Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
Intel

Intel exiting graphics chips market 83

KEM writes "According to this news piece, Intel is giving up on the graphics market " They are going to still make integrated chipsets, but will no longer be making discrete graphics chips. Intel chalks it up not being able to keep up with the other chip makers, despite their purchase of Chips and Technologies in 1997, which was supposed to give them that edge.
This discussion has been archived. No new comments can be posted.

Intel exiting graphics chips market

Comments Filter:
  • When Intel started making chipsets for mother boards how many chipset companies went out of business or left that market? Unless you are talking about non Intel processors how many boards do you see with non Intel chipsets? Don't count super 7 boards even if they can run an Intel processor because they are really built for Amd chips. I'm sure they planned on doing this with video cards as well. Unfortunately for them competition in this arena was a bit more fierce than the motherboard chipset arena. Also they have had a big brother named FTC take a greater notice of them. Last thing you want to do at this point is be seen forcing competition out of another area. So what do you do? Exactly what they did. Play a low profile (in the graphics area) after a while, and then claim to get out of the business. Siting fast paced competition as the reason. At least they are smarter than Bill at this game.
  • Well, this theoretical new PCI bus would be (gasp!) backwards compatible - kindof like plugging SCSI-1 devices into a FastUltraMegaSuperDuperSCSI-3 adapter. (with the exception of the wide-narrow SCSI differences, which CAN be done if you're a wiz with termination - um and determination.)


    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • PC with 3 PCI slots =
    A video card, a SCSI card, an ethernet card, or two video cards, and a SCSI card, or use onboard video, and two SCSI cards and an ethernet card, or two ethernet cards and a SCSI card.
    Versitility, a graphics workstation, a file server with heavy storage, or more network capacity.

    PC with 2 PCI slots and 1 AGP slot = Less choices in the above!!!
    = a market where vidboard manufacturers running scared from Intel make AGP cards instead of PCI cards, so the latest board isn't avail. for your legacy hardware with PCI only.

    Make a better PCI bus, chuck AGP in the trash.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • Wow, go figure. And now there's nothing left of this evil legacy other than AGP, which is just a faster but much, much less versitile PCI bus. I thought the whole point of PCs was versitility. Oops, wrong, the whole point of PCs was cornering markets, and sucking money from consumers. I hope this AGP crap disappears and is replaced by a better, multi-purpose bus, or at least an updated PCI.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • The boards were hypothetical examples.

    in 99% of the cases, PCI isn't the bottleneck anyway.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • Okay, so I'm being a little overly dramatic on this issue. But the limitations of PCI would have better been addressed by a better multi-purpose PCI bus, not a specialized AGP bus built by intel to give it an edge in trying to take over a new market. Thank goodness it didn't work, and good riddance to intel from the video chip market. I can't wait until they exit the CPU market too! (I can dream).

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • I didn't realize that Microsoft made Intel do anything. Do you have a link to a statement to this effect?

    I always thought that it was just an oversight in the 80286 design and fixed in the 80386. Kinda how that Motorola (I think) processor had undefined opcodes that would disconnect the bus and short itself out internally. Probably meant as a test of some sort but got left in production and had nasty effects in production boxes. An oversight, and fixed in the next release.
  • Why the HELL is it a security risk to drop from Protected to Real mode?? The mechanism is a Ring 0 sequence so the code would have to be run by the kernel. If your OS lets you run code in Ring 0, then it's the OS's fault, not the processor's.

    Gimme a break. The code doing so would have to reset all the interrupt vectors, segments and stack pointers. it isn't as simple as flipping va switch and keeping everything running.

    'sides, triple faulting a CPU was pretty damn fast, considering IBM's decision to go through the keyboard controller instead. It just required BIOS help, which was the main problem.

  • They also have some pretty good video cards...

    For 3dfx they seem to head for the way down, the inovation that they once had seem to be gone now...

    And let's not forget the laptop computers that need a video card too (mainly NeoMagic, C&T, there where some Cirrus Logic too)...

  • I don't think this is true. The 64 bit addressing won't be used, that is all. Most 64 bit machines don't have the amount of memory alone to justify the 64 bit addressing, but it does improve throughbut.
  • I hate AGP with a passion, sure it is faster and communicates with RAM better but why make it such a limiting bus. AGP limits people, for instance someone buys a motherboard and later would like a second AGP video card/monitor they can't just just go and buy another one to put into their machine. They are stuck buying a new MB or getting a special new video card. The point behind PCI was it's versitility. I don't want to have a separate bus just for my NIC/modem/video/scsi/whatever. "Specialization is for insects." Heinlein

    How about a new PCI bus. I'd like to see 64 bit PCI (I know there are some out there but unless Intel markets this technology it won't become mainstream too easily) and how about being able to hot swap PCI cards. Well that might take a bit of work but still wouldn't it be nice.
  • It was only about a year and a half ago that I read in the news that Intel was coming out with some graphics chips to essentially put S3, Matrox, Diamond (etc) out of business. Reality is, nobody ever really used their chips. Their performance was always behind other products, and buying one of their graphics products was almost a guarantee that you'd have something semi-supported or not at all.

    Everyone was worried that Intel would start producing boards with their graphics chips built in (and integrated directly to the bus controller chips). I guess this proves that wrong.

    I think 3Dfx and nVida are the only choices, even though 3Dfx looks like the Microsoft-wannabe of the graphics chip world. Matrox, S3 and ATI are great too; I use all three and think they're great.

    --

  • Poor Intel. What a tragedy. I guess they will have to look for something new to manufacture on those fabs they built ;-).
  • Sorry to burst your bubble of twisted reality, but it took a while and a lot of cajoling to get Matrox to release enough information for the damned card to work with XFree86.
  • Well ok they exit the business and incorporate the instruction set in the next gen Pentium. MB's are now cheaper to make and it drives the quest for ever faster ever newer ever premium priced consumer machines. This to me is obviously what they have in mind for positioning the FPU performance of the IA64.
  • >I think 3Dfx and nVida are the only choices, even >though 3Dfx looks like the Microsoft-wannabe of >the graphics chip world

    I have to disagree with this. They may be trying to maximize profits by refusing to sell their chips to other manufacturers, but they are definately not taking after MS.

    There is more support for 3DFX cards than any other 3D card in Linux. There are less compatibility issues with other hardware than nVidia. (Try using a TNT2 in an ASUS mobo some time) And they're more reasonably priced for the performance you get, IMNSHO.

    I do like nVidia, and I like the TNT and TNT2, and I own cards with each of those chipsets, but I have to say my main workstation has a 3dfx card in it. (mostly because of linux support, but even for the Wind0ze partition games)

  • It was obvious from the start that the Intel i740 chipset was no match for the chipsets from nVidia (TNT, TNT2, Ultra TNT2), Matrox (G200, G400 and G400 MAX), S3 (Savage4 and Savage4 Extreme), and ATI (Rage Pro and Rage 128).

    The newest of the non-Intel graphics chipsets I mentioned can do AGP 4X, 32-bit graphics acceleration, and so on. Given that type of competition, no wonder Intel threw in the towel.
  • AGP is cool when compared to 32bit PCI that you can get with a 32bit processor. Yes, 64bit PCI exists - its in Alphas (and Sun's?) and its also runs at a faster buss speed - that is its 100mHz insted of 66mHz.

    But you cant use a 64bit expansion card with a 32bit processor. So something else had to solve the bandwidth problem of PCI32.

    Hopefuly, when 64bit intels come on the street Ill be able to plug in insanly fast sun or alpha videocards. 4d60t *growl*

  • Because we'd be stuck with crap like AMD


    AMD make bloody good processors that are much more cost-effective than Intel's offerings of 3 times the price.. and as the other Intel / AMD topic on /. today talks of, Intel can't stand being beaten in that field so have to play dirty. Dunno why you'd want to side with the 'market bully'..


    And who ARE Transmeta these days? :>


  • I concede from a system integrator's standpoint, the availability of an AGP port at the expense of a PCI slot sucks. I try to steer clear of motherboard manufacturers who do this.

    But on the other point you are incorrect. Video card manufacturers are choosing AGP not because they are scared of Intel, but because you can do a whole lot more with AGP, and in order to keep up with the competition performance-wise they need to use it.

    There are simply some things hardware-wise you just cannot do with PCI.
  • You make it quite clear in your post that you have no idea what benefits AGP gives to a video system, and are in general quite clueless about the things you are trying to talk about.

    In what way exactly is it less "versitile"? AGP in fact is more versitile than PCI from a technical standpoint, given its ability to make use of system DRAM more efficiently than PCI. That along with sidebanding and write-combining makes AGP much better a choice than PCI for anything (not just video cards) that has high bandwith requirements and/or memory requirements.
  • Nice little pop today in TDFX, NVDA, and ATYT. Especially NVDA. Curiously, a much smaller response from SIII/DIMD.
  • Yeah, exactly, I think that explains it better. Although AMD people would have a more personal relationship and dependency in their company because of its size and potential growth factor.
  • Well it just seems to me that AMD is catching up pretty damn quick compared to their size and budget of intel. I'm just proposing this as an idea. Maybe as another user said, they have a more personal involment in the development of the chip.
  • I have a theory that the reason why such big companies don't have as good an edge as what they think they might is due to the work ethic of those who work there. AMD for example has yet to produce a profit. They're working like dogs to produce fast chips cheaply. Intel on the other hand has already established themselves, and the work is distributed among many people. They may have deadlines, but its not the end of the world if they miss them (unlike AMD). Its not as much a monopolistic view as it is a financial view. You can see the same thing with MS. They can miss deadlines, however if they were to reduce the devlopment team to 100 people and give them ultimatums, I think they'd produce a better system. Its just a theory of course, and there are obviously companies that can contradict. Any comments? FYI, I won't respond to flamebait, so if you want a response from me, be polite.
  • Were you using a non-intel chipset on your motherboard? I remember having some compatibility problems running the i740 chipset on my Tyan 1590s Trinity motherboard, which used the VIA MVP3 chipset. They released a ton of patches trying to make the video card work properly, the patch version 2.8 actually ran decent (the latest patch crashed my computer bad). Finally gave in and the other day got a cheap Voodoo3 3000 and threw it in there. No more compatibility problems, and tons better framerate anyhow. Now back to slaying Young Kodiaks...
  • Your last sentence made me curious: is there a SCSI card that plugs
    into an AGP port? Most SCSI devices need high bandwidth and often
    have memory requirements, else you could use a cheaper PCI. So
    how about an AGP to UW-SCSI adapter? Anyone?
  • Hmphh...I didn't know they were into graphics chips in the first place.
    Afterthough: If you don't have any nice chips to make, don't make any at all.

    ---------------------------
  • and intel dropping out of business would be a bad thing.. why?
  • Remember, although Intel may not be designing and producing their own graphics chips, they still probably own a fair number of patents and the prior art being used in others'...

    As long as any such patents remain in their portfolio, they stand to gain revenue through licensing.

    Just a little food fer thought. :-)
  • I have used the i740 chipset up until now and I have to say, it was horrible.. I couldn't even run Win98 without it crashing every hour.. unless I switched it into 16 color @ 640x480..

    One less chip maker to worry about hyping their chipset
  • Uhh... you ever worked at Intel? I work about 10-12 hours daily, and other engineers do that too. I'm sure AMD engineers do likewise in this crazy industry. Designing and getting a processor to work, AND making it manufacturable in high volumes is an incredibly difficult job, and I'm constantly amazed that these things actually work :)

    Andrew

    p/s: This is just my opinion, and obviously does not reflect my employer's views on things.
  • Don't forget "creating hardware bugs we won't acknowledge and certainly won't fix, and then making them standard" Ever try to use a monochrome video card for debugging in a machine with an AGP video card?
  • Fast, fully supported 3d, AND a TV Tuner. they make great boards, fast as all hell, now if they can just figure out how to do 32-bit rendering, and get fully into the OEM market, I'll buy their stock.

    BTW: I thought their commercials kicked ass, except for the thing that nobody can quite get right...30 second of screen shots please. I can't stand game commercials that have 2 second of actual gameply and 28 of some idiot actors doing stupid sh*t. The QuakeII commercial is a perfect example of how NOT to do one.

  • I've been using an original Millenium 8MB for the longest time, and it still is very fast, although the design is over 8 years old. I got a Voodoo 2, so 3D games are no problem (but lack of PCI slots is :-( )

    I saw some Xmark performace figures with the Xfree86 pre 4.0 release on a G400 a while back - around 53 Xmarks! Thats really good for a consumer card that costs less than $200.

    Anyway Matrox always has had THE BEST 2D performance of any video card, and the best Xfree86 drivers.
  • That's close, but not quite right. Somewhere sitting around my house is an old issue of Boot magazine (I think they morphed in maximum pc, or something like that), anyways, in this issue they had a big 3d showdown between all of the various cards of that time. From what I can remember, the Voodoo 2(yes it was out at the same time as the i740) won for speed, and the i740 won for visual quality. On the fps graphs, the i740 came third, beaten by the Voodoo 2, and the Riva 128.
  • > Must be a user error...

    More likely that it would be hardware incompatability, unless this particular user started stabbing a butter knife into his i740.

    If you ask me, "Must be a user error..." tends to be one of those knee-jerk reactions like "Well of course it doesn't work, it's Windows!" :)

    Not that I disagree with the latter, in most cases...
  • This does remove some of the potential Microsoft 'it's a feature effect' from the market. Intel always had the potential to sell their graphics chips at cost or at a loss, just to get market share. Now everyone is basically at the same level, meaning none of them can afford to sell at a loss. So hopefully the competition will be in the performance, feature, and support areas. Except at the very low end of the cost equation where Intel will still be a player.
  • Intel are now having to drop products after commiting to Linux support and almost forgetting that most people still use Windows...

    Yea. If only Intel hadn't fallen for all this "Open Source" Linux hooey. Suddenly, a very paranoid and market-aware company went off the deep end. If they had only stayed in a Windows-only environment where a cruddy product isn't as important as good marketing, they'd still be making graphics chips today! Damn those Linux people and their crazy talk about performance!

  • Hmmm.. After reading all the posted comments it seems the discussions are aimed a little low. Can you say " Playstation II "? Sure you can. Your lives are all about to change...
  • We have 64bit and 66MHz PCI. Join them together and Yum.
  • The i740 was a kick-ass chipset for its time. Back then the 3D landscape was Riva 128, Rage Pro, and Voodoo 1. TNT was stil in development, and the GXXX were not announced yet. At a time when you had the choice of good visual quality (Voodoo), faster speed and bigger textures (Riva 128) or neither (Rage Pro) the i740 was almost as fast as a Riva 128, had better visual quality than a Voodoo and supported AGP 2X like the Rage Pro. Unfortunatly that only lasted until the Voodoo 2 came out but still.
  • Thats bull, the only reason the TNT did not work on Super7 motherboards (not just Asus) is becuase the first generation were poorly designed (the motherboards) and did not allocate enough power to the graphics card. The 3Dfx Voodoo3 chips are just overclocked Voodoo 1s that don't really add anything that we haven't seen before. C'mon 16MB of memory AND no AGP support! What no 32 bit color! Plus the hercules dynamite TNT 2 ultra card is the fastest for D3D and the G400 MAX (I think with the new drivers and ICD) is the fastest GL card.
  • Actually I'm still right, i740 had a jump on Voodoo 2 by 2-3 months, by the time that article was published Voodoo 2 had started the 3rd gen accelerators. i740 came at the very end of the 2nd gen chips. 2 issues before that, i740 won a kickass award. BTW. The Verite won for best visual quality.
  • I actually ran 2 i740 boards on my two computers, and they ran sweet, and still do.

    On the other hand, what is going on with the new Intel graphics chip they just brought out not too long ago? Are they ditching that too, including support?
  • I've had no problems with the i740 in either Windows or Linux. In fact, I'm running it in Linux on my home computer right now (16BIT -- not COLOR -- mode in 1600x1200). Must be a user error...

  • There would be no point in connecting a SCSI card to an AGP port. The SCSI specifications only allow for data transfers speeds ( 20/40/80? MB/s ) that are but a small fraction of the PCI bus speed (135? MB/S), so it is not the PCI bus that is the bottleneck with SCSI devices. This is the reason why some disk RAID-systems work through multiple SCSI connections (on the same PCI bus) to allow for faster data transfer. There is a company that sells RAID systems that connect directly to the PCI bus, circumventing the SCSI bottleneck
  • two competent companies, would fight each other driving performance up and price down...

    but we wouldnt have all these Mofos with different cards totally screwing compatibility and stuff.

    :)
  • no way dude, I put big bucks on the next 3dfx chip being some hard core cool stuff :)

    as for the Matrox, eh.

    Too many cards, too much pain for developers, too many games programmed for the lowest common denominator.
  • intel forgot people use windows.
    yeah.
    and this somehow relates to them making video chips...yeah.

    WHAT ARE YOU TALKING ABOUT WEIRDO!!!
  • 1. I havent heard of any linux drivers for that card, so they didnt put all that much effort into them

    2. drivers are shoddy for all cards right now due to the pace of development

    3. if the core market is shit, its nice to know that some comapnies have the balls to try and make a transistion. the almighty buck is not the only thing in the universe. theres quake and sex too.
  • by jlund ( 73067 )
    So Intel leaves the discreet market to focus on integrated chip sets the same day that nVidia announces an agreement with ALI (Acer Labs) for northbridge integration...

    http://biz.yahoo.com/bw/990818/ca_nvidia_2.html

    Ouch...

  • Really, Intel should have to add such comments to these announcements. Things like, "well, we're no longer making network cards. Ever. Or, at least until next year. We'll continue R&D on them, but our stuff gets pounded in the marketplace, so we'll say we're 'leaving the market' until the R&D guys come up with something decent."
  • Matrox does rock. Good performance, absolutely solid drivers.

    3dfx is dead. Everything's going the way of OpenGL and DX.
  • They have even released the specs on the THREE-D portions of their latest cards, thus guaranteeing that hardware accelerated OpenGL for Linux for the Matrox cards is just around the corner (with XFree86 4.0). NVidia and 3dfx won't do that, saying that keeping their 3d engine secret gives them a 'competitive advantage'. 3dfx cards work in OpenGL mode under Mesa only because of the proprietary GLIDE libraries, where 3dfx munificently allowed someone under non-disclosure to port GLIDE to Linux (as closed source).

    -E
  • The current PCI bus spec tops out at 33Mhz. The current AGP bus spec tops out at 66Mhz. Intel now has a 66Mhz PCI spec (i.e., just as fast as AGP), but a) it is currently available only in their latest motherboard, and b) in that motherboard, it requires eliminating the AGP slot. As for 64-bit PCI, both Compaq and AMI have motherboards that do 64-bit PCI with Intel 32-bit processors. Remember that the width of the bus does not have to be the same as the width of the processor... for example, current chipsets already have a 64-bit bus to main memory, and serialize it to a stream of 32-bit words when it comes time to present the data to the processor. And some Alpha motherboards have busses as wide as 256 bits to main memory, and again, serialize it to a stream of 64-bit words when it comes time to present the data to the (64-bit) processor.

    A year ago Linux did not support the 64-bit PCI spec even on Alphas (which did have 64-bit PCI slots even back then). But I have not checked things out lately, too many other fun things to do and I'm not in that business anymore.

    -E
  • Yeah, I just sold my NVDA today! What a pleasant surprise! Any idea why all that happened?

The only difference between a car salesman and a computer salesman is that the car salesman knows he's lying.

Working...