Doing a Parhelia or How to Break Promises

[jackzod]

Originally posted by Wombat
I'm glad I sold my P while it was still "going to be" a DX9 part.

Same here...

[Ragnarok]

Well, I simply don't understand why everyone's complaining 'bout "DX9 feature or not", "broken promises" and other things like that.
Just take the card "as it is"!!
No matter how many times someone's telling you, that the P. was intend to be a "gaming machine", it's time to get real!
The P. is primary a desktop-card, better used for high quality dvd & video playback, any kind of multimedia processing and visualization, InDesign-product workabouts and last but not least workstation based construction and design computing.
And it's really good at that!
Sure, there are better cards in that kind of use, but as twice as expensive, like 3DLabs products.
The P. is a workstation-card with gaming-support, N O T a game-card with workstation based software support and was never ment to be that way.

Of course, there were and still are problems with the kind a way Matrox handels supporting issues, but show me any other company perfect in it.
ATi? No, don't make me laugh.
Thinking about RagePro128-times or better RageFuryMAXX .... horrible.
Or nVidia? Newer driver-releases sometimes 20% slower than older ones? If you call that perfect driver-support, I don't wanna know the opposite of that.
Or performance promisses? Thank god that there's ATi and nVidia, where for example the 9600 Pro ( slower than the 9500 Pro ) and the GeFoFX5600 Ultra ( just crap at all ) are slower in a new Game ( Chrome for example ) than the over 1 year old P.?
Yes, ATi and nVidia keep their promisses, releasing an endless bunch of buggy cards in 50MHz-steps just to have 10 "different" models. Great. A "must-buy". *lol*

But do as you please ... sell all your P's and get happy with nVidia or ATi. And then start posting your complaints in their forums, cause there will be enough to talk about for sure.

Greetz,
R.

[warp]

Ah well... good news is that nVidia users are in the same boat.

nVidia (still) claims that the FX line is fully DX9 compliant while it's not. That's why all major gaming software houses have to revert to special coded nv30 paths instead of using the standard ARB2 path

[typedef enum]

On the topic of ATI:

Come on, it's been how many years since either of those 2 products were released? Geez, they've released the 1st generation Radeon, the 8500 series, the 9700, the 9800, and now the 9800XT. That's 4 generations of Radeon's.

To a man, nobody will dispute the fact that ATI has come full circle in terms of driver support. Take a good look around and see what real developers are saying about ATI drivers.

On the nVidia side, I have been very critical of their whole operation for the last 18 months or so. Let's face it, they've really tarnished their image. Having said that, the statement about their products NOT being DX9 compliant is absolutely not true. The fact that developers are providing special codepaths is NOT indicative of the product being non-compliant at all; what it does mean is that the FX lineup does not handle DX9 very gracefully without special care, especially in the area of pixel shading 2.0.

[Umfriend]

Well said. I do think it is a bit of a strech to call 9700, 9800 and 9800XT 3 generations though. Might as well add the 9000 then.

[Tomasz]

The GeForceFX is fully compatible with DX 9. Its pixel shaders are just slow in general and very slow when used in full 32 bit precision. To gain speed you have to either fall back to 1.1 shaders or selectively use partial precision shaders. It is a pain for developers, but it is done because of the very large number of people who own nvidia cards.

No one is going to write a special rendering path for the Parhelia because of its scant market share. It would significantly extend development time and would probably benefit only a few hundred people. Furthermore, without DX 9 pixel shaders, the Parhelia will never run a games' DX 9 path; it will always fall back to DX 8. Even if matrox enabled VS 2.0 in drivers, game manufacturers would still have to add a custom PS 1.3 and VS 2.0 path for it to be used. That is just not going to happen.

Regardless, just for the sake of their reputation and in order to appease their loyal customers, Matrox should still enable VS 2.0 in drivers. Who knows, someone might find a use for them. Otherwise, it seems as if they are breaking promises, and this rightfully upsets many people who shelled out nearly $400 for this card.

[RedRed]

I honnestly dont see that much of a problem - the card is what it is - good for business

If it DID support DX 9 it would probably only manage 10 FPS anyway - what would be the point?

Its like a company going for ISO9000 - they can promise 99.9 % of calls answered within their target - but if the target is 3 days - does it matter?

RedRed

[rylan]

So you guys don't have a problem with the new disclaimer for 10% clock variation in addition to the other specs that were removed from the matrox site?

[Tomasz]

Originally posted by rylan
So you guys don't have a problem with the new disclaimer for 10% clock variation in addition to the other specs that were removed from the matrox site?

The clockspeed issue is the most sleazy. Pragmatically speaking, the VS 2.0 issue is inconsequential, but it is a PR nightmare.

[warp]

Originally posted by Tomasz
The GeForceFX is fully compatible with DX 9.

Sorry, but its not (according to Microsoft themselves):

Developers at the Meltdown Microsoft DirectX conference heard Nvidia make an interesting comment about its drivers.

One of the developers using FX 5900 hardware for an upcoming title tested the new shaders 2.0 and made a comment that he'd had problems with them.

Apparently, some 2.0 shaders simply do not work making Nvidia's claim that this card is DirectX 9 hardware puzzling, as Pixel Shader 2.0 is one of the essences of DX 9.

nVidia seems to be screwing their customers too

[Tomasz]

Well, Valve has stated on Beyond3d that the GeForceFX 5900 will use only PS 2.0 in HL2, although most shaders will be partial precision.

Does the "Mixed Mode" use any PS1.4 shaders in place of PS2.0 shaders used in the default DX9 path?

[Brian Jacobson] Not for the 5900. Yes for the 5600 and 5200.

Does the mixed mode use the precision hint on all available instructions, most instructions or are the majority still full precision?

[Brian Jacobson] The majority are partial precision; about 10%-20% are full precision.

PS 2.0 on GFFX seems to have some problems but it still appears to be mostly compatible with DX 9.

[sparky001]

As I remember it a requirement for DX9 compatibility was a 8 pipe architecture. but I am not certain about that.

[TransformX]

I voted "don't care".
It's simply history repeating itself (G200), this is why I didn't get a Parhelia in the first place (so much money for promises once more ? No thanks)...

[warp]

Originally posted by Tomasz
PS 2.0 on GFFX seems to have some problems but it still appears to be mostly compatible with DX 9.

Yes, but that is *only* true for the FX5900/5950. And even these high-end models are not 100% DX9 compliant.
The mainstream FX cards (5200, 5600, 5700, etc.) are not true DX9 cards (PS1.4 is DX8.1 spec...not DX9) but nVidia still does advertise them as true DX9 cards.

Also keep in mind that compatible is NOT the same as compliant.
Heck, even the GF2 and the G400 are DX9 compatible, but both are not DX9 compliant.

Edit: Fixed some typo's.

[Che Guevara]

The P is definitly a good piece of hardware, but the Matrox action with the VS 2.0 is also definitly a shame!

Further, that's not an execusion for the Matrox guys, but the support of the other companys is not better. LoL @ any guy suggesting me to buy a ATI product again , I've not forgotten R.F. Maxx (2 driver releases ever! no win2000/xp/linux driver ever!! for a 600 DM card!!!) and I see massive problems on the R 8500s two of my friends use, since the release of the cards till today, ATI drivers are just terrible.