Apple + Nvidia Speculation + Recap

Posted:
in Future Apple Hardware edited January 2014
Since I have a feeling (as do others) that Apple and Nvidia are working closer than has been made public... I have detailed some of the news rumors and speculation (in date order) that have been about Nvidia and Apple...



All links have been given so you can read any/all of the items but I've cut what I thought was most interesting so we have it all in one place...



If anyone has anything else to contribute post it here...



Dave

Comments

  • Reply 1 of 13
    davegeedavegee Posts: 2,765member
    2002 Mar 1st [http://sci.newsfactor.com/perl/story/16589.html]

    ------------------------------------------------------

    Now, Apple (Nasdaq: AAPL) may be poised to push another standard called 3GIO, widely heralded as the successor to PCI and AGP expansion slots. Those slots, located on a computer's motherboard, allow insertion of video cards, networking cards and so on to enhance the computer's capabilities.



    Pundits have begun to speculate that Apple is planning to incorporate 3GIO into the much-anticipated revision of its aging high-end Power Mac architecture.



    3GI0 stands for "third-generation input/output." Originally called Arapahoe (named for a workgroup that included Compaq, Dell (Nasdaq: DELL) , IBM (NYSE: IBM) , Intel (Nasdaq: INTC) and Microsoft), the standard is expected to start shipping with desktop computers in 2004. Development tools and specifications are likely to appear in 2003.



    The PCI Special Interest Group (PCI-SIG), of which Apple is a member, is attempting to push the 3GIO standard while retaining support for the still-emerging PCI-X standard. The group is telling users that PCI-X is optimized for high-end applications used on servers and workstations, while 3GIO is targeted toward general-purpose applications that run on desktops and mobile devices.



    According to the PCI-SIG, a number of upcoming technological developments will require a faster internal input/output scheme.



    "Technologies such as CPU speeds that will exceed 10 GHz, faster memory speeds, higher-speed graphics, 1-Gigabit and 10-Gigabit LAN, 1394b, InfiniBand fabrics and others will drive the need for much greater internal system bandwidth," according to 3GIO product literature.



    Apple already includes a direct bus to the system controller on current Power Mac G4 systems, providing sustained throughput of 215 Mbps. The company claims that bottlenecks in most other PC architectures cause a slowdown to 133 Mbps when using certain applications.



    Current specifications for 3GIO show data running directly from an expansion device (such as a graphics card) through a memory bridge and into the CPU, avoiding any potential I/O bridge bottlenecks.



    The PCI-SIG claims 3GIO sports an initial frequency of 2.5 Gb/s/direction, which is expected to increase as silicon technology advances to 10Gb/s/direction (the theoretical maximum for signals sent via copper wire). 3GIO purportedly also will feature 100 MB per second per pin data transfer, compared with AGP 4X's 10 MB per second per pin throughput.



    HyperTransport Complement



    AMD is developing its own high-speed interconnection standard, independent of 3GIO, dubbed "HyperTransport." This I/O standard, which AMD officials said is not intended to compete with either 3GIO or PCI-X, will deliver 12.8 GB/sec.



    Apple is a member of the HyperTransport Consortium, lending credence to speculation that the company will choose to integrate some form of AMD's technology into its high-end systems. NVidia, which supplies virtually all graphics cards for current desktop Macs, is also a member of the HyperTransport Consortium.

    ------------------------------------------------------





    2002 Apr 16th [http://www.architosh.com/news/2002-04/2002c1-0412-applegraph1.phtml]

    ------------------------------------------------------

    Rumor: Apple developing advanced dual-engine graphics card, Nvidia in the picture?



    While Mac and PC users both have much to say about Apple's shortcomings in the 3D and special effects worlds, it has been clearly indentified for more than two years now that Apple's two chief performance concerns amount to: one, the lack of 'workstation level' graphics card options and, two, low bandwidth on the motherboard.The G4, itself, is not necessarily complained about. In fact, the G4 architecture has always been cited for its strong floating point performance and AltiVec processing unit.



    With this project Apple is apparently serious in making up for lost time and lost marketshare in the Hollywood studio market.



    According to sources of ours who have actually seen the unit, Apple has been quietly developing a powerful new graphics card which will finally rectify one of the chief complaints of highend video and 3D pros who say it keeps them off the Macintosh platform, despite the fact that they want to be there.



    In a special invitation only event (no press allowed) Apple held a 'Studio Summit' back in the early part of this year. Inviting all of the major Hollywood studios and effects houses, Apple made its best business case for Mac OS X and asked for critical feedback.



    At this special event Jobs candidly asked the question: what will it take for all of you to move over to Mac OS X? Apple was specifically looking for constructive information from which to make future plans. In a note from sources familiar with the private event, Jobs apparently said Apple was prepared to do whatever was necessary to satisfy the needs of this elite user-base.



    Chief 'wants' among the studios is for more powerful hardware. Specifically, they want 'best of breed' graphics performance and options there. They also want a 1U or 2U industry standard, rack-mountable Power Mac for render farm building and servers. And they want only duals and quads -- single processor Macs are not desired except at the portable level.



    So what does Apple have planned? Well our news comes from a source who says even he could not get definitive information, though apparently this new graphics card was inside of a G5 test box (though our sources say it may have been a future G4) which has been in one location for several months. Details of the inside of this box, which we have only touched upon in general terms, were recently discussed in greater technical detail in an AppleInsider forum. The details match up very nicely with information which we have published previously.



    Apple is apparently working on a superfast graphics card with twin-engines. Our sources tell us that the card may or may not be co-designed with Nvidia. Each GPU is fitted out with "heaps of ram" our sources say, estimating the amount at 128 MB for DDR-RAM for each engine (GPU). Besides a curious daughter-card slot on the card, our sources say that they know very little about the rest of the card?other than to say its performance on the test box was nothing short of astonishing.

    ------------------------------------------------------





    2002 May 1st [http://zdnet.com.com/2100-1103-896996.html]

    ------------------------------------------------------

    SAN FRANCISCO--Playing down accounting issues and an ongoing pricing dispute with Microsoft, Nvidia CEO Jen-Hsun Huang said an upcoming graphics chip should help propel the company to continued growth this year.



    Speaking on Wednesday at a chip conference here, Huang characterized the relationship with Microsoft as "very strong," adding, "We just asked a third party to help us solve some stuff."



    Nvidia disclosed Monday that it was in arbitration with Microsoft over the pricing of its chips used in the Xbox, a move Nvidia said could either lead it to recognize additional revenue or force it to sell some chips at a loss--depending on how the issue is resolved. Microsoft's Xbox accounts for 15 percent to 20 percent of Nvidia's sales. Nvidia also said Monday it was restating its earnings for the past three years because of accounting issues.



    Also on Monday, Nvidia raised its financial outlook for the just-ended quarter, and Huang said he sees continued market share gains this year leading to more growth. Some of that will come from a new graphics chip slated to arrive in August.



    The new chip will be manufactured on Taiwan Semiconductor Manufacturing Co.'s latest 0.13-micron manufacturing process, Huang said. Huang did not reveal the name or specific features of the chip, but did say it was a fundamentally new architecture from the GeForce 4 Titanium introduced earlier this year.



    "It is the most important contribution we've made to the graphics industry since the founding of this company," Huang said, speaking at the Merrill Lynch Hardware Heaven Technology Conference.



    Huang also noted the inroads the company has made in the Macintosh market, saying that the company expects to have the largest share of that market "very soon." Huang said that Nvidia's chips are in the eMac that Apple Computer introduced Monday.



    "You are going to see many more exciting introductions in that space," Huang said.



    The company also plans a follow-on to its nForce chipset, which works with Advanced Micro Devices' Athlon processors but not with Intel's Pentium 4. Huang said the company will continue to focus on chipsets for that market, rather than pursue a license to work with Intel's chips.



    Huang said the Athlon market represents about 30 million PCs a year, with Nvidia getting revenue in the "mid-$20s to mid-$30s" for each nForce chipset. Plus, he said, you don't have the risk of competing with Intel's own chipsets, as is the case in the Pentium 4 market.



    Chipsets are companion processors that allow the microprocessor to exchange information with other parts of the motherboard, such as memory. ------------------------------------------------------





    2002 May 17th [http://www.f1gamers.com/f1/apanel/view_news.php?id=2649]

    ------------------------------------------------------

    Unofficial specs of upcoming new Nvidia graphics cards.



    ======================================

    NV30: specs

    ======================================

    0.13 micron process

    400MHz GPU

    512-bit chip structure

    AGP 8X

    8 rendering pipelines

    Supports 128-256MB of DDR SDRAM

    900MHz DDR SDRAM

    200 million polygons per second

    Lightspeed Memory Architecture III

    Supports DirectX 9 and OpenGL 1.3

    release in 3rd or 4th quarter of 2002



    ======================================

    NV35 specs:

    ======================================

    0.13 micron process

    500MHz GPU

    512-bit chip structure

    AGP 8X

    8 rendering pipelines

    Supports 128-256MB of DDR SDRAM

    1000-1200MHz DDR or QDR

    400Mhz RAMDAC

    Lightspeed Memory Architecture III

    Supports DirectX 9.1 and OpenGL 2.0

    Release in first half of 2003

    ------------------------------------------------------
  • Reply 2 of 13
    wwworkwwwork Posts: 140member
    Dave is bored and can't wait.
  • Reply 3 of 13
    nonsuchnonsuch Posts: 293member
    Thanks Dave ... glad someone finally took the time to do that.



    (BTW, it's "bint," not "bink." )
  • Reply 4 of 13
    I hope this isn't true.

    1. It is hard enough to get decent graphics boards for the Mac now, we will only have one supplier if this so called tie up takes place.

    2. What has nVidia really done for Apple? They release a very few of their products for the platform, and the ones they do are late. With bad software. Yea, thats the company I want to get in bed with!



    No, this isn't going to happen.



    Just a note: I wouldn't use architosh as a source, they are a fairly poor site for Mac architectural news, and even worse site for predictions. They also predicted servers with G4 and G5 chips, too.
  • Reply 5 of 13
    penheadpenhead Posts: 45member
    Intersting ...



    I also think that the DDR memory controller in the XServe is inspired by nvidia's TwinBank memry architechture for nforce.



    Normal 266 DDR memory bus: 64-bit, 2,1GB/s

    TwinBank: 128-bit (with two sticks in strategic places) 4,2 GB/s

    XServe: 128-bit 2,1 GB/s (!)

    -anyone got more info on this?



    Im pretty sure that nvidia is trying to get into Apples mobo business as well. They already said that if they can't own the mobo market like they do graphics, theres no point in being there



    In any case, the cozy relationship between them will prolly let apple learn and exchange technologies.
  • Reply 6 of 13
    eskimoeskimo Posts: 474member
    What was the point of that first news posting? To indicate that Apple might be smart enough to include industry standard interface technologies developed by Intel into their computers? 3GIO now called PCI Express will be the standard by which expansion cards interface with computers in the future. It was invented and developed by Intel just like AGP. The fact that Apple would use it in future systems is them facing the reality of the market, not driving anything.
  • Reply 7 of 13
    davegeedavegee Posts: 2,765member
    [quote]Originally posted by Eskimo:

    <strong>What was the point of that first news posting? To indicate that Apple might be smart enough to include industry standard interface technologies developed by Intel into their computers? 3GIO now called PCI Express.....</strong><hr></blockquote>



    Well no point really but just found it interesting to include... What with RapidIO and HyperTransport being talked about so often I just though I'd add it to the mix. You are correct that it's a tad off topic as far as Apple+nVidia goes..



    Maybe I should pull that ref.



    Dave
  • Reply 8 of 13
    programmerprogrammer Posts: 3,457member
    3GIO and HyperTransport are going to be complementary technologies... HT in particular is intended and designed for point-to-point chip communications on the same board. 3GIO (PCI Express) is intended for longer distances and removable boards, with many devices sharing the bus. RapidIO is more like HT than 3GIO, but it gets a little fuzzier because RapidIO creates more than just point-to-point links. In fact even HT has defined a bunch of message passing features in their protocol that allows daisy-chaining of linked chips to propogate messages around. But the intention is for HT/RIO to be how chips communicate with the 3GIO bridge or tunnel chips.
  • Reply 9 of 13
    programmerprogrammer Posts: 3,457member
    [quote]Originally posted by Thai Moof:

    <strong>I hope this isn't true.

    1. It is hard enough to get decent graphics boards for the Mac now, we will only have one supplier if this so called tie up takes place.

    2. What has nVidia really done for Apple? They release a very few of their products for the platform, and the ones they do are late. With bad software. Yea, thats the company I want to get in bed with!



    No, this isn't going to happen.



    Just a note: I wouldn't use architosh as a source, they are a fairly poor site for Mac architectural news, and even worse site for predictions. They also predicted servers with G4 and G5 chips, too.</strong><hr></blockquote>





    1) Apple will likely provide it only as a BTO option, and have it in only one model PM. Such a monster card will be relatively expensive.



    2) nVidia provides the PC market with chipsets and the board vendors and OEMs can pick and choose which chipset and memory configurations they use. Apple is alone in the Mac market, and they don't want to have a wide variety of graphics boards, so there are fewer options. This is not nVidia's fault. ATI is different, they actually build and sell retail graphics boards directly. nVidia also fairly recently started building drivers for the Mac, which is a non-trivial proposition (at least in MacOS9 its a pain). Further, Apple has restricted the use of vendor specific OpenGL extensions -- they want to have standardized ones that take a long time to develop and bring to market. As for bringing the hardware to market, yes it appears that the Mac gets things later but we aren't privvy to why that is. My guess is that nVidia's longer standing contracts get priority and since Apple was slow to adopt nVidia they end up getting the short end of the stick. Perhaps with this rumoured new board that will change.



    nVidia isn't perfect, but they don't deserved to be slagged like that. Not any more than ATI does, at any rate!



    As for the rumour about an Apple/nVidia board... nothing stated there implies that it is anything but the next generation nVidia board (i.e. nv30)! The key part is likely not yet mentioned, like a HyperTransport connection in place of the AGP bus or something exotic like that... otherwise its just Apple being aggressive (finally) about truly getting a leading edge card out right away. And 128MB of VRAM is not any more than you can currently get on a GF4Ti.



    The comment about nVidia continuing to chase the AMD chipset market is particularly interesting... that means they are going to be making HyperTransport chipsets because the new AMD processors (with and without memory controllers on board) use HyperTransport links to the chipset. If Apple had Moto replace MPX with HyperTransport then they could use the nVidia chipset.



    Things that make you go "Hmmmmm".





    BTW: I've seen nothing saying that the Xserve memory controller's memory bus is anything more than 64-bits wide. The 2.1 GB/sec claim by Apple certainly reinforces that. What is your source about that?
  • Reply 10 of 13
    zazzaz Posts: 177member
    [quote]Originally posted by penhead:

    <strong>Intersting ...



    Im pretty sure that nvidia is trying to get into Apples mobo business as well. They already said that if they can't own the mobo market like they do graphics, theres no point in being there



    In any case, the cozy relationship between them will prolly let apple learn and exchange technologies.</strong><hr></blockquote>



    Considering that the nForce is a really good first effort on nVidia's part... 2.0 will be a sure winner... and most of it, short of PPC chip control, is something that would be entirely beneficial to Apple products.



    Their target is OEM's mostly.



    The board minus the AGP and PCI slots are ripe for iMac style units. Withh them add a little FireWire on board and gigabit enthernet with PPC controlers and it all looks pretty good.



    The OSX foundation makes the premise reasonable, yes?



    And with it comes 5.1 audio, high bus and memory speeds and a development cycle on a large scale by a company looking to aggressively get into the market.



    I am sure there are caveats to the endeavor, but on the whole an Apple, nVidia, IBM alliance would make a fine new trinity. :cool:
  • Reply 11 of 13
    davegeedavegee Posts: 2,765member
    [quote]Originally posted by Programmer:

    <strong>BTW: I've seen nothing saying that the Xserve memory controller's memory bus is anything more than 64-bits wide. The 2.1 GB/sec claim by Apple certainly reinforces that. What is your source about that?</strong><hr></blockquote>



    Well this from the xserve tech specs seems to agree with Programmer:



    256MB or 512MB of 266MHz PC2100 DDR SDRAM with up to 2.1GB/s throughput

    Four DIMM slots supporting up to 2GB of DDR SDRAM using the following:

    ?\t128MB or 256MB DIMMs (64-bit-wide, 128Mb technology)

    ?\t512MB DIMMs (64-bit-wide, 256Mb technology)



    Dave
  • Reply 12 of 13
    zazzaz Posts: 177member
    [quote]Originally posted by DaveGee:

    <strong>



    Well this from the xserve tech specs seems to agree with Programmer:



    256MB or 512MB of 266MHz PC2100 DDR SDRAM with up to 2.1GB/s throughput

    Four DIMM slots supporting up to 2GB of DDR SDRAM using the following:

    ?\t128MB or 256MB DIMMs (64-bit-wide, 128Mb technology)

    ?\t512MB DIMMs (64-bit-wide, 256Mb technology)



    Dave</strong><hr></blockquote>



    That is the theoretical limit of the ram itself, thus the PC2100 name.



    It does not change the atributes of the MPX 133mzh bus that currently tops out at 1GB/s



    [quote]from Tech Specs:

    133MHz system bus supporting over 1GB/s data throughput

    <hr></blockquote>
  • Reply 13 of 13
    penheadpenhead Posts: 45member
    Sorry, my bad on xserve memory





    [quote] from apple.com/xserve/specs.html:



    Full 128-bit internal memory data paths <hr></blockquote>



    This is AltiVec i guess ...



    I just reasoned that since they were making their own memory controller, something like this could be easy to do, and since nvidia does it, why shouldn't Apple? I've been asking this a lot lately ..

    (Im building an AMD system based on nforce nowadays for "my other" job: nforce takes more than $100 off the total price)
Sign In or Register to comment.