NVidia's Dispute over Nehalem Licensing - SLI not involved

[Davros]

just to clear up, it requires 256 megs of video memory (as in it uses up) and not it will only work with an igp with access to 256mb of system memory

if so why would it be the case ? if the igp was being used i would only be using 2d (like web browsing ect) so wouldnt i just need X*Y * 32 no ?

 

[ChrisRay]

Correct. If you set the IGP to 128 megs or 64 megs. It will NOT work. I never got a technical explanation to why. The IGP can be set to 64, 128, 256, and 512, But will only work if you have dedicated at least 256 megs of system memory.

Which is the main downside I see to it. Basically it means if your looking to build a fast 2 gig system its not really practical. For 4 gig systems its not quite as big of a deal.

Chris

 

[Davros]

this is from the duplicate thread posted by shtal
http://forum.beyond3d.com/showthread.php?t=48413

Quote from Nvidia's Director or PR, Derek Perez
"Nvidia’s bold decision to keep SLI for itself."

"Nvidia has confirmed that it is officially becoming a victim of what VIA, AMD and many other companies have felt for a long time, as Intel tends to repress competition when threatened by superior products and technologies."

That Derek Perez is one two faced bar steward intel is just doing to nvidia what nvidia is doing to intel
isnt intel's decision to keep Nehalem a "bold decision" ?
intel are just deciding to not license their tech they are not deliberately blocking competitors ala sli

and why do nvidia need a license anyway thats like needing a license to make refils for epson printers ?
surely the point of the pc is its open didnt licenseing chipsets die with ibm's mca bus ?

 

[BlackAngus]

didnt licenseing chipsets die with ibm's mca bus ?

Ive got a question over what this is all about anyway.
So I would imaging that Nvidia can make a piece of plastic that has the correct pin outs, and also that they could get a Nehalem and reverse engineer the layout to some degree or an other. So Im wondering what technologies are not covered by the current agreements. Knowing NV's PR can be quite .... head spinning.... Im not sure we are hearing the whole truth.

I would guess that either CSI is the source of the licensing issue or that the IMC layout and protocols are.
Anyone have any idea as to what the exact issue is?

I believe that NV would have to use the CSI protocol for Nehalem to function correctly, and that is what is not covered under the licensing agreement. The license is likely the FSB technology license from intel that NVidia has, and Intel believes that the current agreement does not cover CSI. (Or whatever CSI is called now.)

Just a guess, has anyone heard any more details?

 

[Jawed]

Ive got a question over what this is all about anyway.
So I would imaging that Nvidia can make a piece of plastic that has the correct pin outs, and also that they could get a Nehalem and reverse engineer the layout to some degree or an other. So Im wondering what technologies are not covered by the current agreements.
Knowing NV's PR can be quite .... head spinning.... I would guess that either CSI is the source of the licensing issue or that the IMC layout and protocols are.

Memory should be irrelevant if it's directly connected to the CPU socket.

So that just leaves QPI I reckon, without which it is presumably impossible to connect a southbridge, or to interface with PCI Express.

Jawed

 

[Davros]

(Or whatever CSI is called now.)

Just a guess, has anyone heard any more details?

I beleive the latest one is called CSI - New York

 

[Blazkowicz]

Well, at least triple buffering isn't quite as grave of an offense as disabling vsync...

disabling Vsync is more than okay on a >= 85Hz CRT. 60Hz really makes Vsync off unbearable, maybe it would look good on a 100Hz LCD. (waiting for a 100 or 120Hz - real Hertz, LED back-lit, fast xVA panel with no input lag, under the 200 euros mark for a 24" please . will still be using the CRT till then, 'cos I like it)

 

[Shtal]

NVIDIA skips LGA 1366 Nehalem plaform, supports SLI by chipset

According to news reports from the web, NVIDIA has acquired the rights to use the Intel QPI technology. The interesting development is that NVIDIA has no interest in developing a new chipset to support the LGA1366 Nehalem chipset. Instead, it will supply it's nForce 200 bridge chip (BR04) to have SLI support on the Intel X58 platforms.

As we already know, SLI technology is nothing other than disabling support using software means. NVIDIA, by enabling SLI support with a so called hardware chip is just going to add more costs to manufacturers who intend to develope Intel X58 boards that supports both CrossFire and SLI.

With NVIDIA reluctance to develope a chipset for the LGA1366 and the mid range LGA1160, it looks like we will only be seeing INTEL for INTEL, AMD for AMD real soon.

It seems that NVIDIA will soon be exiting the chipset business with no more new NV chipsets support both Intel and AMD platforms.

http://my.ocworkbench.com/bbs/showthread.php?p=433716#post433716

Looks like war between intel vs. nvidia is finally over!!!!!

Intel won!!!!!!

 

[Arun]

Those guys seem quite ignorant. Anyway, I've said this times and again: the right strategy for both LGA1366 and LGA1066 is a BR05 chip which is capable of: "IN: 16xPCIe Gen2, OUT: 3x16xPCIe Gen2" with an on-chip IGP-level GPU for HybridPower. In addition to that, you can add a gigabit MAC and RAID5 acceleration or whatever to further differentiate with motherboards without that bridge.

There are two complexities with this approach: first, it is theoretically possible for Intel to block that. Second, it's only good enough for the high-end, and you miss out on the mid-range and low-end completely. In a way the former would be good, because it'd be a very effective and easy to prove arguement to highlight monopolistic behaviour; and honestly, if Obama is elected president and say, Edwards, becomes Attorney General, you want as many as those as you possibly can get because it'll be by far your best weapon against Intel as the definition of 'monopoly' gradually changes.

There were rumours of a socket 715 for third party chipsets, but I don't know what NVIDIA and Intel's plans are there; it could be Intel decided to just remove that socket or perhaps it never even existed. Certainly they could still grab a substantial part of the Intel market with bridge chips for Socket 1160/1366 and single-chips for Socket 715; in fact, that would be the vast majority of the addressable market! (i.e. S1366 will be low in terms of volume and S1160 will just be small southbridge chips; not much differentiation possible or desirable there, and the ASPs are unlikely to be gradiose especially if there's any competition).

Also remember that it might simply be too late for NVIDIA to create a S1366 chipset until, I don't know, Q2 2009. That's a fair bit of marketing momentum and sales lost already, making it much harder to amortize a completely new chipset...

 

[Broken Hope]

Why is the bridge chip even needed? I thought the only thing stopping SLI working on an Intel chipset is Nvidia blocking it in software? Since I'm pretty sure there used to be hacked drivers that allowed it.

 

[Arun]

Why is the bridge chip even needed? I thought the only thing stopping SLI working on an Intel chipset is Nvidia blocking it in software? Since I'm pretty sure there used to be hacked drivers that allowed it.

It's not needed, but NVIDIA needs to figure out a way to make money while still adding some value so that it's semi-justifiable. A BR04 does add value in the Socket 1160 market (instead of 2x8 PCIe, you get 2x16 PCIe; this is equivalent to the difference between the nForce 750a and the 780a, and the price gap can be pretty big amusingly enough!) - although it pretty much doesn't in the Socket 1366 market (except 2xBR04 for Tri-SLI, I guess) and you'd need something like my speculative BR05 to do the trick.

 

[INKster]

It's not needed, but NVIDIA needs to figure out a way to make money while still adding some value so that it's semi-justifiable. A BR04 does add value in the Socket 1160 market (instead of 2x8 PCIe, you get 2x16 PCIe; this is equivalent to the difference between the nForce 750a and the 780a, and the price gap can be pretty big amusingly enough!) - although it pretty much doesn't in the Socket 1366 market (except 2xBR04 for Tri-SLI, I guess) and you'd need something like my speculative BR05 to do the trick.

So, in essence, a shrunk BR04 ("BR05") would have a business model similar to any other add-on chip for a motherboard out there (Ethernet, Firewire, HD Audio, etc) ?
It could work, but they need to seriously reduce the chip packaging footprint (yes, it's smaller than a PLX, but still too big for easy integration into a motherboard design) and overall power consumption for that to be anywhere feasible, IMHO.

 

[ChrisRay]

While Arun is not "Dead on" hes pretty close to being right on alot of his speculation regarding nvidias vested interested in the first Nehalem platform and costs of supporting the platform.

Chris

 

[Shtal]

Also remember that it might simply be too late for NVIDIA to create a S1366 chipset until, I don't know, Q2 2009. That's a fair bit of marketing momentum and sales lost already, making it much harder to amortize a completely new chipset...

Would you say Intel behind all of this trouble, just like upcoming USB-3.0 specification - Intel kept hidden as long it wanted.

 

[Arun]

So, in essence, a shrunk BR04 ("BR05") would have a business model similar to any other add-on chip for a motherboard out there (Ethernet, Firewire, HD Audio, etc) ?

Right, it's the same business model. My speculative BR05 isn't just a shrunk BR04 though...
BR04 - IN: 16xPCIe Gen2, OUT: 2x16xPCIe Gen2
BR05 - IN: 16xPCIe Gen2, OUT: 3x16xPCIe Gen2, with IGP & HybridPower

It could work, but they need to seriously reduce the chip packaging footprint (yes, it's smaller than a PLX, but still too big for easy integration into a motherboard design) and overall power consumption for that to be anywhere feasible, IMHO.

Well, I doubt power consumption would be lower given the additional on-chip GPU, but they do need to keep that in check, yes - ideally it needs to be done on 40nm, obviously.

Would you say Intel behind all of this trouble, just like upcoming USB-3.0 specification - Intel kept hidden as long it wanted.

Yes, definitely - in fact, it's actually a fair bit worse than what I'm implying here; there's a reason why NVIDIA didn't seriously consider the bridge chip strategy previously, for example, heh...

 

[INKster]

Yes, definitely - in fact, it's actually a fair bit worse than what I'm implying here; there's a reason why NVIDIA didn't seriously consider the bridge chip strategy previously, for example, heh...

Such as... ?

Do delays and/or cancellations have anything to do with it ?

 

[hoom]

Such as his signature

Anyway so I've been not keeping proper track of chipsets for a while, is ATI completely locked out of Intel chipsets too? (except for Intel chipsets being able to do Crossfire)

 

[bowman]

Such as his signature

Anyway so I've been not keeping proper track of chipsets for a while, is ATI completely locked out of Intel chipsets too? (except for Intel chipsets being able to do Crossfire)

ATI hasn't made an Intel chipset since they were bought by AMD, and I don't suspect they will again unless AMD sells them again..

 

[Blazkowicz]

I like that little DOS rig with an Intel chipset and an AMD K6 I made.

 

[trinibwoy]

http://biz.yahoo.com/prnews/080714/aqm063.html?.v=50

New SLI motherboards will feature the NVIDIA nForce® 200 SLI processor, Intel Bloomfield CPUs, and Tylersburg (X58) chipsets. The nForce 200 SLI processor features patented SLI technology for graphics bandwidth management and multi-GPU peer-to-peer communications, both required to optimize graphics performance.

"With GeForce-based visual computing application, our customers are experiencing exciting new ways to interact with their photos and videos while also experiencing the bone-chilling realism from our latest graphics processors," said Jeff Fisher, senior vice president of GPU business at NVIDIA. "Our SLI technology allows us to combine the power of hundreds of GeForce processing cores in multi-GPU configurations with Intel's latest CPUs for platforms that are sure to excite our customers."

"It's great to see that NVIDIA opted to enable SLI on the future Intel Bloomfield platform," said Rahul Sood, CTO Voodoo Business Unit, HP. "Make love not war I say ... and NVIDIA's enablement of Intel chipsets to support SLI will make our jobs much easier in offering a better customer experience as we continue to evolve HP's award-winning Blackbird 002 and Voodoo Omen platforms."