The russians cant listen to NDAs either.

[Marshmallowman]

Me thinks: pixel shaders are for making polygons ...etc look rounder and have the correct grain etc.., If you displacement map, all these features have already been rendered in*3D*, doing pixel shading is just the extra icing on top op the cake
(and also compatability with obsolete video cards )

[mattsteg]

So sad to see that people disrespect NDAs just to try to gain a bit of notoriety. The product is definately looking good so far though, as long as I can afford one. If not, I'll just be jealous.

[ALBPM]

They got their Noteriety all right. I think we noticed what a bunch of ****oles these people are.

Nationality aside...An ****ole is an ****ole!!!

And for those with any Integrity NDA means:

<FONT SIZE=4> No Damn Answers!!!<SUP>tm</SUP>...
<FONT SIZE=1><SUP>tm</SUP> courtesy of xortam <FONTSIZE=2>

I'm sure Matrox's info on May 14 will be better than this crap. At least it will be for real.

[Geek]

Yip, the May 14th Matrox announcment 100% real.

I still take that russian stuff with a pinch of salt.

Mark

[rugger]

Gee,

Stop it guys, your making me drool, and thats not good for my IBM keyboard!!!

[otlg22]

Well that all depends on which model keyboard you have.. if it's a Model M, I think it can take it

[rugger]

Originally posted by otlg22
Well that all depends on which model keyboard you have.. if it's a Model M, I think it can take it

Yep, a model M, even more reason to not wreak it

[otlg22]

Model M was my keyboard for a long time.. however I must admit MS Natural has won me over (but the one with the correct placement of the arrow keys)

[KngtRider]

no mention of if the mipmapping is improved to nvidia levels

that superfilter whatever its called is nice but prolly in supported apps, and what cost, see the ansio bug with the gf4

like very hard for me to find a game with bumpmap let alone embm............

also now we seebn the triplehead in action, does it have gap compensation between the displays do you think, and they got it to work in GL too, d3d is supproted in dualhead not gl atm

I whish i was under NDA

Sorry guys, the only way i can make the other aussie webmasters and gimps to eat hat is by beating a gf4's 3dmark score.

[mariner]

I'm just ignoring all this Russian stuff (didn't even bother to read)
and waiting for Tueday. I want to learn about Parhelia in Matrox Way!

[IcedEarth]

Originally posted by Ali
Just did the same with the 3Dmark pictures, and noticed that the cone of light comming out from the bottom of the plane isnt smoothed. Probably only works on 'real' edges, not alpha modifications.

I wonder if Matrox will allow the use of Super Sampling AA as an option in the drivers as well. If they have the fill rate to run 3 monitors at decent speed, they also have the power to run at 3X screen resolution.

It would be nice to have FAA-16X on by default, and use SSAA if you want the highest possible image quality, even if it means a performance hit

Just one mans thoughts.

Ali

I think this has more to do with resizing and compressing the picture than with FSAA.

[otlg22]

I tend to agree with Ali.. it looks as though the alpha effects weren't aliased.. probably something that would get fixed by release driver time.. remember those shots are _OLD_.

[Toastywheatman]

Originally posted by otlg22
TDB:

The pixel shaders are only 'techincally' not dx9 compliant. They _can_ do dx9 level work, but the DX spec calls for 4 shaders capable of doing a certain level of work, and the Matrox shaders appear to chained together so you only have 2 of them for more complex shading operations. (i.e. 4 simple shaders or 2 complex shaders, kind of like the decoder on the Athlon, it can issue a certain number of 'simple' ops a clock, or a smaller number of complex ops a clock)

I'm pretty certain Parhelia will not be able to carry out all PS 2.0 instructions. This goes way back to MS Meltdown last year where they released a great deal of info on DX9. The webpage is http://www.microsoft.com/mscorp/corp...sentations.asp and the DX9 powerpoint presentation is at http://www.microsoft.com/mscorp/corp...1/ppt/DXG9.ppt .

Here's some spec comparison:
PS 1.1/1.3 PS 1.4 PS 2.0
Maximum texture instructions 4 6 16
Address operations/instructions 4 8 32
Constant registers <=8* 8 32

* I could not find this value but I imagine it has to be less or equal to PS 1.4

Unifying two PS 1.3 class shaders does not yield the same number of registers or instruction count required for PS 2.0. The instruction set is also fundamentally different. PS 1.4 and PS 2.0 are much more RISC than PS 1.1. I recall all the experts saying the R8500 PS 1.4 engine is fundamentally different than the PS engine of the GF3 & 4, and yet MS and ATI advertise it as a big step towards PS 2.0.

I believe that Matrox wanted to launch a product before the DX9 generation of hardware was unleashed from ATI and Nvidia, and so they chose to go with the PS 1.3 model instead of waiting for the 2.0 model to be finalized. Ultimately I think this is in their favour because they will avoid a three-way race in which they would be the underdog to both other companies. This will allow Matrox to claim the performance crown for at least awhile (to increase mindshare of the company) and allow them to price more conservatively until R300 and NV30 are out the door.

-Toasty

[otlg22]

I see no reason why the Pixel Shaders can't handle both instruction sets.. Additionally the Parhelia is shown as having 256 Constant registers, which to me seems to meet the criteria.

[nggalai]

Hi Toasty,

I concur with you that the Parhelia won't be able to support PS2.0 for all the reasons you've already mentioned. In addition, PS2.0 ask for floating point values, while PS1.3 won't. No matter how many PS1.3 units are daisy-chained, you can't change that.

It's highly possible, though, that many DX9-style effects can be supported or approached by using dependant register reads and multipass rendering.

I believe that Matrox wanted to launch a product before the DX9 generation of hardware was unleashed from ATI and Nvidia, and so they chose to go with the PS 1.3 model instead of waiting for the 2.0 model to be finalized. Ultimately I think this is in their favour because they will avoid a three-way race in which they would be the underdog to both other companies. This will allow Matrox to claim the performance crown for at least awhile (to increase mindshare of the company) and allow them to price more conservatively until R300 and NV30 are out the door.

My thoughts exactly. Of course, this will only be a sensible thing to do if the Russian site was wrong and Parhelia will ship long before NV30. Let's hope for the best.

ta,
.rb