In-depth article on Parhelia @ Digit Life

[McElvis]

Looks quite a good article, though have not read it all yet.

http://www.digit-life.com/articles/parhelia/index.html

Warning ! Lots of pictures.


Interesting comments so far...

Vertex Shader 2.0, 4 parallel execution units (The current drivers support only v1.1 (not 2.0 because of the lacking DirectX 9.0), and a speed of operation of vertex shaders makes me doubt that 4 pipelines are enabled)

I must say that the Vertex Shaders 2.0 should be expected with the DirectX 9.0. It is possible that 4 units of these shaders will begin working to their full capacity only with the DX 9.0.

There are also a lot of components controlling stability of the card. The developers have done their best to prevent any pickups which can be caused by such a complicated layout (as you remember, the first high-speed cards on the GeForce2 Ultra looked terrible: ripples were well noticeable on the screen).

The TV-out works excellently! Apart from settings in the drivers which have everything necessary for TV image optimization, the Parhelia offers the DVDMax technology which allows displaying movies on TV in a full-screen mode, thus, making free the desktop on a monitor. Well, everyone has a too long way to go to the level of quality and capabilities of TV-out from Matrox!

[fav]

Here is a snapshot from (private?) conversation between Digit-Life/IXBT and unnamed person from Matrox (DL claims so) in reply to the article:

That was quite a long review, you must not have slept for the past couple of
days. Your review was very in-depth and an interesting read. However,
I just wanted to mention a few things that struck me while reading it.

With regards to our TMU's and vertex shaders, the driver is enabling their
full use on the chip, in other words, nothing has been disabled. However,
if a specific application doesn't call for quad texturing or deep vertex shaders or pixel shaders, a significant portion of our architecture will be sitting idle. These idle portions of the chip essentially future proof us for future applications that will take advantage of such features and we do anticipate the release of such complex applications.

Additionally, this is the first cut of our drivers. We are hoping that future
drivers may be able to more dequately demonstrate the benefits of our
architecture moving forward.
___

(Is it correct to cite private mail exchange in public?)

[tt22]

Abt the vertex shaders,. If he claims the applications don't use the 4 shader pipelines, how come the gf4's 2 seem to work perfect.

I'll be dissapointed if the new ATI r300 only can use 1 of its 8 pipelines

[CaineTanathos]

so basically 3dmark sux ?

[Evildead666]

yup

[R.Carter]

Who would make a 3D engine that requires 4 pipelines when there is only one card that would make use of it and the current market leader only has 2?

For fun, lets take a look at some older versions of 3DMark and compare a GeForce2 Ultra to a GeForce4 benchmark scores.

GeForce4 Ti 4600 3dmark2k.
GeForce4 Ti4600 3dmark 2k1
GeForce2 Ultra 3dmakr 2k

Any particual reason why a GF4 (18404) scores so close to GF2 Ultra (14376)?

If we didnt have 3dmark 2k1 the GF4 would look like complete CRAP!! .. "OMFG ONLY 4000pts HIGHER THEN A CARD 2 GENERATIONS OLD!?!! WTF!!"

[TdB]

don´t nvidia always say they double the performance for each generation?

the reason why we don´t see big performance jumps between nvidia cards is because they basically share the same core.

besides, these scores are kinda extreme, do they really represent the normal performance-delta between the 2 cards?

[Goc]

I could be talking out of my arse, but is there a way to "turn on" all for 4 pipelines via drivers... Don't know... Something like caching vertex shaders operations and then sending them to hardware in quads so that way all four units get their share of work?

[TdB]

I believe all 4 pixel/vertex-shaders already are working, however those shaders are heavily optimized for very complex shader programs, when game devs finds out that shaders can do more than just simple water-reflection/vertex-blending and start doing complex stuff, we will see the true power of those shaders, and not before that, it seems.
unfortunately that won´t happen before some high-level shading-language appears.

nvidias "cg", could just save matrox´s ass, whouldn´t that be ironic

seriously, all graphic-card companies have to agree on a HLSL standard ASAP.
Luckily a HLSL-standard will be a part of dx9(i´ve heard), hopefully it won´t be as nvidia-optimized as cg

[Edward]

I think the problem is most 3dengines are NOT optimized with the Parhelia in mind, I've seen what the card can do in NWO and Max Payne and that's a lot