Controller round-up seeks fastest USB 3.0 solution

USB 3.0 was a long time coming, and it finally seems to be taking hold. Blue SuperSpeed ports can be found on a range of devices these days, from enthusiast-oriented motherboards to ultraportable laptops and netbooks. The selection of controller silicon has expanded, as well. While NEC’s two-port D720200 dominated the scene for more than a year, it now has competition from AMD, ASMedia, Etron, Fresco Logic, and VIA. So, which is fastest?

VR-Zone has put together an interesting roundup comparing the performance all the USB 3.0 host controllers on the market, including a preview of NEC’s upcoming D720201, which doubles its predecessor’s port count. As it turns out, USB 3.0 speeds can vary quite a bit from one solution to the next. The site came away most impressed with controllers from ASMedia, Etron, and NEC. It was also disappointed by the USB 3.0 controller integrated in AMD’s A75 chipset, although to be fair, that config used a different motherboard than the other ones.

The lack of real-world file copy tests makes it difficult to draw too many conclusions about the results, but VR-Zone does have a few interesting notes on compatibility. Not all USB 3.0 cables are created equal, the site says, and it claims to have run into more than a few compatibility issues that impacted performance. I haven’t encountered any problems with the USB 3.0 hardware we’ve had in the labs, but we’ve been drawing from a much smaller pool of devices and controllers.

Comments closed
    • LoneWolf15
    • 8 years ago

    I wouldn’t have seen this article without TR linking to it.

    I found it really interesting, and wanted to say thanks.

    P.S. Mine’s the one with the Etron controllers.

    • Bensam123
    • 8 years ago

    Wish TR would’ve done something on this. I enjoy niche article and this is something quite relevant.

    Adding to this they really only tested bandwidth, but they didn’t test latency at all(!). I think this is a very bad omission as this can vary a lot and for some people that matters, like when using a input device.

    • ronch
    • 8 years ago

    I think this just shows that USB 3:0 is still in its infancy.

    • jensend
    • 8 years ago

    Lotta whining in here.

    A couple of people have already pointed out that the difference between fastest and slowest here is pretty insignificant compared to the difference between USB2 and USB3 (<20% vs 1000%).

    More importantly, the only hardware currently out there where you will see any difference between these controllers is with SSDs connected via USB3; none of these controllers will be a real bottleneck for people using USB3 external magnetic hard drives or thumb drives.

      • NeelyCam
      • 8 years ago

      Yeah, but I can’t let logic get in way of good trollin’…

      (the thing is, if someone had responded with that last sentence of yours, the troll would’ve been over)

        • jensend
        • 8 years ago

        Thumbed this post back up to zero because at least you had the decency to admit it.

          • NeelyCam
          • 8 years ago

          Thanks for the effort, but my guess is that the zero won’t last..

      • Krogoth
      • 8 years ago

      Yep, it is a big first-world problem that only exists for a tiny niche.

      Why would you use external SSDs? It doesn’t make any sense.

      SSDs are build for fast access data storage. It is not something you would do in an external apparatus.

    • Johnny5
    • 8 years ago

    “Not all USB 3.0 cables are created equal, the site says, and it claims to have run into more than a few compatibility issues that impacted performance.”

    Honestly, who screws up a cable? Don’t lie this testing was funded by Monster, wasn’t it?

    • Krogoth
    • 8 years ago

    Geez, people have short memories.

    USB 2.0 went through the same issues when it was new.

      • NeelyCam
      • 8 years ago

      You mean AMD went through the same issues when USB2.0 was new?

      What I remember is that the USB2.0 drivers that were supposed to come with VIA’s AMD board never worked. After struggling with it for months, I gave up, bought an NEC USB2.0 card and everything started working.

      That was the last AMD-based desktop I had. I’ve had Intel since then, and everything has just worked (with the exception of drivers for one or two NVidia crap-hics cards)

        • Musafir_86
        • 8 years ago

        -Err, “VIA’s AMD board”? Then it’s VIA’s fault, not AMD…

          • NeelyCam
          • 8 years ago

          AMD’s fault. Should’ve picked a better partner, or offer chipsets themselves.

        • Krogoth
        • 8 years ago

        The entire industry had issues with the USB 2.0 transtion. It wasn’t until second and third generation controllers that you had most of the issues resolved.

      • Bensam123
      • 8 years ago

      USB 3 has been around for what, three years…?

      That’s hardly new.

    • indeego
    • 8 years ago

    Worse than I imagined. No wonder Intel has ignored this to date.

      • BobbinThreadbare
      • 8 years ago

      This is still 3-4x the speed of USB2 right? What about it is so bad?

        • Goty
        • 8 years ago

        Nothing, really; people just aren’t happy unless they’re complaining about something.

          • indeego
          • 8 years ago

          There were compatibility issues.
          There were speed issues all across the board that depended on cables or setup.
          There are license issues that will fragment the market.
          One of the largest semi’s in the world showed the worst performance.

          Yeah, so it just seems like a crapshoot of mediocrity and a “wait and see if something better” or “needs more fine-tuning” comes out.

          My NEC USB3 has been mostly idle @ home. I’m afraid that I’m using cheap network storage/transfer as it’s universally accessible from anywhere. Printing is dying.
          eSATA is far more prevalent than USB3 and fills transfer needs.

          I’d invest in fiber/10 Gbps before going USB3 at this point.

        • NeelyCam
        • 8 years ago

        Yeah, except you get 4-5x the speed of USB2 on an Intel board.

    • link626
    • 8 years ago

    damn. amd’s usb3 is terrible.

    I wonder if drivers can fix it.

      • NeelyCam
      • 8 years ago

      “But… but… It’s a [b<][u<]native[/u<][/b<] USB3.0 controller!! It's far superior than what's available on Intel's platforms!!" Yeah... AMD fanbois were embarrased again. When will it end? I personally can't wait for the actual Bulldozer benchmarks. That'll be a boatload of fun!

        • BobbinThreadbare
        • 8 years ago

        I don’t think anyone said AMD had the best USB3 implementation.

        The important part is that motherboard makers don’t have to include a 3rd party chip, which costs money, to implement it. So the cheap boards, which are going to sell the largest numbers and thus influence adoption rate the most will have USB3 with AMD, but not Intel.

          • NeelyCam
          • 8 years ago

          Many folks were saying how this AMD’s “native” USB3.0 controller is so much better than whatever comes in Intel boards, because it’s “native”. Well, they were [u<]wrong[/u<]. An Intel platform provides a better USB3.0 experience than AMD's Llano platform, as shown by this test. And price? The cheapest A75 board ([b<]$75[/b<], including shipping): [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16813157261[/url<] An H61 board with USB3.0 ([b<]$65[/b<], including shipping): [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16813138311[/url<] So, the AMD motherboard is more expensive.

        • Goty
        • 8 years ago

        AMD’s USB 3.0 controller is still faster than Intel’s since, you know, Intel doesn’t even have one.

          • NeelyCam
          • 8 years ago

          A pathetic argument that doesn’t address any of my comments. Please try again.

            • Goty
            • 8 years ago

            It’s funny because you just get so worked up about AMD, especially when they’ve got the lead on Intel with something.

            • NeelyCam
            • 8 years ago

            I don’t get worked up because I’m not a fanboi. I’m just pointing out fallacies in AMD fanbois’ arguments.

            And regarding the “lead” (I assume you mean that AMD has a USB3.0 design out there before Intel, and that’s a “lead”), it’s a weak one considering how awful their implementation is compared to others.

            In other words, AMD platform is crippled with this “native” USB3.0, whereas Intel platform comes with a working solution. AMD fail.

            • Goty
            • 8 years ago

            [quote<]I don't get worked up because I'm not a fanboi.[/quote<] I'm sorry, you're going to have to retype your post without this line in because I can't get past it without bursting into fits of laughter.

            • NeelyCam
            • 8 years ago

            Well, I’m happy that my argument brings you joy and happiness. I still see no counterargument whatsoever…

            Or do you think laughing makes you win arguments? Like the last one about Brazos chips using more power than Atom chips when idling?

            • Goty
            • 8 years ago

            I don’t see what kind of argument can arise from the fact that your favorite company doesn’t have a USB 3.0 host controller while AMD does. It seems pretty cut and dry to me. If you really want some competition, though, we could always compare AMD’s fastest implementation to Intel’s. That should be fair, right?

            As for the Brazos vs Atom power consumption discussion, the argument was about TDPs vs power consumption, not about idle power consumption. The fact that you’ve had to twist it around so badly to fit your world view should say something about the actual result of that discussion (i.e. that I was right and proved it with hard numbers).

            Sorry Neely, you lose again. 😉

            • NeelyCam
            • 8 years ago

            My original point was that despite all the hype about AMD’s ‘native’ USB3.0 vs. Intel’s supposedly crippled third-party solutions, AMD’s solution ended up being the worst. People were whining about how Intel is trying to kill USB3.0, yet Intel boards offer the fastest USB3.0 speeds. That’s the bottom line here.

            Part of the Brazos/Atom argument, if you remember, was the discussion of what was the Atom/Brazos idle number… and it came from your less-than-hard numbers and your completely arbitrary claim that you believed (without proof) that AMD is over-reporting their TDP while Intel is under-reporting theirs. Somehow in your fantasy land you really believed Intel would massively underestimate their TDP (like they wouldn’t be called out on that by all the netbook/tablet manufacturers…)

            I tried to point out how stupid this was, highlighting the very small power consumption difference between Atom system idling and Atom system under load. Then you made an even more outrageous claim that Atom somehow idles at crazy high levels (close to the reported TDP), and goes over TDP with that 2-3W increase when loaded…. like that idle number wouldn’t kill every single Atom netbook’s idle battery life. And you were actually convinced that all this was true.

            You didn’t provide any actual hard numbers to prove anything – it was all your pixie dust and fairytales. Now, if you’re interested, we can reopen that argument that you completely lost, but you will have to send me actual links to stuff with those “hard numbers”, so I can tear apart your argument once again.

            And I’ll be here, because I can’t go to sleep when somebody’s wrong on the internets.

            • Goty
            • 8 years ago

            What’s this about not getting worked up again? 😉

            Anyhow, you should go back and read that thread again, because you’ve constructed some strange argument on my behalf that was never made.

            As for the USB 3.0 implementations, there are quite a few AMD boards with third party controllers (I know of boards that use ASMedia, Xfast and NEC chips off the top of my head), so, whoops, Intel doesn’t have a lead there either.

            Keep grasping for straws, Neely; it’s fun to watch. Now, I’m off to bed because I CAN sleep when you’re wrong (and quite soundly in fact).

            • NeelyCam
            • 8 years ago

            [quote<]Anyhow, you should go back and read that thread again, because you've constructed some strange argument on my behalf that was never made.[/quote<] I just explained clearly where you were wrong. Please point to me where I misrepresented your comments. Unless you can, then I assume we both agree that Atom chips consume less power than Brazos both in idle and under load. [quote<]As for the USB 3.0 implementations, there are quite a few AMD boards with third party controllers (I know of boards that use ASMedia, Xfast and NEC chips off the top of my head), so, whoops, Intel doesn't have a lead there either.[/quote<] Point me to one that's cheaper than Intel's alternatives, please? AMD is charging for a crippled USB3.0, and partners are forced to add better third-party solutions. Customers lose.

            • jensend
            • 8 years ago

            Uh- almost all the AMD boards with 3rd party controllers are AM3/AM3+ boards, i.e. they don’t have any native USB3 yet; the only exceptions are high-end FM1 boards trying to load a dozen USB3 ports so they need both controllers to hit their goal. (I don’t understand the point of high-end FM1 boards, and I don’t know what use cases require somebody to connect >4 USB3 devices at once, but mobo makers will slap just about anything in existence onto their higher-end boards so as to have more talking points than the competition.)

            Nobody would bother putting a 3rd-party controller on an A75 mobo just to “replace” AMD’s “crippled USB3.0” because, as we’ve already discussed, A75 USB3 performance is just fine for the use cases that matter. If you’re chaining a dozen SSDs to one computer via USB3, the first thing that should be done is somebody needs to take away your credit card; but failing that, you might as well be using a higher-end system than Llano as the hub of your insanity.

            I thought you were done trolling this thread and had moved on to greener pastures. Just can’t pull yourself away?

            • NeelyCam
            • 8 years ago

            [quote<]I thought you were done trolling this thread and had moved on to greener pastures. Just can't pull yourself away?[/quote<] I was gonna, but the Brazos/Atom argument with Goty pulled me back. Don't worry - as soon as this falls off the Front Page discussions list, I won't bother anymore

            • Goty
            • 8 years ago

            [quote<]Please point to me where I misrepresented your comments.[/quote<] Go right back to the beginning. Good, now read the entire thread again. Let me know where you get lost. [quote<]Point me to one that's cheaper than Intel's alternatives, please?[/quote<] Ah, so now it's down to, "Well, Intel's boards are cheaper!" Sorry, but if we actually look at boards with comparable features (i.e. H61 with USB 3.0, SATA 6Gbps and HDMI out compared to A75 with the same; I'll throw you a bone and leave out the fact that the A75 board has 4 USB3.0 ports vs 2 on the H61), we get a maximum price advantage of $5 in favor of the H61 board (both ASRock boards, too). Given the fact that H61 boards have been on the market for roughly a quarter longer than the A75 boards, I'll be generous and call the prices comparable. Sorry Neely, you'll have to find another invalid argument.

            • NeelyCam
            • 8 years ago

            If you insist continuing with that line of commentary (“let me know where you get lost” and “you’ll have to find another invalid argument”), fine – let’s continue, and I’ll stop being polite.

            Sorry, Goty, but you are wrong. You really should re-read the thread again, but you should have an engineer with you when you do that – it looks like you are one of those physicists who doesn’t actually understand the practical applications. The best part is that your physicist arrogance makes you incapable of actually realizing this, setting you up for more embarrassment.

            Let me help you get started, though. Like I already said, you somehow got this idea that Brazos TDP was “overestimated” and Atom TDP was “underestimated” (and here it seems that you don’t understand what TDP means, or how it is used in practical computer designs).

            JumpingJack politely explained to you that your claim was without basis, providing you with a link that clearly disproved your Brazos claim, and even doing the math for you. He even pointed out the impact of PSU efficiency on the result (something you showed you don’t understand in a different Zacate thread). He also mentioned the VRM inefficiency, but since you don’t understand PSU inefficiency, VRM probably doesn’t even register in your radar. For Atom, he listed numbers he measured himselt (even explaining how the experiment was done so you could repeated it), showing that Atom was idling very low, and under load was below TDP, thus disproving your claim.

            Your response was random arrogant bs, basically claiming that his results were wrong, and saying that he needs to show better proof that his numbers are right (all the while you are showing absolutely zero proof supporting your wild claim about misreported TDPs). The only supporting “evidence” you had was saying (without references) that “in similarly configured systems, Brazos [b<]as a platform[/b<] has significantly lower power draw than Intel's Pine Trail platform. This is not disputable." Of course, you clearly missed JumpingJack's list of items that also consume system power, making the usage of system power as a very inaccurate technique for estimating CPU power.... Furthermore, it seemed that you forgot that your original argument was about CPU power (not "platform" or "system" power). Out of frustration with your lack of focus, understanding, intelligence or whatever, JumpingJack dropped out of the discussion. To bring the discussion back to CPU power, I simply stated that "CPU TDP != platform power." Your response: "Of course it doesn't, but it's certainly a [b<]large part of it.[/b<]" (another invalid argument that, while being true for some systems, is not true for all and is clearly not true for a current Atom system - the exact point JumpingJack tried to point out to you... twice). You went on to say: "Also, using my previous assumption that [b<]chipset TDPs are roughly the same[/b<] ..." (another invalid claim that you yourself later disproved) "...and that the [b<]platforms are similarly configured[/b<],..." (another invalid claim - the cheap Atom platforms are using cheap and power-hungry components, and the motherboard designs between the Brazos and Atom systems in question are completely different) "... my statement is entirely truthful." I responded by pointing out that your assumption was wrong, repeating some of the information JumpingJack was trying to give you in a different way, hoping that you would understand why you are wrong. You went on to list TDPs of the two chipsets, and summing them with the CPU TDPs, somehow trying to use this to validate your claims. You chose to ignore the fact that the test that was linked was using DDR2 for Atom while using more power-efficient DDR3 for the Zacate system, didn't take into account the effect of VRMs (cheap ones tend to be less efficient), and glossed over the other motherboard differences. In all your "support evidence" showing that Atom consumes more power than Brazos, you chose to use the most power-hungry Atom available over a year ago (D525), while ignoring the far more efficient N550 that came out later (still well before Brazos) and had more power efficiency features. As I already mentioned, you chose to ignore JumpingJack's comments about N330 (another more efficient Atom). So, overall, you gave weak supporting evidence for your claims throughout the thread while ignoring strong opposing evidence with unproven and unsupported claims. You completely lost the argument, but still somehow held onto the belief that you "won"... possibly because you didn't have the knowledge to understand the counter-arguments or, alternatively, because you just simply were being intellectually dishonest in the argument. Either way, Jack and I gave up continuing the thread because it was like talking to a brick wall... I only decided to finish it now because you were being a ass about it, and arrogant jerks are a pet peeve of mine.

            • Goty
            • 8 years ago

            Brazos (Zacate + Hudson M1) TDP > D525 + NM10 TDP. This is fact, so let’s start from there.

            Now, how about the motherboards? Well, TR’s own testing used two mini-ITX motherboards for testing, and the board used for the Brazos setup was arguably the more fully featured board, but let’s call power consumption due to the motherboards equal even if the power consumption between the two likely favored the stripped-down Atom setup. So, that’s not going to be a contributing factor.

            Well, how about the PSU and efficiency ratings? Well, TR used two different PSUs on both setups and the power draw of the two is quite close, so even for the less-efficient PSU used, the difference should be small and well within the margin of error. Even then, let’s ONLY consider the brick PSU just so you have one less thing to whine about.

            So then, we’re down to the actual power consumption numbers. What happens when putting the system under load? *GASP* The E-350 system consumes less power! Now, you’re going to cite the fact that the difference from idle to load for the Atom system is much lower than that of the Brazos system, but that was never in question. TDP is essentially a measure of the heat generated (and therefore power consumed) by a chip under normal load conditions, meaning that idle power consumption has nothing to do with measures of TDP, and the idle-load delta is also not a measure of TDP. It comes down to real power consumption under load, and I can show you numerous examples where, in otherwise comparable configurations, Brazos systems consume less power than comparable Atom systems (even when using the NM10 chipset). If the systems are comparably equipped, it is reasonable to assume that load power measured for the system is an accurate indicator of platform (CPU + chipset) power consumption (and therefore of relative TDPs) and it is clear that the TDPs stated by AMD and Intel for their respective systems are not directly comparable, to the tune either Intel tends to provide a more conservative TDP or AMD tends to inflate their TDP.

            Now, I know you have an undying love for Intel and their products, but you’re simply wrong here, and that’s OK! It’s been known for a long time now that AMD and Intel don’t measure TDP the same way. As a matter of fact, Intel has a history of understating their TDPs. AMD started using their “ACP” number in the past just so they would have a number comparable to that of Intel. It’s nothing for you to get your panties in a bunch for.

            • NeelyCam
            • 8 years ago

            Again: no. You keep repeating the same wrong assumptions that I and other people have already proven wrong, and by using wrong assumptions you get wrong conclusions, including the conclusion that I’m wrong.

            Your blind belief in your own assumptions and conclusions in the face of mounting evidence to the contrary makes you the worst kind of a scientist – one that refuses to admit (s)he can be wrong. I really hope that you’re more scientifically honest in your day job.

            This is a waste of my time.

            • Goty
            • 8 years ago

            The only assumption made is that the platforms are similarly equipped and therefore that the peripherals draw roughly the same amount of power. Please indicate how this assumption is flawed, if you can. As for your “mounting evidence”, please note that I’m the only party citing actual numbers. So far your arguments have amounted to “Nope, you’re wrong because I don’t believe your argument.” Sorry, champ, but that doesn’t float.

            • NeelyCam
            • 8 years ago

            Go back to the original discussion (like I asked you to) to find the numbers. You lost that argument, and you’ve lost this one because you’re unable to dispute those numbers that clearly disprove your points.

            Until you go back there, please refrain from making additional invalid arguments.

            • Goty
            • 8 years ago

            The numbers I provided came from TRs own testing and Intel’s and AMD’s own documentation. If you want to dispute the numbers provided by those parties, well, there’s not much I can say for you.

            • NeelyCam
            • 8 years ago

            The TR numbers were never disputed – they are solid. I was merely pointing out that you were using the numbers wrong, jumping into wrong conclusions using your invalid assumptions. JumpingJack provided additional numbers showing exactly [i<]how[/i<] your conclusions were wrong, helping you make the right conclusion (which you refused to do... probably because it would mean you were wrong).

    • dashbarron
    • 8 years ago

    Progress is progress, slow but steady.

    • Mr Bill
    • 8 years ago

    What is the point of USB3 with its funky double connector cable, when eSATA is just as fast? Are USB controllers cheaper than SATA controllers? Or, is it that USB can supply power but not eSATA?

      • LoneWolf15
      • 8 years ago

      Power is a big one. As is compatibility with legacy USB devices people own.

Pin It on Pinterest

Share This