Parhelia slow?

[KingJackal]

http://www.rage3d.com/board/showthread.php?s=&threadid=33622002

Draw your own conclusions - these sure aren't final figures.

IMHO ;):

This was never meant as a gamers card, and Matrox have never made gamers cards ( post 3D - I don't really think even the MillenniumII/Mystique series were that 1337 - even at the time ). It looks like that's set to continue. Their card appears to be lining up for Ti4400-Ti4600 level performance. Oh yeah - except nVIDIA are drivers GODS, and will probably release new detanators so their Ti's can all out-pace it :rolleyes: :D.

However, having said that - I have no doubt that much of the technology being introduced with this card will most definately rock. Matrox DO have a history of releasing features loooonnnng before their time ( EMBM anyone? ).

Regardless, this will be a VERY interesting launch with the R300 now looking to be about the only top-end gamers choice until the NV30/P10 ( miles away ;) ).

[Humantuckshop]

I always thought the card would be aimed towards graphics designers and the like. As you stated KJ the Parhelia is obviously not a gamers card.

I'm sure the 2D quality will be awesome. :p

[Solid Snake]

GeForce 3's and 4's were never that hot on 2D filtration. With Matrox, I believe the filtration will be VERY good, only because this what they also focus on as well.

[KingJackal]

Originally posted by Solid Snake
GeForce 3's and 4's were never that hot on 2D filtration.

Leadtek's are :p :D

[Solid Snake]

I never owned one ... :) Well all said and done, I think my eyes are too bad to even notice these things and I stick with lowly resolutions compared to you KJ.

[mird-OC]

beta hardware, beta drivers, let's just wait for the real thing, shall we? ;)

and i'd agree that some Leadtek cards have better filters than your average nVidia card but don't go comparing them to a /\/\47|20>< :D

[DiscoStu]

Originally posted by mird-OC
beta hardware, beta drivers, let's just wait for the real thing, shall we? ;)
indeed - I remember a few people writing off the radeon 8500's a little too early as well.

[Maverick]

yes we have to remember too that nVidia have been using unified drivers that go back to the Geforce256 days (as far as the geforce architecture goes). Each hardware revision has required an update to the drivers, along with all those tweaks in between, that are now 3 years later very stable, fast, & highly optimised.

Matrox Parhelia is a completely new architecture (as will the P10, NV30 and maybe the R300 as well). With a completely new architecture, drivers start from scratch. It may take at least 6 months for "stable" and "fast" drivers from each graphics house to hit the streets. Thats why second revision hardware is always received better than first revision -the drivers as well as hardware has been bug-fixed and optimised sufficiently.

[KingJackal]

Very rarely do you 'bug fix' hardware. Generally if there's a hardware fault, you code AROUND it. Hardware modification is always expensive, and often impossible.

It is very encouraging to see such a low performance hit for 16X FSAA and anisotropic filtering enabled together ( though it would be nice to know whether that's 2, 4, 8, 16, 32 or 64 tap aniso ;) ), though FAA's inability to work in any ID game ( Doom3 included ) is an interesting problem.

I wonder if ID will work to make FAA usable in Doom3? I'm not so sure changing the shadowing method would be so easy in an engine that's mainly based on it's lighting and shading.

Hmmmm - so many interesting possibilities.... :)

[Ragnor]

I think I recall Carmack saying no to compatibility with FFA in favour of keeping the better looking / faster way of doing the shadowing... not 100% sure on that tho..

And.. on the Matrox.. it's highly likely it's just driver efficiency/optimization that is serverly cramp'n its style, since the hardware specs are impressive.

BTW.. check out the lastest review/article @ H]ardOCP (www.hardocp.com) which tests the new ATI Drivers (Catalyst) on the Radeon 8500. Damn the 8500 is keeping up with the gf4 Ti4200 in most benchies now...

Compare that to a few months ago when the 8500 couldn't beat a gf3 Ti200

[DiscoStu]

LOL - I was just about to post that ragnor ;)

[Volodkovich]

damn...i didnt think they were getting that close to a GF4 ti 4200. I spose if u look at there specs there not far off.

[CrazySurfaNZ]

Sweet thats good news for me considering i just bought a radeon 8500 :)

[Maverick]

Originally posted by KingJackal
Very rarely do you 'bug fix' hardware. Generally if there's a hardware fault, you code AROUND it. Hardware modification is always expensive, and often impossible.



LOL, yeah, like nVidia's S3TC bug. Did they ever bother fixing that for the Geforce4? Probably after they got it wrong on the GF256 all subsequent NV10 coding in games would have coded around it -resulting in more problems if they did fix it.

What I was referring to was how the second revision usually improves and optimises features introduced in the first revision of hardware. Geforce2 is basically a GF1 with some better shading capabilities (EMBM etc) and an improved T&L engine, that and the better design enabled higher clock speeds. GF1 was quite a low speed GPU (120MHz) that ran damn hot.
We saw this again with the Geforce4 having a much more heavily optimised memory controller and a further tweaked shader pipeline than the GF3.

In reference to Matrox Parhelia, P10, NV30, R300 -the first editions among them with completely new achitectures and new features will not be fully optimised in hardware till at least a second revision -eg: NV35. ParheliaII etc. By then of course drivers will have had sufficient coding time as well to develop.

Another thing of note, is that Matrox will not be sacrificing on IQ for speed. Unlike nVidia or ATi driver teams will do from time to time;)