Matrox confirms Parhelia only AGP 4X

[NFS4]

Fixed link, sorry about that

 

[Pocatello]

It seem as if the Matrox PR machine and the engineer people have a little mix-up. AGP 8X must be twice as fast as 4x, the PR people thought, so that's what we'll have.

 

[Electric Amish]

Originally posted by: Pocatello
It seem as if the Matrox PR machine and the engineer people have a little mix-up. AGP 8X must be twice as fast as 4x, the PR people thought, so that's what we'll have.

Pretty much sums it up.

amish

 

[NFS4]

Originally posted by: bdog231
AGP 8x is pretty much pointless right now. You may not know it, but your shinny new GF4 is only running on a 2x AGP bus, and so is that GF3 and 8500. Last I heard they disabled the 4x feature in their drivers due to stability problems.

Where did you hear that at????:Q My GF4 runs at AGP4x with sideband enabled

 

[Pocatello]

I think it's all depend on when intel is going offer an AGP 8x support chipset, and Matrox will produce an AGP 8x card just in case people don't want to be left out.

 

[NFS4]

Doesn't SiS have their AGP8x card out already?? The Xabre whatevertheflip it's called Anyway, they say it will work on upcoming AND and Intel AGP8x boards. So if SiS can do it, it can't be too hard

 

[grant2]

NSF4 does your posting of related articles mean you admit that Perhelia is no longer "vapourware"?

 

[Pocatello]

With all the people pimping the Parhelia like it's the second coming of Christ and that it will bake the perfect french bread, how can you have any doubt?

 

[aka1nas]

AFAIK, whatever AGP speed you run your board at has nothing to do with voltage, thats the AGP spec, not the AGP speed. It's the different AGP specs that have changed voltages on us. As long as the AGP 3.0(AGP8x) spec doesnt change the voltage and they do not change the connector, then all older cards running at said voltage should work fine. The 8x label is kind of a joke, because no matter how fast they crank that AGP bus, if you have to access main memory for your rendering, you will take a large performance hit.

Does anyone know how AGP Pro fits into all this? Is this spec still current and are any cards being developed for it?

 

[randomlinh]

they say they have it.. but i want to see production samples No really, i do.. i need compitition to drive prices down for my doom3 upgrade

 

[Leon]

AGP will become increasingly more important, as we approaching current AGP limits when trasfering verticles across AGP bus. With current standard, we max out @ 20-30mpolys at best (depends on different factors). There are workarounds (vertex_array_range type of low level extensions in OpenGL), but they are not sufficient.

Whoever claims that AGP8x or AGP in general is BS should go to any OpenGL developer board, and ask them. You will get the same answer.

Leon

 

[Electric Amish]

Originally posted by: Leon
AGP will become increasingly more important, as we approaching current AGP limits when trasfering verticles across AGP bus. With current standard, we max out @ 20-30mpolys at best (depends on different factors). There are workarounds (vertex_array_range type of low level extensions in OpenGL), but they are not sufficient.

Whoever claims that AGP8x or AGP in general is BS should go to any OpenGL developer board, and ask them. You will get the same answer.

Leon

For gaming we are nowhere near approaching the limits. I play all my games at AGP 1x and I have no problems.

amish

 

[Leon]

I play all my games at AGP 1x and I have no problems.

I can play games on PCI Voodoo5 and will not have any problems either. What is this have to do with what I've said above? Did I say AGP slower than 8x will cause gaming "problems"? No, I did not.

Obviously, you lack basic understanding of 3D technology, and go for what Matrox PR feeds you, so I will not bother explaining further. Be happy with what you have, but learn to accept the facts.

Leon

 

[Electric Amish]

Originally posted by: Leon

I play all my games at AGP 1x and I have no problems.

I can play games on PCI Voodoo5 and will not have any problems either. What is this have to do with what I've said above? Did I say AGP slower than 8x will cause gaming "problems"? No, I did not.

Obviously, you lack basic understanding of 3D technology, and go for what Matrox PR feeds you, so I will not bother explaining further. Be happy with what you have, but learn to accept the facts.

Leon

Wow, are you psychic? You don't even know me, but from 1 sentence you have deduced that I know nothing about 3D Technology. Impressive.

I don't know much, but I do know that the difference between 1x and 4x is minimal in gaming.

amish

 

[BFG10K]

AGP 8x is pure BS... just like 4x. There's hardly anything that even pushes 2x.

False. Don't be one of those "anything my card doesn't have is BS" zealots. AGP x4 is close to being saturated right now thanks to vertex data from exploding polygon counts and texture swaps from larger and more textures and the next generation of games like Doom3 and Unreal2 will further stress AGP.

In addition to have twice the theoretical bandwidth, AGP x8 has been optimised for streaming transactions for this very reason. You might think it's useless but the likes of Intel and nVidia know it's not and I'm guessing that they know just a wee bit more about 3D technology than you do.

I don't know much, but I do know that the difference between 1x and 4x is minimal in gaming.

Not if you look in the right places and run the right kinds of benchmarks.

 

[Lord Evermore]

You might think it's useless but the likes of Intel and nVidia know it's not

Or they might just know how to make you think they know what's really needed because it's in their best interests to make you think you absolutely must have whatever newest tech they come up with.

 

[Sunner]

Can't say I care much about AGP.

Ever since it's introduction we've been fed with loads and loads of reasons why it will be more important in the near future.

Then came AGPx2, and there was no difference between AGPx1 and x2.
And finally, AGPx4 came, and still there was little to no difference, even between x1 and x4.

So, please excuse me if I don't care a terrible lot this time around.

 

[Pariah]

"You might think it's useless but the likes of Intel and nVidia know it's not and I'm guessing that they know just a wee bit more about 3D technology than you do."

Wasn't Intel the one that told me I needed a P4 to raise my web surfing experience to a new level?

"Does anyone know how AGP Pro fits into all this? Is this spec still current and are any cards being developed for it?"

AGP Pro is not a performance enhancing standard, all it adds is addition power lines increasing the power it can utilize from the standard 25W up to 50W. It's used for highend rendering boards and the like.

 

[Chumster]

From a recent Gamespy article on Unreal 2:

Unreal II will reportedly push 100X the number of polygons as Unreal Tournament. As we were told, just one tree in Unreal 2 could have more polygons than an entire SCENE in UT.

You can read the entire article here.

While it may not be needed now, it would seem (at least to this layman), that the next gen games could very much utilize the additional bandwidth.

Chum


[edit]formatting stuff[/edit]

 

[NFS4]

Originally posted by: BFG10K

AGP 8x is pure BS... just like 4x. There's hardly anything that even pushes 2x.

False. Don't be one of those "anything my card doesn't have is BS" zealots. AGP x4 is close to being saturated right now thanks to vertex data from exploding polygon counts and texture swaps from larger and more textures and the next generation of games like Doom3 and Unreal2 will further stress AGP.

In addition to have twice the theoretical bandwidth, AGP x8 has been optimised for streaming transactions for this very reason. You might think it's useless but the likes of Intel and nVidia know it's not and I'm guessing that they know just a wee bit more about 3D technology than you do.

I don't know much, but I do know that the difference between 1x and 4x is minimal in gaming.

Not if you look in the right places and run the right kinds of benchmarks.

I think that pretty my sums it up right there (the part I put in bold)

 

[Lord Evermore]

Show some numbers proving that in real world use the AGP4X bus is "saturated" with anything. Just saying that a game can theoretically pump out enough to do it isn't enough, nor is saying that the "right" benchmarks can saturate the bus.

 

[BFG10K]

Wasn't Intel the one that told me I needed a P4 to raise my web surfing experience to a new level?

Some of Intel's SSE2 instructions have been designed to speed up the rendering of web pages and multimedia. So yes, they were very much correct.

Show some numbers proving that in real world use the AGP4X bus is "saturated" with anything.

Obviously you can't show that AGP x4 is saturated until you have something faster to compare it to.

Nevertheless, you can show that there's a difference between AGP x1/x2/x4 even with today's games. Try some RTCW, JKII or Serious Sam benchmarks (all games that heavily utilise T&L) in CPU limited situations on a fast system and you'll see easily a difference between the different AGP speeds. The ideal test of course is the Unreal2 technology test as that'll give you a glimpse into the future.

 

[Sunner]

Some of Intel's SSE2 instructions have been designed to speed up the rendering of web pages and multimedia. So yes, they were very much correct.

Oh, phuleeeease....

 

[ElFenix]

what is saturating AGP 4x?! my athlon's memory has a whole 2.1GBps available to it at any time... processor + AGP + everything else in the comp can not exceed 2.1GBps. its simply not possible. if AGP 4x is sucking down its 1GBps then my processor and PCI devices are getting 1 GBps between them. and AGP 8x is worse. theres no way any PC2100 DDR athlon without some kind of dual channel could even think about saturating AGP 8x. if all this vertex data takes up so much space then it takes up that much space when its sent over the FSB as well. the processor handles all sorts of other things as well, AI, physics, windows background tasks... unless you're swapping textures over AGP all the time the processor is getting more memory bandwidth than the AGP bus. maybe an rdram P4 board or a dual channel DDR setup has enough left over after the processor gets it share to make use of something more than AGP 4x.

 

[ElDonAntonio]

Some of Intel's SSE2 instructions have been designed to speed up the rendering of web pages and multimedia. So yes, they were very much correct.

AHAHAHAHHAHAHAHAHAHAHAHA that was one of the funniest thing I read in these forums since a long time...

1) I browse the web on a "crappy" laptop Celeron 800 and a broadband connection and it's running without a hitch. You can throw whatever humongous Flash animations at my system, it'll take it without any problems. Don't feed my any P4 bullshit about a "new web experience". Only a total noobie would believe that.
2) SSE2 has not "been designed to speed up the rendering of web pages" at all. That sentence shows a huge level of ignorance towards assembly language and optimization. SSE2 is just a extension to the set of CPU instructions that adds better integer SIMD capabilities than MMX offers and over 128 bits rather than 64. Maybe you mean that IE has been designed to take advantage of SSE2 (not the other way around), but I really don't think IE can benefit from SIMD, and never heard it was optimized.

As for the 4X/8X debate, I'd also like to see some clear demonstrations that 4X can be saturated, and that 8X can provide better performance given the FSB limit as Fenix pointed out.