Matrox Parhelia truth..

seppo is wog(ston)/orange and me is stefan/dxm.

now we are introduced (who ever you are are).


anyways.. i hope matrox stays alive, but preferly doing what they where good at: delevering what they should, and having good d3d drivers. add to that topnotch image quality etc.. and you should have matrox.. but the matrox parhelia just makes no sense...

as i said, i asked for cash refund.. i hope they will make a parhelia II where they will support all feutures they claim to have.
 
I feel kinda stupid now. ;)

I just read your first post more closely, and you asked about the SDK. Have you gotten a response about this yet? (I do see you asked for a refund, so it may not matter, but I'm curious.)
 
keegdsb said:
I just read your first post more closely, and you asked about the SDK. Have you gotten a response about this yet? (I do see you asked for a refund, so it may not matter, but I'm curious.)

mr.Omar claims it doesnt exist, so only sales etc knows about it.. or its all just rubbish (my guess).
 
stefandxm said:
seppo is wog(ston)/orange and me is stefan/dxm.

now we are introduced (who ever you are are).

sorry that I didn't introduce myself... I am just Nappe1. (I am as well as known as biggest and hypest Bitboys / Matrox fanatic in the whole net... ;) ) and actually only guy you can mix up me is the Tronic, because he has identical real name. In any case, I am long time demo scene lurker, (actually all way down to 1993 or 1994 I think.) but it took all way up to this year Assembly to make my own entry, which didn't even make to big screen. :)

things that I have heard about Parhelia II project aren't really convincing, but it seems that there's some hope about Parhelia 8x that should fix majority of Parhelia's first revision problems. (FAA with stencil, AGP8x support, and some other improvments...)

So there's hope. Stefan: you might want to take a look of Matrox forums. on "Industry chat" there has been some discussions about Parhelia's VS 2.0 support and getting it on drivers, so that might be good way to start. you know, you aren't only one wanting DX9 drivers for Parhelia. :)

for example:
http://forum.matrox.com/mga/viewtopic.php?t=1812

EDIT: I just found out that you already have posted to bug report forum so... you might have seen already thread that I linked above...
 
I will sabotage Stefan's attempts to be "angry", he been telling non-stop how he loves the image quality on the card. The fact that it's the card he has on his system tells that he's not against Matrox by large margin.

When I saw the displacement mapping videos, the character skinning and the landscape.. my jaw dropped to the floor, since that kind of functionality is precisely what I been waiting for.

If I may rehash to the first post, basicly, we need two things:

1. sampler2d on vertex program/shader, this will allow us to write something like this (Cg/HLSL follows):

// read displacement value
float disp = tex2D(dmap,v.texture0);

// displace vertex
float4 pos;
pos.xyz = v.point + v.normal * disp;
pos.w = 1.0;

// ... rest of the vertex program

This would implement only vertex-accuracy displacement mapping, but still, the reads from displacement map would be filtered, etc.etc. and done inside the GPU. Currently similiar trick needs to be done in software using FPU, SSE or similiar. The especially slow part is moving geometry through the bus.

OK-- presampled displacement is possible already, if we store in the vertex presampled displacement value, and have a global "uniform" in HLSL terms variable with strenth. Atleast we can modulate the effect strenght. But this isn't anything different from tweening anyways and nothing new.

The "cost" is same, float per vertex, or in the displacement texture. The displacement mapped version consume texture coordinates.. but often it is arranged so that different channels share coordinates (possibly with different bias and scale). In landscape texcoords can be computed, and is very easy, especially if little texture coordinate packing on steep slopes isn't a problem (texel density is variable of slope on planar mapping).

Yadda yadda.. the (1) combined with 2. Adaptive Tesselation, would "0wn", since this would allow the data to be streamed from GPU's local memory, which is very high-performance. The interpolation of texture coordinates when tesselating triangles would sample unique values from the displacement map, not just filter the existing ones from vertices. This is why both features are needed for high-fidelity results.

...

Just random thoughts on the topic. Then some gaming related thoughts:

- Halflife 2 can't be released a day too soon

- Max Payne 2 also has beautiful graphics what I seen from published screenshots and if it plays like the previews indicate, that game can't be released a day too soon either

- Then DOOM III looks absolutely brilliant aswell

Just amazing stuff coming down. Halo2 on xbox.. must work NOW as much as possible, because I'm going to be 'sick' suddenly when these games begin to come out... anyone else a hopeless game-addict?
 
Doesn't the Parhelia have VS 2.0 hardware? That won't be exposed if they don't have DX9 drivers.

BTW, I was quite impressed when I found out that the integrated ProSavage8 core has DX9 drivers. No DX9 features, of course, but still impressive.
 
It would be better if they finally manage to make a driver which supports T&L the Savage 2000 has in hardware, or take care of the Savage XP, both of which are desktop chips.
 
parhelia said:
It would be better if they finally manage to make a driver which supports T&L the Savage 2000 has in hardware, or take care of the Savage XP, both of which are desktop chips.

They don't sell any desktop chips. It's a waste of time and money to develop drivers for dead hardware that nobody uses. Although IMO companies should make detailed specs and driver code of old hardware available to the public, so that public domain drivers could be written if there's still some interest in the hardware.
 
stefandxm said:
keegdsb said:
I just read your first post more closely, and you asked about the SDK. Have you gotten a response about this yet? (I do see you asked for a refund, so it may not matter, but I'm curious.)

mr.Omar claims it doesnt exist, so only sales etc knows about it.. or its all just rubbish (my guess).

There's or at least there was an SDK that exposed Dismplacement Mapping trough DX8.1.
It used some tricks (passing non-standard data trough vertex declaration, and using a spare texture stage).

I have v0.9 of the SDK, I don't know if they ever done a v1.0 of it.

Oh, yeah and as I know it Mr. Omar no longer works for Matrox...
 
I'd say Matrox has much more important driver-side things to worry about than some unused displacement mapping feature :)

Their drivers are a slow, bug ridden mess. Adding more potential bugs isn't going to help anyone.

I'd also say they need some management turnover ;)
 
incurable said:
parhelia said:
"They don't sell any desktop chips"

What about DeltaChrome?
Nothing. Since it's been announced. Half a year ago. Kind of enforces his point, doesn't it? ;)

cu

incurable

DeltaChrome was planned to be available on H2/2003. (they talked about getting developer samples out on early Q3/2003 at Cebit.)

so, if we look from their point of view, they are still in schedule, but is that schedule too slow is another question...
 
parhelia said:
"They don't sell any desktop chips"

What about DeltaChrome?

I'm sure that when it's out they'll support it. But currently they don't sell any desktop chips and haven't sold any since the "old" S3 died. It's therefore quite reasonable of them to concentrate on driver development for what they do sell.
 
Back
Top