Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
Graphics Software Intel Hardware

Intel Licenses NVIDIA SLI Technology For P55 Chips 63

adeelarshad82 writes "NVIDIA announced that Intel has licensed the company's SLI technology for inclusion in upcoming products — as have a slew of major hardware partners such as ASUS, EVGA, Gigabyte, and MSI. This means the P55 chipsets that power those new socket LGA 1156 motherboards, which are based around the next-gen Nehalem architecture, will let you build systems using two or four NVIDIA-powered GPUs. Specifically, the licensing agreement covers the Core i5 and Core i7 microprocessors."
This discussion has been archived. No new comments can be posted.

Intel Licenses NVIDIA SLI Technology For P55 Chips

Comments Filter:
  • Monopoly? (Score:2, Informative)

    So does Intel hold enough share of the chipset market, for this to become an antitrust issue?

    Unless nVidia will license that same technology to ATI, it sounds like it freezes ATI out of the multi-GPU-on-Intel-chipsets market.

    • Re:Monopoly? (Score:4, Insightful)

      by networkBoy ( 774728 ) on Tuesday August 11, 2009 @04:20PM (#29030067) Journal

      ATI has crossfile.
      and no, this is not an antitrust issue (unless it's against nVidia), as Intel is paying nVidia for the tech.

    • by Reapman ( 740286 )

      Well considering ATI is actually AMD now.. not too sure AMD would even want that. I'm fairly certain you can get SLI going on AMD platforms still.

    • Re:Monopoly? (Score:5, Informative)

      by verbalcontract ( 909922 ) on Tuesday August 11, 2009 @04:22PM (#29030077)

      First of all, it's AMD.

      Second of all, ATI is owned by AMD. Because of them, AMD makes graphics cards. AMD also sells technology that allows two AMD cards to be used on one motherboard. Therefore, AMD will probably not pay money for the same technology that lets people purchase and use two competitor's products.

    • Re:Monopoly? (Score:4, Insightful)

      by sofar ( 317980 ) on Tuesday August 11, 2009 @04:23PM (#29030099) Homepage

      Unless nVidia will license that same technology to ATI, it sounds like it freezes ATI out of the multi-GPU-on-Intel-chipsets market.

      s/ATI/AMD/g

      why would AMD promote SLI when they can sell crossover? It seems they would cannibalize their own GPU market by supporting SLI on their chipsets.

    • Re: (Score:1, Redundant)

      by Desler ( 1608317 )

      Unless nVidia will license that same technology to ATI, it sounds like it freezes ATI out of the multi-GPU-on-Intel-chipsets market.

      Why would AMD want to license nVidia's SLI when it is a direct competitor to its own Crossfire technology?

      • Yeah, it sounds like a conflict of interest to want your competitor's cards on YOUR chipset that supports YOUR cards already. On the other side of the coin, though, I'd rather have my competitor's cards on MY chipset than lose a sale to my competitor's chipset because they support my competitor's cards.

    • Isn't this exactly the same as the P45 chipsets? Those things had SLI support but not crossfire; if you want Crossfire, you got/get an X48 or whatever. Seriously, this doesn't seem like news to me, other than a new chipset will be produced in the future. Now if it said there will NOT be a chipset for the new generation ATI cards... then that'd be news. Did I miss something?
      • Re:Monopoly? (Score:5, Informative)

        by Chabo ( 880571 ) on Tuesday August 11, 2009 @07:10PM (#29032017) Homepage Journal

        You have it backwards. Crossfire works on pretty much any motherboard with multiple PCIe slots, provided they have sufficient bandwidth between them. SLI requires the motherboard manufacturer to purchase a license from Nvidia to support SLI on the board.

        If I'm not mistaken, Intel's X58 board was the first Intel-made board to support SLI, and I don't think anyone made a P45 board supporting it.

      • by Briareos ( 21163 ) *

        I'm happily running 2 ATI Radeon 4850's in CrossFire (when needed) on my P45-based Gigabyte mainboard, so I think you've got something mixed up here...

        np: RJD2 - Hidden Track (Deadringer)

  • First this [slashdot.org] and then this [slashdot.org] and now this [slashdot.org]
  • Jen-Hsun Huang (Score:2, Interesting)

    by vivek7006 ( 585218 )

    So this is how the Nvidia CEO intended to open a can of whoop ass on Intel. What a dumb-ass...

    • Re: (Score:1, Informative)

      by Anonymous Coward
      Mods, the parent is not trolling. This is what Nvidia CEO said a while back http://techreport.com/discussions.x/14538 [techreport.com]
    • Re: (Score:3, Interesting)

      by Hal_Porter ( 817932 )

      Intel and NVidia have a complex relationship. On the one hand Intel and NVidia are jointly a natural alternative to AMD now it has bought ATI.

      Right now of course Intel sell vast numbers of low end GPUs and rely on NVidia for the low volume high end stuff - i.e. SLI for gamers. So Intel pretty much has to support SLI.

      On the other they are both hinting they will compete directly. Intel has been talking about CPU/GPU hybrids (i.e. Larrabee) for ages and there have been persistent rumours NVidia will launch an

  • by hubert.lepicki ( 1119397 ) on Tuesday August 11, 2009 @04:16PM (#29030025)

    Is it Intel that paid NVidia or the other way around? Having support for SLI is defo good thing for NVidia and for Intel, the question is who should care more.

    • by Sycraft-fu ( 314770 ) on Tuesday August 11, 2009 @04:37PM (#29030287)

      nVidia does charge for SLI licenses. Reason being that they are also in the motherboard chipset market and want you to buy theirs. Intel wasn't all that pleased with the situation and so refused to license QPI to nVidia, which would mean no Core i7 chipsets. Well, that got all resolved and licensing started happening both ways. My guess is neither side is paying the other all that much.

  • by Anonymous Coward

    TFA says that the Intel chipsets will be limited to 8 lanes instead of 16 to give Nvidia an advantage for thier own chipsets.

    Why is a license needed to interface with an IC in the first place?

    • Re: (Score:3, Insightful)

      Because intellectual property is sacred...
    • by rsmith-mac ( 639075 ) on Tuesday August 11, 2009 @05:12PM (#29030715)

      TFA says that the Intel chipsets will be limited to 8 lanes instead of 16 to give Nvidia an advantage for thier own chipsets.

      If that's what TFA says, then it's full of it. The design of Bloomfield/P55 is such that there are 16 PCIe lanes coming straight from the CPU, and then another 4 lanes coming off of the P55 chipset, routed back up to the CPU through whatever interconnect Intel is using (it is in effect just a Southbridge, with the Northbridge integrated in to the CPU). This is because Bloomfield was intended to be used with a single x16 PCIe slot, it's a mid-range product. Nehalem/X58 is Intel's high-end product, and it has a much faster QPI connection coming off of the CPU to interface with a proper Northbridge to feed more PCIe lanes.

      So what you have to do is split the 16 lanes in to two sets of 8 lanes, and then use those lanes to make two x16 slots that only have half the bandwidth they're supposed to (something the PCIe standard allows). That's why the Intel chipsets will be limited to 8 lanes.

      The NVIDIA chipset mentioned is the NF200, which is a PCIe bridge. It would sit at the end of the 16 lanes coming from the CPU, and in turn offer 32 lanes (2x16) for PCIe slots. This gives you the full 16 lanes of bandwidth to each slot, but it doesn't get you any more bandwidth to the CPU. You still only have 16 lanes of bandwidth to the CPU. The only advantage to using a bridge chip is that it means the full bandwidth of the CPU can be dynamically allocated to a single PCIe slot, and that two PCIe devices can communicate with each other at full speed (the NF200 also has a few SLI commands that make sending data out from the CPU faster by automatically replicating it to the two slots). This does little to solve the fact that you have too little bandwidth in the first place, which is why you won't see the NF200 used too much. Plus bridges add latency and complexity to motherboard designs.

      As for why a license is needed at all: because NVIDIA says so. Their products won't work in SLI mode unless they see a license or a NF200 bridge chip (which is an automatic buy-in for SLI). It's a scummy system really, the license doesn't actually do anything. At best it means some token testing was done to make sure SLI worked, when if you build to PCIe spec it would work anyhow.

  • by Anonymous Coward on Tuesday August 11, 2009 @04:21PM (#29030071)

    This is just a weak form of DRM. Nvidia's drivers check the motherboaïrd against a whitelist and if the mobo is on the list the driver allows SLI. Naturally, chipset/motherboard makers have to pay protection money, er licensing fees, to get on the list.

    • What prevents me from spoofing another manufacturer's id?
      • Re: (Score:3, Informative)

        by BrentH ( 1154987 )
        Nothing, which is why there is at least one Windows application that does exactly that.
      • Severely ticking off the maker of certain components? Think of what Palm and Apple are doing, however, Palm and Apple really don't have to communicate much, Palm does their thing and Apple does their thing, on the other hand, many motherboard makers have some components provided by third party manufacturers or licensed to them.
  • the relentless flash pop-up ads on that site (that re-open immediately upon clicking close) made it impossible to read the article. Something tells me this isn't what the site had in mind when it sold advertising space there ...
  • ... to get this to be a reality or I would expect to see Apple grabbing some of AMD/ATi harmony as the alternative.

    Apple is pushing OpenCL and it's rather limited [CPU-core bound OpenCL only] if none of Apple's systems have SLI support.

    • by guruevi ( 827432 )

      Wasn't SLI only supposed to be for being able to talk to multiple chips for a single output? OpenCL (and NVidia's CUDA) however don't need SLI in order to talk to multiple GPU's. I have a machine with 4 GPU's and CUDA works just fine. I really don't see the advantage of SLI other than bragging rights and the difference between 200fps and 300fps, most games (and other real time video outputs) have enough with a single video card and all others (scientific output) like to save their output on disk so stuff do

      • While I basically agree with the premise of what you are saying, doesn't the SLI bus provide for having a shared memory segment between all the cards? I'm not 100% sure that is a feature of that specific interconnect; but I've always assumed it was. The difference being that if n cards with x memory would have a single (n*x) memory pool local to the processors plus the overhead of whatever locking semantics that would require. v.s. having n spearate x sized pools plus whatever work predivision overhead and/
    • Apple doesn't support SLI, and if they wanted to they could "license" it directly from Nvidia.

      • by tyrione ( 134248 )

        Apple doesn't support SLI, and if they wanted to they could "license" it directly from Nvidia.

        Apple has had no need for SLI until now. The same option for Crossfire if Apple wanted to use Intel boards with Crossfire support. OpenCL changes all bets.

  • by billcopc ( 196330 ) <vrillco@yahoo.com> on Tuesday August 11, 2009 @05:56PM (#29031285) Homepage

    Frankly, I've been less than pleased with the whole SLI vs CrossFire debacle. It's a friggin PCI-Express bus! This is nothing more than software and/or firmware enforced lock-in, and it stinks. I would have preferred for Intel to reject the SLI tax entirely. NVidia is the small player here, they're the ones who should be bending over backwards to get the big guys to promote their products. I don't want to pay an SLI tax on my motherboard, NVidia should be plenty glad that I'm buying two expensive GPUs instead of one, and they should consider a partnership with Intel like a divine blessing because Intel is 10 times larger and has far greater reach into every single market.

    The situation is simple: right now, I own a perfectly fine motherboard that doesn't support SLI (Intel P35). I also have a perfectly fine Geforce 8800 and would have loved to add a second, but I can't because my board isn't on the SLI whitelist. My options are:

    A. buy the second card, and replace my motherboard with an overpriced unstable NForce 750 board

    B. fuck NVidia and buy two brand new AMD cards

    Assuming equal performance, option B would cost me far less, even though I would prefer the NVidia GPU. Their SLI lock in has thus resulted in a lost sale.

    Now I'm just one guy, but here's the funny part: I used to sell gaming rigs... lots and lots of 'em. When people heard about SLI, all the hardcore guys wanted it, but when they found out they had to taint their lovingly assembled systems with an NForce board, most of them backed off. It wasn't even about the money, it's about NVidia's awful track record in the chipset biz. They make even SIS look good. They never really fixed the NF2/3/4 disk corruption glitches, and they trashed the one good thing they had going for them: Soundstorm. That was a long time ago, but the way they handled those very public screwups left a bad taste in everyone's mouth.

    • Re: (Score:2, Informative)

      by coxymla ( 1372369 )
      Even if you had a SLI board already, buying another 8800GT is a horrible idea. Far more cost effective would be to sell the 8800GT you already have and upgrade to a better card (probably a 4870.)
      A single new card draws less power, makes less noise and heat, takes up less space, and probably runs faster in almost every game (but maybe not synthetic benchmarks.) SLI is a bust.
      • by MogNuts ( 97512 )

        1) New better performing cards don't always draw less power than an older single card (GTX 285 needs more power than a 8800 GTS 512).
        2) Yes, they do usually make less noise and heat
        3) Less space--never. Go from a 8800 GTS 512 to a GTX 285 or a 4870, check it out.

        I agree with you that a single card is better for heat and simplicity (installation, drivers, & game support), but lately I've noticed that SLI/Crossfire really is the future. In some games you just can't get the framerates you want at the detai

        • 1) New better performing cards don't always draw less power than an older single card (GTX 285 needs more power than a 8800 GTS 512). 2) Yes, they do usually make less noise and heat 3) Less space--never. Go from a 8800 GTS 512 to a GTX 285 or a 4870, check it out.

          He was talking about a single new card compared to two older cards in SLI.

          1) A GTX 285 needs more power than a GTS 8800 512, but it DEFINITELY does not need more power than two GTS 8800 512 cards.
          2) Newer cards almost invariably make less noise and heat than older cards, and certainly less than two older cards.
          3) I don't care how big your single new card is, it still takes less space than two older cards. Even the ridiculously-sized GTX 2XX series only take up about the same amount of space as two other

          • by MogNuts ( 97512 )

            3) No. The problem now is that of length. Everyone's cases nowadays can hold two video cards. Unless u buy a midsized or micro sized case, u can fit 2 cards. But all the new cards like the 285, 275, 4870x2 are all really damn long. I can fit a 8800 gts 512 in, but NOT a gtx285. It's just too long to fit in my case the way it's set up inside.

            And btw actually ur wrong on the other front as well. Pop in a 285 and it takes up two PCI slots. But slightly less powerful cards only take 1. So 2 cards fit where only

    • I agree with you. I was an Nvidia fanboi until I built my current system 6 months ago and it was because of the cost. I built a X4 940 system with a 4870X2 for about $600 less. I didnt have to replace my RAM and I didnt have to replace my motherboard. I have had 0 problems with my system. My coworker who had to have Nvidia has had nothing but problems with it. ATI has really come around with the help of AMD.
    • I have an newish model Asus NForce board, and it is top-notch. Someone who refuses to buy new NVidia hardware because some old NVidia hardware was flakey is someone who is out of sync with reality.

      I agree that NVidia should be giving away SLI licenses to support the purchase of their GPUs. But if you want SLI right now, just go for an NForce board--they rock.

      • by Goodl ( 518602 )
        You have got to be kidding, Nforce boards are dreadful!!! I was so glad to see the back of my Striker II Formula once X58/Nehalem came out, If I could i'd have taken a double barrel shotgun to it just to pay it back for a year of freeze ups disk corruption and bluescreens. Nforce boards rock indeed, what a crock!
    • by sznupi ( 719324 )

      Uhm, SiS chipsets were actually quite good at least from the time of 735. The only real problem with them was that they were usually put on cheapest motherboards...which still worked perfectly until failure from non-chipset related issues (just now caps failed on my ASRock SiS 746fx motherboard, which gave me flawless 5 years otherwise).

      Even X360 has SiS southbridge... (well, perhaps not the best example reliability-wise ;p, but the issues with it also seem to be not related to SiS)

      • I get what you're saying, and I agree that SiS chips are almost universally paired with shitty ass boards, but my gripes are with the actual chipset's functionality and driver support. I feel they are today where VIA was 7-8 years ago in terms of stability.

        • by sznupi ( 719324 )

          Weird, for me, and for 7 or so years they are on par with Intel in stability & being non-problematic, while just slightly slower (in strictly chipset related areas - say, PCI bus)

          Even better if you managed to get one of few good motherboards (MSI 745 for example)

          PS. "today"? SiS disappeared from the general market ~3 years ago, unfortunatelly...

Every nonzero finite dimensional inner product space has an orthonormal basis. It makes sense, when you don't think about it.

Working...