Matrox confirms Parhelia is AGP 4X

[Greebe]

Parhelia runs fine at AGP4x, so what's your point?!

[rugger]

Originally posted by ralf
The history of AGP 4x shows there is no future for AGP 8x.
Clever.
Besides, my benchies show 4x has an impact.

This thread started as a discussion on Parhelia being a 4x or 8x chip, and the possible advantage of 8x. So it´s a question of what the near future will bring us in terms of apps and games, and my point is that the evolution is towards very high polygon intensity and that this calls for the speedier AGP bus, since vertex data is sent through AGP.

But this seemes very hard to understand


I guess if Matrox hadn´t verified the Parhelia to be a 4x chip (at this time) there wouldn´t have been all this arguing about the virtues of AGP 8x.

No, the industry is not moving towards very high pologon counts. Polygon counts are going to continue to go upwards as they have so far, but the real focus (as shown in Doom III) is to move toward better shading and texturing effects. Displacement mapping is another technology that shows the movement towards this direction (although the card still tessellates into lots of polygons, the game/application does not need to describe them all)

High polygon count models are just hard to do when you have to do them for thousands and thousands of objects.

As for the future, it is ussually much like the past, so its best to start from there and predict forward. With that in mind, AGP 8x is going to be ANOTHER yawner for the graphics industry.

[Technoid]

I'd just wan't to point out that the 16MB limit of W2K's gart seem to be easy to go around.....

Even Nvidias can do it...

[rugger]

Originally posted by ralf
Well rugger, it comes down to curved surfaces vs polynomial surfaces, and while N-patches are a clever solution to low polygon count models they don´t bring more detail which parametrical surfaces do – and the latter allow smooth LOD changes on the fly.
And the winner is... yet to be decided.
So you vote for curved surfaces, that´s fine with me but I don´t.

And I think reasoning along the line “X didn´t make a difference therefor Y is useless” is a dangerous path since it can be transferred to a lot of situations and in essence negate development, e.g. “we´ve had dx8 for a year and it´s made no difference, so dx9 is just pointless”, “look, there is no HD nearly capable of r/w at 133 mb/s so with ATA133 who needs serial ATA. Worthless”.
And we could go on and on.

(rugger, by the time you read this I wouldn´t be surprised if it has already happened:
I´m being flogged for talking down dx9 and serial ATA )

No, I mean using the past as a model for the future, rather than relying on speculation (which says every technology is the next big thing and you cannot live without it). Its not simply saying everything that is going to happen in the future is not worthwhile. It actually taking the avaliable facts and making an educated guess on how a new technology will perform. It may not always be right, but it often pretty close.

Serial ATA has nothing to do with Parellel ATA. It could be quite good, given motherboards ship with plenty of connectors and there isn't any gotchas to them. Of course, it is too early to be investing in them, since the standard is still a moving target. And the benefits of Serial ATA will have nothing to do with transfer rate.

With Direct X9, I AM using the past as a base to move from. Direct X8 has been out a long time, yet games that extensively use and rely on its features have not been released yet. Since Directx 9 is at least 6 months away, you can't expect games to require it for at least another 18 months, or probably closer to 2 years from now. So for now, yes, Directx 9 really doesn't matter.

About AGP 8x. That is virtually a cut and dried case. Each AGP improvement has had absolutely no effect on performance of first cards to carry that standard. Even later cards to carry each AGP standard have only ever been able to extract a small performance improvement (AGP 1x to AGP 4x on the latest cards seem to only improve the card by 20%) Since AGP 8x is only a bus speed increase like all other AGP "improvements" so far, why would it break the mold created by previous "improvements".

[rugger]

Originally posted by ralf
*sigh*
..,and I thought I was quite obvious.
I have nothing against serial ATA or dx9, on the contrary. I was just trying to make a point of what can follow by extrapolating the current state of affairs as a method of determining the future, hence the citation marks.
You don´t have to argue dx9 and whateverATA with me. I´ll take´m as they come. See?

As for AGP 8x you´re really avoiding the issue by using just the afformentioned extrapolation.

No, your extrapolations over dx9 and SerialATA were based off the extrapolation of AGP's past and future, not off their own futures.

AGP 8x on paper just isn't exiting enough to bother about.

[Wombat]

Not to mention that IMHO the <I>least</I> important offering of Serial ATA is the speed advantage over ATA/1xx. It's everything else about the standard that has potential.

[Mehen]

the inquirer has pretty bad support for this claim that its only agp4x anyways
"Matrox said that one of the reasons why its chip is AGP 4X part is because it is shipping Parhelia before broad AGP 8X introduction and this is something that we can agree, since Intel hasn't launched its AGP 8X hardware and the brood of Chipzilla chipsets is mighty indeed."
i thought it seemed stupid that they would be ready to "ship" the parhelia already, considering there has been no way to order it in the first place, and therefore they wouldnt know if, when, or where to ship it, ... they might not ever be done the card!! i think its pretty presumptious for the inquirer to say that they were about to ship it and then went "duhh, theres no agp8x boards out now..." i really dont think matrox is that stupid... so i wouldnt believe that claim that its only going to be agp4x

[Pace]

Sure AGP 8x isn't an issue RIGHT NOW, but by the end of the year when motherboards supporting it roll out...

I'd say AGP 4x still isn't an issue. And even if my MAX did AGP 4x, the card is still pretty much limited in itself anyway.

"talking about" being the operative words here.

But we're talking about AGP 8x here...so, by the time we get AGP 8x Parhelia's, we'll have threads wanting Matrox to upgrade it to NGIO...

And also, Mr Wombat, what do you like about sATA then? Just a curio, as I might as well learn something while my study effort is going crap. Man, Uni is boring...

[ChronosZero]

I personally think the most immediately useful thing about serial ATA is the increased cable length, which if memory serves jumps from .5m to 1m (maybe 1.5m... can't remember).

[DGhost]

problem with serial ata is that you can only have one drive per port on the motherboard right now. it does not support any sort of hub/switch so you cannot daisy chain any hard drives. if you have two serial ata ports, you can have two drives. Serial ATA 2 is gonna fix that.

as far as the speed benefit, with only one drive per port, its pointless except with burst transfers. point me to an ATA drive that can do more than 50MB/sec sustained and i'll gladly eat my words on this

the lower wire count in the cable as well as the increased lengths are good, but not enough of a benefit to switch over immediately.


About the AGP 4x/8x stuff - Intel is sampling the AGP 8x chipset - there is no reason for them not to be able to release a board that can do AGP 8x. like it has been pointed out SiS is shipping both chipsets and cards that support it, and Via is already shipping chipsets. the company that isn't is Intel - but they are far enough along it shouldn't be an issue.

personally, i would rather have a card that can do both AGP 2.0 and AGP 3.0. the major reason is that if chipset manufacturers stop supporting older AGP specs (like intel did with the 850 - couldn't use a Voodoo 5 in them), i would still like to be able to use the card.

will AGP 8x make a difference? as games start using higher resolution and higher quality textures and start using more polys in a scene, it probably will start to make a difference. plus, if you have a 64mb card it would make a difference if you are texturing from system memory.