Report: AMD to outsource chipset design to ASMedia

AMD’s acquisition of ATI in 2006 eventually spelled the end of third-party chipset designs. Now, according to industry sources quoted by DigiTimes, things may be about to change.

This time, the lucky outsourcee would be ASMedia—one of the subsidiaries of Asustek, Asus’ parent company. Word got around last month that AMD might partner up with ASMedia on SATA Express controller design, potentially through a licensing agreement. Earlier this week, however, DigiTimes said its “sources from the upstream supply chain” revealed that the partnership “may advance further with an outsourcing of the whole chipset R&D.”

When asked to chime in, AMD told DigiTimes that it “does not comment on market speculation.” I’d probably take this rumor with a healthy helping of salt, all things considered.

Outsourcing chipset R&D might make business sense for AMD, which remains in a somewhat precarious situation financially. The stakes in terms of overall system performance and stability could be relatively low this time, as well, since modern chipsets are little more than glorified I/O hubs. There’s no chance of a wonky third-party memory controller crippling a new CPU.

Still, as Geoff pointed out last month, ASMedia’s USB and SATA controllers typically have lower performance than those built into Intel chipsets. I suppose we’ll have to wait and see—if DigiTimes’ sources are right—how competitive ASMedia controllers can be when connected directly to the processor.

Comments closed
    • UnfriendlyFire
    • 7 years ago

    TR should do a review on the I/O stuff after ASMedia starts making stuff for AMD.

    What good is a computer if it keeps BSODing due to s*** firmware?

      • Airmantharp
      • 7 years ago

      I’m willing to bet that ASMedia’s parts, and boards based on them, will be far more solid than a survey of their currently released products would logically lead one to believe.

      ASUSTek has nothing to lose here but profit, and firmware (BIOS, UEFI) implementations appear to be one of their strengths. It’s the drivers we should be worried about :).

    • ronch
    • 7 years ago

    Does AMD need to pay Asmedia anything to make chipsets? Back in the old days it’s the chipset vendors who have to get a license from the CPU makers to design and produce chipsets, right? If anything, Asmedia would be the one who will profit from this.

      • Airmantharp
      • 7 years ago

      They both make profit- otherwise it wouldn’t be business, it’d be extortion (or something else not terribly legal or ethical).

      Really, all ASMedia needs is a license to use Hypertransport; the only thing keeping ASMedia or anyone else is that Intel refuses to license their system bus. But that’s not a bad thing either, in Intel’s case- they may not be completely up-to-date all the time, but they’re solid.

    • Rza79
    • 7 years ago

    Now that AMD finally got their act together and their USB and SATA performance is as good as Intel’s, they decide to change course. Maybe they expect at least as good performance from ASMedia. The choice of ASMedia kinda makes sense. Years ago, VIA’s head of chipset design and design team moved to ASMedia. ASMedia’s office even got raided by the police on VIA’s allegations that ASMedia stole their USB 3.0 IP.

      • Airmantharp
      • 7 years ago

      You might want to re-research your ‘as good as Intel’s’ perspective. If ASMedia can get their drivers solidified and stable, their controllers will be at least as good as AMD’s, and likely better as they have a broader market to service.

      And I don’t know why anyone would have wanted to steal VIA’s USB 3.0 IP.

    • ronch
    • 7 years ago

    AMD – Butter over too much bread.

    • ronch
    • 7 years ago

    If AMD is going back to third-party chipsets they might as well put VIA and SiS on the list instead of relying solely on ASMedia for chipsets. That is, if these two still have the people to do it. Nvidia would be good too but I don’t think they would be interested unless AMD has killer products coming out.

      • Airmantharp
      • 7 years ago

      That’s the issue- no one else is doing it anymore, as AMD gobbled up ATi’s business, and Intel essentially perfected the art and brought it all in house, to great positive effect.

      ASMedia has probably collected what resources (people and loose companies with applicable IP) are left in order to build a viable business, and in this case, to get AMD’s business.

    • ronch
    • 7 years ago

    AMD. ARM. ASMedia.

    Triple A!!!

    • sschaem
    • 7 years ago

    “AMD’s acquisition of ATI in 2006 spelled the end of third-party chipset designs.”

    Didn’t nvidia produce AMD chipset well into 2007?

      • Cyril
      • 7 years ago

      Sure, but the acquisition still spelled the end of third-party chipsets.

      Added an “eventually” in that sentence just for you. 😉

        • sschaem
        • 7 years ago

        Eventually 🙂
        [url<]https://techreport.com/review/14661/nvidia-nforce-780a-sli-chipset/2[/url<] My htpc is using an nforce chipset from 2007, and I'm sure a TR review at the time helped me make that choice. (And still going strong after 7+ years of 24/7 usage)

          • Deanjo
          • 7 years ago

          Actually the last one was the 980a (yes it was a rebrand).

    • chuckula
    • 7 years ago

    If this is a prelude to AMD just bringing the southbridge on-die for most or all of its product line then NBD really.

      • Airmantharp
      • 7 years ago

      You’d expect that for SoCs, but even Intel continues to maintain their ICHs (slowly). I suspect that multi-channel SDRAM controllers and strict power requirements keep them from building higher-end SoCs; if it were a reasonable move, you’d expect Intel to be doing it already, given how much it’d save them in manufacturing costs and how much resulting fab capacity would be freed up, in my opinion.

    • slaimus
    • 7 years ago

    I own an Intel motherboard where the PCI slot is supplied by an ASMedia 1083 bridge. Very few PCI cards work properly without freezing the system, and I had to basically ignore it. Apparently the problem is widespread: [url<]https://groups.google.com/forum/#!topic/linux.kernel/D_ws98CwtGw[/url<] I would advise AMD to keep the QA in house if they do end up outsourcing the design.

      • Airmantharp
      • 7 years ago

      AMD’s in-house chipsets before the ATI acquisition weren’t any better, though, while ATI’s were pretty good. AMD must have seriously nuked their chipset engineering team to be farming out to a third-rate chipset provider.

      SiS, VIA, and hell Realtek are better.

        • just brew it!
        • 7 years ago

        What has SiS done lately? VIA seems to have retreated to the SFF market; yes, their USB, Audio, and LAN chips were better than Realtek’s, but not *enough* better to save them in a cost-driven market. Realtek’s “lowest common denominator” solutions finally seem to have gotten good enough that most people don’t seem to notice.

        I would actually put ASMedia somewhere between Realtek and VIA, TBH.

          • Airmantharp
          • 7 years ago

          Nothing, so far as I know, but their last effort was pretty solid, and they did license the IP for the Xbox 360 to Microsoft for what that’s worth (or not worth, as it may be).

          I’m really just pointing out that they had the potential to develop their IP into something useful had they not been forced out of the discrete desktop chipset business along with VIA and Nvidia.

    • jackbomb
    • 7 years ago

    “There’s no chance of a wonky third-party memory controller crippling a new CPU.”

    Good. I’d much rather have AMD’s wonky integrated memory controller cripple their new CPU.

    To add something worth reading to my snarky comment, I never really had trouble with the memory performance of NVIDIA, ATI, and even VIA chipsets. The only real offender was SiS, but approximately no one used them.

      • Airmantharp
      • 7 years ago

      The last one they put to market was actually outstanding, both in memory controller performance and in overclocking performance, but I agree that that wasn’t the norm for them.

    • Deanjo
    • 7 years ago

    They should have never stopped 3rd parties from making chipsets for their processors. They got greedy, cut everyone off, and now find that they do not have the resources to handle a task.

    • NeelyCam
    • 7 years ago

    Makes sense. AMD doesn’t make much money off chipsets – money comes in from CPUs and GPUs.

    And although others are pointing out that AMD’s USB and SATA controllers aren’t quite as fast as Intel’s, it doesn’t really matter – they are fast enough. Differences in real life only matter to benchmark geeks like us.

      • Deanjo
      • 7 years ago

      [quote<]AMD doesn't make much money [s<]off chipset[/s<].[/quote<] FTFY

      • Airmantharp
      • 7 years ago

      Intel probably doesn’t make much money off of their chipsets either, though likely more than AMD.

    • just brew it!
    • 7 years ago

    At first glance this may seem to be a bit of “tail wagging the dog” syndrome, but I’m going to contradict the sentiments others have expressed so far, and say that this probably makes sense.

    At this point, AMD needs to be lean and mean if they are to survive. Leveraging ASMedia’s IP for peripheral interfaces allows them to focus on their core strengths. Let’s face it, AMD/ATI chipsets have always been pedestrian at best, so I don’t think they are “trading down” here.

    And FWIW the ASMedia USB3 controller on my M5A97 seems to work reasonably well in Linux, so it isn’t quite the Open Source apocalypse that some people seem to be anticipating either.

      • nanoflower
      • 7 years ago

      I agree with what you’ve said. I think this also says something about AMD’s future or at least what they see as their future. It doesn’t include things that don’t provide significant revenue which includes chipsets. As more functions are brought on-board the CPU/GPU/APU there’s less for a chipset to do and so the potential revenue is bound to decrease. I think AMD sees this so why not farm the work out to a company dedicated to this type of work?

      • Airmantharp
      • 7 years ago

      I think that the biggest concern is with ASUS’ ability to write proper drivers and accompanying software, you know, like ever.

        • just brew it!
        • 7 years ago

        Yes, that’s a concern. Although I might strike the “and accompanying software” bit, since that really could be someone else’s responsibility (assuming the drivers are up to snuff).

        I’d also be perfectly happy with them publishing register-level hardware specs (or their reference driver source) in a timely manner, and letting the Open Source / hacker community write or improve their drivers for them. It’s a win-win — they save costs on driver development, and users get functional drivers. I’m not just talking Linux here either; the same development model could work for Windows drivers IMO.

          • Airmantharp
          • 7 years ago

          I totally agree- I put it in there because ASUS’ software is usually as buggy or buggier than their drives (see all Xonars, for example).

          But if they’re going from addon crap to full-fledged OEM production, ostensibly showing up in the Opteron and other professional lines as well, I’d be willing to bet for a more positive outcome- and since nothing about chipset controllers is terribly hard or proprietary any more, it would make sense for them to provide a functional framework to the community so that it can be refined and distributed en mass.

            • just brew it!
            • 7 years ago

            I’ve been waffling on whether to get a Xonar for my Linux box… need to do a little more research on whether the Open Source drivers are in reasonable shape. They’re using an off-the-shelf codec, so there’s hope.

            • f0d
            • 7 years ago

            whats wrong with xonar drivers? never had any issue with mine
            i am using it for windows and not linux though so maybe its a linux thing? not sure, either way never had a single issue with mine its just been set and forget

            • Airmantharp
            • 7 years ago

            Compare Xonar’s drivers and software to Creative’s for the last three or four years- it’s a night and day difference.

            Particularly, the contortions necessary to use some of the Xonar cards’ special functions require extensive trips to Google, whereas with Creative (and say Realtek) the damn switch just works as intended.

            • f0d
            • 7 years ago

            i have had the exact opposite experience

            never had to “google” anything with my xonar – everything i want it to do it just does, its been set and forget since i got it years ago

            yet when i had a soundblaster the driver was bloated with junk and i had all sorts of driver issues and problems and things not working

            • Starfalcon
            • 7 years ago

            New sound blaster Z is a very nice card and works just fine with the basic driver…you do not even have to install anything besides that for it to work. Might want to check it out when you upgrade from you old PCI xonar.

      • BaronMatrix
      • 7 years ago

      ATi had the best chipset right before Intel kicked them to the curb…

        • NeelyCam
        • 7 years ago

        [quote<]NVidia had the best chipset right before Intel kicked them to the curb...[/quote<] ftfy

        • chuckula
        • 7 years ago

        Nvidia had the best chipsets* until AMD kicked them to the curb. Oh, and Nvidia should sue AMD for the monopoly practice of illegally bundling the memory controller with the CPU that killed Nvidia’s right to make memory controllers! Oh the horror!

        * Yeah, I know I’m trying to keep a straight face, but messing with the ol’ Styrofoam Baron is just too much of an opportunity.

          • Airmantharp
          • 7 years ago

          To my memory, ATi’s chipsets (and the motherboards based on them by top-tier manufacturers) were quite effective with the last non-IMC Intel CPUs, but they weren’t Intel.

          The biggest problem for them was that for a while there, there was very little objective reason to buy Intel anything- you paid more to get a slower system with a higher TDP. Kind of like AMD today, except that they couldn’t possibly charge as much and move any product like Intel did.

      • JustAnEngineer
      • 7 years ago

      [quote=”just brew it!”<] AMD/ATI chipsets have always been pedestrian at best, so I don't think they are "trading down" here. [/quote<] I remember my AMD 760D chipset as slowish but bulletproof. The faster VIA chipsets available shortly thereafter were bug-ridden.

        • just brew it!
        • 7 years ago

        Yup, they were generally solid but slow. Early ones (the 750 “Irongate” IIRC) had dodgy AGP implementations though.

        I actually considered putting my old Athlon MP dually back into service as a file server a few years back, until I measured the idle power consumption. The motherboard, RAM, and CPUs promptly went back into storage.

          • JustAnEngineer
          • 7 years ago

          Let’s review that…

          · You had some old top-of-the-heap hardware lying around that was still functional enough years later that it might serve in a low-end duty.

          · You evaluated the cost of the power consumption for running the old equipment and realized that it was completely unreasonable to operate such inefficient equipment in the current decade.

          I was right there with you this far, then…

          · You decided to put the obsolete hardware back into storage until after Obama’s carbon tax triples the price of electricity in parts of this country.

          It seems to me that you had made the rational evaluation that the old equipment was no longer useful. That’s when you should have gotten rid of it (or donated it or put it up on e-bay).

            • just brew it!
            • 7 years ago

            I may bring it to the swap-fest this year. Starfalcon would probably still take it, irrespective of the power usage. 😉

            • Starfalcon
            • 7 years ago

            Yup 🙂

        • srg86
        • 7 years ago

        My AMD 760 chipset board (an ABIT) developed a memory controller fault (at 512MB) after 1 year. Other than that it was fine. I replaced it with an nForce 2 Ultra 400 (that was bullet proof in my experience).

    • Prestige Worldwide
    • 7 years ago

    I don’t see how outsourcing R&D could ever be a good idea.

    Here in Quebec, the government gives sizeable tax credits for R&D, and I doubt you would qualify if you outsourced such activities. It would be much more worthwhile to do it internally.

    Then again, AMD is not based here so perhaps my comment is completely irrelevant.

      • xeridea
      • 7 years ago

      If it is outsourced to a company that specialized in something, their R&D costs will be potentially substantially lower, making it a lot cheaper in the end. Also, being able to focus more on core strengths is a big plus.

    • maxxcool
    • 7 years ago

    Oh yes that is exactly what the “future high performance” cpu needs a third party chipset … this takes me back to via chipsets, bluescreens, usb support nightmares.

      • who_me
      • 7 years ago

      The nForce chipsets were rather good though…

        • just brew it!
        • 7 years ago

        If you ignore the nForce 2…

        Edit: And their horrific NIC drivers.

          • Airmantharp
          • 7 years ago

          And 3… they finally started to get their crap together with the 4/5/6/7 generation, which were all basically the same thing, but then politics killed them.

            • Deanjo
            • 7 years ago

            Had both the Nforce 2 (Abit NF7-s) and 3 (Asus K8N-E Deluxe) and they never gave me issues.

            • Airmantharp
            • 7 years ago

            I’ll say that you got lucky :).

            Had a Gigabyte NF3-type with socket 754, and it was rough for the first year or so.

            • JustAnEngineer
            • 7 years ago

            NVidia’s Active Armor was horrible. It invisibly (or not-so-invisibly) ate parts of your packets.

            • Deanjo
            • 7 years ago

            Fortunately you didn’t have to even install Active Armor (which was the cure).

            • Airmantharp
            • 7 years ago

            I do recall that the common advice was to not use that feature- but otherwise, their accelerated Ethernet controllers were up there with Intel’s.

            • JustAnEngineer
            • 7 years ago

            It wasn’t just the software. I had to put a piece of tape over the NVidia ethernet port when I handed that motherboard down to my folks so that they would use the [b<]much[/b<] more reliable Marvell port.

            • Airmantharp
            • 7 years ago

            Well then. If Marvell was the definition of ‘reliable’ in that era, then I’ll have to agree with you- Nvidia’s solution could have been nothing less than horrific. Had nothing but a stream of BSODs from the Marvell solution until they released an actually stable driver over a year after I’d bought the board- and that was nForce3.

            • Deanjo
            • 7 years ago

            A lot of “chipset” issues back then was more from crappy BIOS implementations. I have to say that the K8N-E deluxe was one of the rare trouble free Asus boards that I had and Abit was Abit, masters of offering solid boards without “proprietary whiz bang” features with a BIOS engineering crew that has yet to be matched.

            Abit:
            “Oh you found an issue on Linux? We will have a fixed beta version for you to test this week.”

            Everyone else (except Tyan)
            “Oh you found an issue on Linux? Please install windows. Issue resolved. Closing”

    • derFunkenstein
    • 7 years ago

    If ASMedia’s USB 3.0 controllers (compared to Intel’s built-in USB3.0) are any indication…then nothankyou.jpg

      • Airmantharp
      • 7 years ago

      You know, I agree with you on a pure performance perspective, but the (second/third gen?) USB3 controller on my ASRock Z68 board actually works incredibly well outside of the fact that you cannot boot off of it.

        • derFunkenstein
        • 7 years ago

        For things that normal people do, it’s fine. My system with the 3570K is on a Gigabyte Z68 board with an ASMedia controller and it saturates my external hard drive. I do have to be careful about what I plug into it, though – my Mbox doesn’t like it, sputtering and stuttering with audio nonstop without crazy high latencies (where the drivers have been updated for USB 3.0 and it works fine on my wife’s new Gigabyte H87 board, when I tested it just for kicks). Might be Avid’s fault, but I’d just as soon stick with Intel’s stuff where possible. I just don’t want to buy a Z77 board or a whole new setup with a new CPU.

    • Farting Bob
    • 7 years ago

    Jesus, how bad is AMD doing that they have to outsource their chipset designs? Im guessing both of the engineers left at AMD are overworked.

      • Airmantharp
      • 7 years ago

      It’s probably not that they’re doing so bad, but rather that since they have to outsource chipset fabrication anyway, unlike Intel, they might as well let someone else apply the elbow-grease to keep the IP up to date.

    • Sahrin
    • 7 years ago

    Terrible move for AMD; if they keep outsourcing system-level components they’ll just fall farther behind Intel in the power consumption race.

      • Deanjo
      • 7 years ago

      Keep in mind that when AMD was outsourcing their system level components, the offerings were competitive. Had they continued outsourcing, you would more than likely see features implemented far faster and on updated processes. They have been stuck on 55/65nm for eons.

      • nafhan
      • 7 years ago

      Not true at all. Most of the “system level components” are on die at this point. Look how simple some of the Kabini MB’s are.

        • Sahrin
        • 7 years ago

        And yet the power spec for the PCH’s is like 5W.

    • PerroLito
    • 7 years ago

    I don’t see how this could make sense… AMD already has to design the i/o hub integrated into SoCs. Are the discrete so different that they cannot be developed alongside?

    • Ryu Connor
    • 7 years ago

    I’d note some of the worse behavior from the early half of 2000 3rd party chipsets was not – typically – from the memory controller. It was a problem with the very I/O hub you reference: poor AGP support, poor USB support, and busted PCI controllers.

      • Deanjo
      • 7 years ago

      [quote<] It was a problem with the very I/O hub you reference: poor AGP support, poor USB support, and busted PCI controllers.[/quote<] That isn't any different then what AMD offers now or in the past. Subpar USB and ide/sata performance, their AGP performance with Irongate had huge issues limiting it to 1x and when it came to PCI, for the longest time, many card manufacturers would only guarantee their product to work on an intel chipset. That being said, and knowing ASMedia's track record and AMD's disappearance from non-Windows OS development, I would say that if they do this don't expect a trouble free experience in a non-Windows OS.

        • derFunkenstein
        • 7 years ago

        Did you use a chipset with, for example, ALi Aladdin 5 on a Super 7 motherboard with a K6-2? What we have today from AMD, while still behind Intel, is a far cry beyond that nonsense. [url=http://ee.gigabyte.com/products/page/mb/ga-5ax_52/<]This[/url<] was my first motherboard in my first hand-built PC and I thought I was doing something wrong. PCI video cards like the AMD Xpert 98 I started with were fine, but as soon as I tried to use the AGP slot with a Voodoo 3 3000 the system was awful. Buggy, constant reboots, blue screens...no review I ever read said anything about that, but as it turns out no review I read used anything other than Pentium IIs, which ran on Intel's chipsets. It took us about 15 years to get as close as they are. lol

          • Deanjo
          • 7 years ago

          I actually ran a K6-3 on an ALi chipset for years (first as a desktop then migrated to a mail server that was reliable until the fan died). Those issues you mention could also be a result of multiple things such as a poor bios implementation, power supply issues etc and not necessarily due to chipset errata.

            • just brew it!
            • 7 years ago

            Agreed. Later ALi chipsets in the Super 7 era were decent. One of the most reliable systems I ever had was a K6-III+ on an ALi-based Micronics motherboard, with a tweaked BIOS (to recognize the III+ mobile CPU). It is part of Starfalcon’s collection now. 😉

Pin It on Pinterest

Share This