So the G800 will "support" 8x eh?

[Rob M.]

Personally I'd rather have a stable 4x card than a piece of paper basically saying my card will work in a 8x socket. They should be focusing on a proper implementation for the current standard, especially considering how useless 8x is.

[Mark F]

I hope not.

Or we will not see it for a year and a half or two!

Remember the party line about 4X. Even though MATROX had a roll in the specs, early boards would not run in 4X motherboards at 4X; 'how could they make 4X video cards when there were no 4X motherboards'. Yes, that is only paraphrased but about what was said, whenit was found out they coudn't run and/or be upgraded to run at 4X; they needed a different hardware revision.

So therefore we would need to have 8X MBs available for the G800 to run 8X, maybe 1.5-2 YRS.

I don't want to wait that long for a G800!!

Mark F.

PS. they are part of the stearing committe, setting the spec.



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

[superfly]

As far as i know,the 8x agp spec is still under developement and we'll see motherboards equipped with only by this time next year.

There's absolutely no rush in having 8x now,since games aren't even taking advantage of agp 4x as it is.

the bandwith advantages that agp 8x boasts will only become usefull maybe in 2 1/2 or 3 years time.
And the cool part is that by then we'll have graphics card's 10 times more powerfull than anything available today.

[Snake-Eyes]

Uh, yeah.. right.

And by that time, since the graphics cards are that powerful, the 8x AGP bus will still be slower than using on board memory.



------------------
Ace

[superfly]

There will never be a bus protocol that will perform as well as local memory on videocards.

It usually takes the better part of two years to develop a motherboard chipset,which only gets developed once all the standards are agreed upon(agp spec,pci spec,ide spec,overall bus speeds,licensing issues,etc...).

the problem is that by the time the boards are released,there will always be faster bus speeds and memory available to video card makers,since they only have to worry about the agp slot specification and nothing else.

[xortam]

Hell with on-board memory ... go with on-chip memory.

[Jorden]

Btw, who said the G800 will support 8X AGP? All the news said was "This extra bandwidth will benefit the many innovative graphics technologies being developed for future Matrox products, as much for home as for business applications."

As the G800 news has leaked out already, it isn't really a future Matrox product anymore, so the next card we might expect it on is the G1600 then (in the least).

And much can change in a year, so next year we might all have a P4-1.5GHz CPU, 512Mb of RIMMs on a mobo that does 8X AGP, so we're ready for the new Matrox card, sporting at least anything we hype it to

Jord.

[superfly]

The agp 8x spec won't show up with the p4 as well,at least not while the p4 is built on a 0.18 micron process.

i'm guessing that the 8x agp won't show up at least until the p4 goes for it's 0.13 micron die shrink which,as far as i know,won't happen until this time next year and will require a new chipset as well.

the agreement that intel has with rambus forces them(intel)from making chipsets that support DDR sdram,that contract ends late next year i believe,so i'm guessing that agp 8x could show up when intel starts making their own DDR sdram chipsets themselves,without having to resort to licence the p4 bus to via and the other chipset makers like it will have to when the first p4 systems launch in a couple months.

This must really drive intel nuts right about now but i guess they want to sell p4's,and if it means losing a good part of the chipset market(at least temporarily)so be it.

Coming back on topic,are the rumored G800 specs then final???.

[Rob M.]

Umm, what does CPU die size have to do with the AGP implementation?

I could see maybe it could do with chipset die size, but as far as I can see, the cpu die size has absolutely nothing to do with it.

[Rob M.]

Besides, I'm 100% sure the G800 will support agp 8x, the same way that my G400 supports agp 4x like it says on my box...

[superfly]

It doesn't really,since the first p4 systems will run on intel's 850 chipset,which only supports rambus memory.

what usually happends is that when cpu's go for their die shrink(and hence lower voltage requirements)they usually introduce a new chipset with new features.

I'm sure that many have heard that,for instance,the p3 will go for yet another die shrink(0.13 micron)and have a faster bus to go along with it(200 mhz)that will require new chipset.

Most motherboard makers won't even bother with initial p4 support since the first versions will be built at 0.18 micron,and with a chip that boasts 47 million transistors,that makes the chip almost 2 1/2 times bigger than the p3(217 sq mm vs 100 sq mm),that makes p4 quite a bit more expensive to build.

And since intel will move to 0.13 micron for the p4 as fast as they can,probably by this time next year,it will also require less voltage(as usual),but the thing that pisses off mobo makers the most,is that it will use a new socket wich,by necessity will also use a new chipset with yet again new features(agp 8x???)

So what do mobo makers do,support the initial chipset(intel 850)that has maybe a 6 month life span and still uses expensive rdram and therfore won't probably sell in very high numbers.

They have the option to use alternative chipsets,especially since intel has licenced the p4 bus,but those new chipsets won't show up at the same time(probbly a few months later).

Since the agp spec is alredy at revision 0.9,which is close to the final spec,i'm guessing that agp 8x will make it's apperance no later than when the p4 goes for it's die shrink,again by this time next year.

And hopefully it will be by then that we'll see the first agp8x compliant video cards.

[Pace]

superfly: regarding the RAMBUS contract not allowing Intel to use DDR SDRAM would it not be possible to interleave SDR banks to double tha bandwidth? Just a thought - I know the i840 does this and I believe that it is with SDRAM - or is it only RDRAM what can do it? Anyone?

Paul.

[Mark F]

I know the i840 does this and I believe that it is with SDRAM

NO, the i840 doesn't use SDRAM. It 'could' with a MTH, so I guess that is the equivellent to not able to use it.

Not sure about this but I think some VIA 133a MoBo have a 'interleaved memory banks' option. Any one else that can confurm this?

Mark F.

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

[superfly]

The 840 does indeed interleave it's memory regarless of what memory is being used.

The 840 does require an MRS(memory repeater hub-sdram)to support sdram.

there is no point in having pc 133 sdram over pc 100 on an 840 board,since the sdram memory channels run at 100 mhz regarless.

the one big drawback of interleaving memory is that memory modules have to installed in pairs,just like in the old days with simms.

if the 840 supported sdram directly it would acheive pc 200 memory performance(1.6 gig/sec),because it has those 2 channels.

but since it needs those MRS chips,it suffers about 15/20% performance hit per channel,but even with that handicap,it still beats pc 133 memory performance.

If no one has figured it out yet,i have an 840 board with sdram(supermicro p3 DME),with dual cpu's,and the performance is outstanding.

[Ali]

There was a review of cheap intergrated chipsets at tomshardware the other week, that had interleafed dimms. I cant remember what it waas, but I was looking for cheap computers for work, but the ones reviewd arnt available yet.

Cant be bothered looking for the link, cos Tom pissed me off.

Ali