Will Multi-VPU require an ATI motherboard chipset?

[trinibwoy]

Thoughts? There is a non-SLI DFI dual-16x Nforce4 board and it's mighty attractive. Will Multi-VPU be possible with non-ATI chipset mobos?

 

[Geo]

There is a non-SLI DFI dual-16x Nforce4 board and it's mighty attractive.

What is the purpose for the existance of such a board? It looks like an Oxymoron to me on its face, so obviously I'm missing something.

 

[trinibwoy]

There is a simple mod to activate SLI on Nforce4 Ultra chipsets. Several motherboard manufacturers took advantage of this since the Ultra chipsets are cheaper. Nvidia has since taken steps with its newer drivers to prevent this but the fact remains that a high-quality overclocker's dual-PEG board exists out there for less than $140. So if it works with MVPU it would be even more attractive, hence the question.....

 

[Tim]

There is a simple mod to activate SLI on Nforce4 Ultra chipsets. Several motherboard manufacturers took advantage of this since the Ultra chipsets are cheaper. Nvidia has since taken steps with its newer drivers to prevent this but the fact remains that a high-quality overclocker's dual-PEG board exists out there for less than $140. So if it works with MVPU it would be even more attractive, hence the question.....

But it is not dual 16x it is one 16x PCIx + 4x PCIe (both with a 16x slot)

 

[trinibwoy]

But it is not dual 16x it is one 16x PCIx + 4x PCIe (both with a 16x slot)

There was an article at Anand that dealt with the issue that actually said it ran at 16x + 2x but the performance in SLI was close enough to true 8x + 8x operation. Might be different for MVPU though since it's supposed to tax the PCIe bus more for inter-card communication. Thanks for the info.

 

[IgnorancePersonified]

dunno really but if this Inquirer article has any semblance of truth then it sounds like a more flexible architecture.

I'm sure since you posed the question here there will be an inq article on it soon....

 

[neliz]

I'm sure since you posed the question here there will be an inq article on it soon....

There are entities here (and at rage3d) that notify theinq of some "insider" stuff.. but then again, l'inq's latest post about AMR seems to be an extract of the general concensus at r3d..

they must think "ah, MB's agree : quoted for truth"

 

[Demirug]

Maybe Fuad is fallen in love with the supertilling procedure too much.

The regkeys in the new driver leave me to doubts if ATI will use supertilling as the solution for PC multi GPU.

 

[Geo]

Maybe Fuad is fallen in love with the supertilling procedure too much.

The regkeys in the new driver leave me to doubts if ATI will use supertilling as the solution for PC multi GPU.

More, please. . . what are you seeing (or not seeing) in those regkeys that gives you doubt?

 

[Demirug]

Maybe Fuad is fallen in love with the supertilling procedure too much.

The regkeys in the new driver leave me to doubts if ATI will use supertilling as the solution for PC multi GPU.

More, please. . . what are you seeing (or not seeing) in those regkeys that gives you doubt?

There are two keys called "MVPUSplitRTarget_X" and "MVPUSplitRTarget_Y". The X Value have a default value of 50, Y is 0(please don't ask). There are some caclulations that let me believe that this two values control a split line between two chips as a per cent value.

 

[digitalwanderer]

There are two keys called "MVPUSplitRTarget_X" and "MVPUSplitRTarget_Y". The X Value have a default value of 50, Y is 0(please don't ask). There are some caclulations that let me believe that this two values control a split line between two chips as a per cent value.

Mebbe those are the grid coordinate values for tiling and the driver currently breaks the screen into 4 tiles?

Sort of a longshot, but a thought none the less.

 

[PatrickL]

It seems a new chipset is needed according to digitimes.

In addition, ATI will introduce two high-end chipsets supporting its dual graphics card Multi-Rendering platform to compete with Nvidia’s nForce4 SLI (scalable link interface) platform, motherboard makers indicated. The RD400 for the Intel platform will be available by the third quarter of this year, said the makers, adding that ATI will launch its RD580 for the AMD platform soon after that.

From http://www.digitimes.com/mobos/a20050418A6032.html

 

[Geo]

It seems a new chipset is needed according to digitimes.

In addition, ATI will introduce two high-end chipsets supporting its dual graphics card Multi-Rendering platform to compete with Nvidia’s nForce4 SLI (scalable link interface) platform, motherboard makers indicated. The RD400 for the Intel platform will be available by the third quarter of this year, said the makers, adding that ATI will launch its RD580 for the AMD platform soon after that.

From http://www.digitimes.com/mobos/a20050418A6032.html

I didn't necessarily get that out of that article. That ATI needs a new chipset to support dual graphics card, yes; that there won't be interoperability between the two solutions, unclear (from this article, at least).

Having said that, I wouldn't be shocked if the first rev of the drivers only support an all-ATI combination in both cards/mobo, particularly if R520 has the whip hand performance-wise this time around (say thru the end of the year). Then a little chin-music on how of course they did verification with their own solution first, but that as their engineers complete verification more vendors solutions will be supported. That way they spin it as protecting customers and end-users by not supporting a combo solution until they are sure it works flawlessly (or some such kind of language).

Then timing to follow based on market factors.

Call me moderately cynical, but if (note big if) R520 is a clear-cut performance winner I can see where they might like to leverage that initially for a little while to see if they can move some of the gotta-have-the-best-at-any-price folks to their mobos. How quickly that verification of mixed combos follows could depend on how loud the screams are from users and how much pressure they get from the system builders.

Tho Dave is hinting otherwise, and I hope his analysis is the one that prevails. I'm just leery of initial revs of small-niche solutions for a few reasons. There are often legitimate technical challenges, smaller resources to deal with them, R&D expenses to recoup, and business-case guys with short-term vision lobbying mightily for proprietary synergy. EDIT: And one way you can straddle all that is to be short-term propietary to cherry-pick the cash-flush early-adopters while pointing only moderately disingenuously at the technical/resource challenges and promising that the support is inherent in the solution and will be turned on as soon as you can responsibly do so for the greater good, etc.

 

[neliz]

Hardocp has the RD480 pictured

So,... I guess they're benchmarking it *RIGHT NOW*

http://www.hardocp.com/image.html?image=MTExMzg1MTE5NjBXVjFGenE2NmFfMV8xX2wuanBn

Judging by the size of the chipset cooling, we don't have to expect huge fan setups for the R520 either...
This line is a dead giveaway

I would expect to see many motherboard builders launch these in late May when the Computex show in Taiwan kicks off.

 

[Geo]

Hardocp has the RD480 pictured

So,... I guess they're benchmarking it *RIGHT NOW*

http://www.hardocp.com/image.html?image=MTExMzg1MTE5NjBXVjFGenE2NmFfMV8xX2wuanBn

Judging by the size of the chipset cooling, we don't have to expect huge fan setups for the R520 either...
This line is a dead giveaway

I would expect to see many motherboard builders launch these in late May when the Computex show in Taiwan kicks off.

Well! The new Leader in the Clubhouse for the release date derby is once again the Old Leader in the Clubhouse. Who brought up that May 10 date? Let's lynch him!

They wouldn't announce them separately, would they? (i.e. R520 and the mobos to support it in MVP).

 

[neliz]

Well! The new Leader in the Clubhouse for the release date derby is once again the Old Leader in the Clubhouse. Who brought up that May 10 date? Let's lynch him!

They wouldn't announce them separately, would they? (i.e. R520 and the mobos to support it in MVP).

I think that the plan was to introduce the R520 first and the Mobo's later, but that was a plan from last januari, no need to hold onto that anymore.

Then I think it's time to lynch vr-zone.. HOcp, in my opinion, is quite reliable...

 

[Geo]

If you have heard any rumors surrounding ATI's upcoming "RD480" motherboard chipset, you will know that it supports dual PCI-Express video cards from ATI.

That's a rather ominous choice of wording.

 

[neliz]

Yep, so at least we know it's the end of may, and every day before that is only a bonus

 

[Jawed]

If you have heard any rumors surrounding ATI's upcoming "RD480" motherboard chipset, you will know that it supports dual PCI-Express video cards from ATI.

That's a rather ominous choice of wording.

Oh, come on, you don't seriously expect NVidia SLI'd boards to work in a non SLI-chipset mobo, do you?!!!!

Jawed

 

[neliz]

If you have heard any rumors surrounding ATI's upcoming "RD480" motherboard chipset, you will know that it supports dual PCI-Express video cards from ATI.

That's a rather ominous choice of wording.

Oh, come on, you don't seriously expect NVidia SLI'd boards to work in a non SLI-chipset mobo, do you?!!!!

Jawed

I think we do... just like the Nforce4 mobo's with "sli unlocker" bios upgrades...

Sure, I've bought an NForce motherboard, but if NV doesn't need my money, they're not getting it... and I think a lot of people in the industry feel the same..