 will increase. A prime example of the bother that sort of thing can be to a developer is the NWN OpenGL extensions problems Bioware had that I mentioned earlier.

 

[Rodéric]

So what carmack did was open up GL programming and GL consumers to a much wider audience. I'm just not sure that has any pratical benefit to the gamer.

You would prefer a dictator ?
Having the choice is a good thing.

 

[OpenGL guy]

I think gamers went through quite a bit of pain during the whole "general l ICD suckage" era. What carmack did, IMO, was shift the burden away from himself for taking up the challenge to write a D3D engine....

...and passed the burden on to several IHVs to develop GL ICDs. I actually see what Carmack did as kinda selfish.

I don't see it that way at all. From what I recall of Carmack's comments on the matter, he had spent two weeks trying to port his software engine to D3D and gave up in frustration. He then looked at OpenGL and had the port completed in a weekend. Now, if he had used D3D to begin with, maybe MS wouldn't have improved the API to the extent that it has. Can you imagine if all we had was retained mode programming?! Egads!

Carmack made the right choice, then, because OpenGL was a better tool than D3D was. However, I think that has changed, except that OpenGL is the only cross-platform 3D API (you won't find D3D on Linux or Solaris).

 

[gordon]

Vendor Specific Vertex Extensions, not paths..vendor specific paths are still there.

I just found a link to the plan and youre right. Could have sworn he had made a statement similar to my original comment though.

 

[Doomtrooper]

I have come to think of OpenGL extensions as basically a very bad idea for the OpenGL API--I can remember that Carmack was roundly criticized for supporting them originally, although originally I thought they represented if not a good idea then at least a decent one. But it seems to me that it is very likely that extensions will only serve to fragment the API and to make things more difficult for developers (as happened with NWN)--which is bad for the general future of the API itself, IMO. To this end, I would like to see Carmack scrap the vendor paths entirely and strictly go to ARB2 to support in his software. The main question I have now is why he hasn't already done that.

Extentions allow the chip companies to expose the functionality of their hardware as it comes out. That's what had allowed Carmack to do any shader development in the first place. OpenGL is created by a committee, and if everyone had to wait for them to agree on a standard and finalise it (like ARB2), they might as well have moved to Direct3D.

I believe that Carmack hasn't scrapped these code paths because he has already developed them, and they provide better performance. He would be betraying his customers if he would scrap performance paths. It'd be an altogether stupid thing to do.

Vendor Specific fine, Proprietary..NO.

Vendor Specific still allows the IHV to use their hardware to its full extent, but by putting proprietary BS on it makes it bad overall for gamers, and their has been enough games released recently that show that, like NWN.

http://oss.sgi.com/projects/ogl-sample/registry/NV/texture_shader2.txt

vs.

http://oss.sgi.com/projects/ogl-sample/registry/ATI/fragment_shader.txt

You can see how many proprietary extensions Nvidia has vs. the rest of the ARB (which is none)

http://oss.sgi.com/projects/ogl-sample/registry/

 

[Joe DeFuria]

You would prefer a dictator ?
Having the choice is a good thing.

Isn't "dictating" exactly what Carmack did? Why didn't he code BOTH GL and D3D? That would've given ME (gamer) a choice.

As a GAMER (repeat 5 times: G-A-M-E-R. Not gmae "developer", not "3d Professional", not "demo coder"....GAMER). No, I don't care about having a choice between D3D or GL. I just want to play a game with minimal fuss.

Carmack had less fuss for himself by coding GL.

IHVs had more fuss by coding Gl drivers, (pass cost onto consumer) and consumers had more fuss with downloading GL drivers, mini-gl drivers, GL drivers for this game, for that game.....). At one point, there was a concerted effort (forget the name) "to make sure you got the right GL driver...)

 

[Hellbinder]

I bet you wouldnt be saying that if it were ATI which performed lower on ARB2.

Your probobly right...

In this case though, its not fair that both IHV's support the hardware necessary to do the job right, while one get the benefit of *completely* pandering and specialized coding and the other has to do it the right way.

Which results in Thousands of people talking smack about ATi and how much better Nvidia hardware is for Doom. And how better suited Nvidia hardware is for Future Games. and blah blah blah..

When exactly the Oposite is the real Truth.

 

[Joe DeFuria]

I don't see it that way at all.

OK, we can agree to disagree.

From what I recall of Carmack's comments on the matter, he had spent two weeks trying to port his software engine to D3D and gave up in frustration.

Right.

He then looked at OpenGL and had the port completed in a weekend.

Well, I don't know about the specific effort / amount of time, but basically, you are correct.

Now, if he had used D3D to begin with, maybe MS wouldn't have improved the API to the extent that it has.

Or (alternative view) maybe D3D would have been improved faster with Carmack's direct guidance and help. Instead, even though he believed it would "suck less" with each revision, he basically swore off D3D, in hopes of not "dragging the industry through another ill-birthed API". (I remember those little snippets pretty distinctly. )

Can you imagine if all we had was retained mode programming?! Egads!

Do you really think MS dropped retained mode just because Carmack dropped D3D?

Carmack made the right choice, then, because OpenGL was a better tool than D3D was.

Again, I just disagree. He certainly made an understandable choice though.

 

[Hellbinder]

Btw,

Carmack's quotes which say that its really hard to note the differences.
And using lower precision != reduced quality if your eye cant detect the difference. In fact it is called intelligent programming since you are only using the precision you need and improving performance. It is not Carmack's (or NV's) fault if ATI decided on doing fp24 all along and giving him no choice in situations where fp16 or even fx12 precision would have been enough.

Nice attempt at a Spin Job. I think the democrats could use a man like you in 2004

Make no mistake, Nvidia uses lower Percision modes becuase THEY HAVE TO.

 

[Hyp-X]

I don't see it that way.

The 3rd party wrapper is just a piece of software. Are you saying any software writer must get nvidia's permission to use their extensions? ATI doing it in their drivers would be different, because the drivers are a direct interface to their hardware.

The wrapper just a "translator."

A driver is a piece of software.
Being an interface to the hardware is just the function of this piece of software but it doesn't make it any less software.

Actually a driver is a "translator" since it's unlikely that the public APIs are closely matching the hardware registers / states etc.

Both a driver and the wrapper implements APIs. If there's a software patent to an API than you need permissions to implement it. (At least in countries where people were content to let such stupid thing as software patent exist in the first place.)

 

[Joe DeFuria]

Interestingly, I recall (I think it was Creative?) at one point had a GLIDE wrapper for some of their nVidia products back in the day. It was even promoted as a feature. I think there was a law-suit over that...but don't know how it turned out.

Just for further info...just to make sure each fragmented thought in this thread continues , some related info:

http://www.zdnet.com/products/stories/reviews/0,4161,2290046,00.html

http://www.cdmag.com/articles/020/118/3dfx_sues_creative.html

I don't know what became of the law-suit, but I believe the basis of it was that 3dfx claimed that Creative used 3dfx copyrighted source code to create the wrapper...not that the idea of the wrapper itself was illegal.

 

[FUDie]

Interestingly, I recall (I think it was Creative?) at one point had a GLIDE wrapper for some of their nVidia products back in the day. It was even promoted as a feature. I think there was a law-suit over that...but don't know how it turned out.

Just for further info...just to make sure each fragmented thought in this thread continues , some related info:

http://www.zdnet.com/products/stories/reviews/0,4161,2290046,00.html

http://www.cdmag.com/articles/020/118/3dfx_sues_creative.html

I don't know what became of the law-suit, but I believe the basis of it was that 3dfx claimed that Creative used 3dfx copyrighted source code to create the wrapper...not that the idea of the wrapper itself was illegal.

Hmm Aureal sues Creative and goes out of business. 3dfx sues Creative and goes out of business... Creative is the tar baby!!

-FUDie

 

[Joe DeFuria]

Both a driver and the wrapper implements APIs. If there's a software patent to an API than you need permissions to implement it. (At least in countries where people were content to let such stupid thing as software patent exist in the first place.)

Surely, 3dfx had a "software patent" on Glide? But their law-suit was not based on the premise that "wrappers" were of themselves illegal. (Wouldn't that be the most direct route to take?)

3dfx's arguments stemmed from illegal use of their copyrighted code / SDK.

http://www.glideunderground.com/61699/CL61699.html

All I'm saying is, there is no clear indication to me, that wrappers themselves are illegal. Are "Console Emulators" illegal? They don't appear to be by definition.

 

[Hyp-X]

Must've been DX3 or DX5. DX6 was the first with 'good' D3D

Actually in a lot of things DX8 had the first good D3D...