ant, is this the official release?

[Opus]

Jeeze is this it? After all this time it's still going to take them till the third quarter to produce a card that has only a slight speed improvement over the Max? I take it will be like the difference between the G400 and the Max rather than that between the G200 and G400.

I'm a bit like Cmag in that I used to post more but for some time now I only occasionally check back on the site to see whats up. It seems Matrox is falling behind the other card manufacturers (though they are still No.1 for image quality). Are they losing the plot? Couldn't they get some AMD guys in

God, that meams I'm gonna have to stick with my trusty G200 for a while longer and there's no way I'm paying over £200 for a GeForce2 card...

Cheers

[Kruzin]

CMAG, et al,

While I can't say anything about G800 (Matrox has never even announced such a card will ever exist), I will say I am greatly looking forward to their next card after G450. Rather than saying "they are falling a gen behind", I am looking at it from the POV of "in 6-8 months, they are gonna be a 1/2 to a full gen AHEAD".

I don't think we'll be disappointed in what Matrox has coming down the line....

[Dyre]

I hope your right Kruzin, i really do.

[Mark F]

If that's 1/2-1 gen ahead of the then generation, it sounds great. If it's 1/2-1 gen ahead of current generation then it'll be compeating with it's pears(at best). I hope it's the first!

So bring it on

Mark F.

------------------
OH NO, my retractable cup holder swallowed a DVD...
and burped out a movie

[Himself]

I think it will be "compeating" with it's oranges, not it's pears.

"You can really appreciate a good pear, because a bad pear is so very bad."

I think they had better aim for 4 times the specs of the G400 if they want to get anywhere in the gamer market, at least. 6-8 months is an eternity, they really need to get on the ball, no wine before it's time won't cut it. If they don't have T&L, they better have FSAA or tiled rendering. Pixel shaders would help but it's more of a spec sheet tick than anything. (I seriously doubt anybody other than ST Micro and 3DFX will come out with a working tiled architecture soon, and only 3DFX because they bought GigaPixel).

[Hugh G Recshun]

If I remember reading correctly, about a year ago Matrox purchased some licensing deals that had to do with tiled rendering. I believe that this may be what Matrox is up to. I think they are looking very seriously at tiled rendering and have been for some time. I remember an interview somewhere that had a Matrox PR guy talking about what the next gen. aims were, and he specifically mentioned overcoming the current memory bandwidth problem, and the FSAA performance hit. Using tiled rendering properly, you can reduce both of these problems using current technology, and I truly think this is where they are headed. Just think, tiled rendering, matrox graphics quality, high speed mem, t&l (as useless as it is currently), embm, and dual head. Sounds pretty yummy, doesn't it?

Hugh G

[Homercles]

Just a quick note: Q3 starts in 2 weeks

Q4 starts in around 100 days

DX8 is still in early beta.. don't get your panties in a knot yet

Homercles

[superfly]

well,whatever the G800 turns out to be(performance wise)it will be a very expensive card costing upwards of at least 550 canadian,how do i know this,it's simple really,a couple of days ago i finally cracked under the presure and bought a hercules 64mb GF2,and needless to say it's a very impressive card(1600*1200 q3 at 80 fps anyone),but what shocked me most was it's price...630$ Canadian(before taxes of course),other wise it comes out to 730$...i honestly believe that a G800 with 64 MEG ram will also be priced around that level.because in order for the card to have a decent lifespan it will have to be equipped with at least that much memory,so that you'll be able to play games at high resolutions and with much higher quality textures,the chip itself should be at least as powerful as the GF2 is(if not more so),i don't even dare think how much would a potential dual chip G800 cost,especially considering that each chip has to have it's own memory bus.

[Galvin]

Well if the G800 uses tiled rendering it won't need as much ram, since tiled rendering uses less ram to do the same job.

[superfly]

i agree that tiling does have it's advantages,just see the kyro(powerVR 3 chip)card in action for proof,especially in terms of bandwith needs and has no overdraw problems,but it's desing is inherently complex and it's alredy on it's third generation,the first 2 coudn't keep up with the other cards on the market at the time,i'd like to believe that matrox has the ability to build one,but the question here is,can matrox build one that's at least as fast(if not faster)than anything available on the market when it get's released(Q4/2000 to Q1/2001),besides the current rumor has it that the G800 will use a type of memory that's even faster than DDR,why would matrox want that memory if they will be using an HSR(hidden surface removal)approach.to make matters worse if indeed matrox releases the G800 around xmas,it will be going up against nvidias nv20,these are very tough times indeed as far as consumer level gaming cards go.

[The Rock]

All this talk about the G800 is causing rumbling in my nether-regions. That stinks because after buying MDK2, I can no longer defend my poor ole G400. This is the first title that has indicated that I need a new video card...and soon. The next wave of games are gonna choke the life outta the best video card I've ever owned.

The Rock
www.3dforce.com

[Himself]

On the one hand, the DDR and faster/different references kind of suggest tiling isn't going to be a solution any time soon, not at 128 bit bus anyway. The main profit advantage of tiling is using cheaper and slower ram with equivalent results. Of course, on the other hand, they could keep the 64 bit DDR ram option to keep costs down even further and certainly supporting the different ram types is just smart to keep options open.

[superfly]

well...it pretty much boils down to those two options really,a conventional approach with very fast memory and preferably a 256 bit wide bus help the chip reach it's potential performance...the advantage of this aproach is that it's how matrox and the other makers have always built their chips,and there are no potential patents being infringed upon(unlike tiling)the disadvantage is that it'll be expensive...the other aproach is of course use tiling or any other hidden surface removal tecnique in such a way that no patents from other companies(powerVR/nec or 3dfx/gigapixel)are being infringed upon,and also deliver real world performance equivalent to anything available from the other makers,but that's something that matrox has never tried before,but it would certainly be much less expensive to build.