Comments Locked

63 Comments

Back to Article

  • zeeBomb - Friday, November 27, 2015 - link

    Whoa!!!
  • dsumanik - Thursday, December 17, 2015 - link

    AT Theres a typo in the headline, here's the correction:

    The ASRock Z170 Extreme7+ Review: When You Don't Test The Headline Feature Of The Motherboard.

    Otherwise knows as: We didn't do our job, but trust us it's awesome and buy it anyways.

    Seriously AT?

    *Facepalm
  • daos - Friday, December 18, 2015 - link

    I completely agree! Hasn't been the same since Anand left
  • pedjache - Friday, November 27, 2015 - link

    In one of AT's previous articles, it was stated that the gtx770 used in tests gave odd results in Shadow of Mordor while the network connection was on, and said the matter will be looked into and reported. Anything on that?
    (more) on topic - nice review, wonderful motherboard, and kudos for praising them engineers, I happen to know they love and can't get enough of it.
  • Ian Cutress - Friday, November 27, 2015 - link

    Still looking into it.
  • flameyyy - Saturday, November 28, 2015 - link

    do the DPC latency issues have anything to do with the bugged intel networking drivers? https://communities.intel.com/thread/54594
  • evilspoons - Friday, November 27, 2015 - link

    The Board Features/Visual Inspection page appears to be blank except for an introductory paragraph, as of 10:32 AM mountain time.
  • evilspoons - Friday, November 27, 2015 - link

    Also, it's a shame the headlining feature (triple M2 x4 in RAID) wasn't actually benchmarked. I would have loved to see 3x Samsung 950 Pro drives jammed in this sucker.
  • eddieobscurant - Friday, November 27, 2015 - link

    They will all share the same pci 3.0 x4 lanes, so there is a cap at 3.2 gb/s for the raid array. The thessdreview has a review of this combination.
  • Flunk - Friday, November 27, 2015 - link

    That's GB/s, which is 8x as much.
  • phorgan1 - Tuesday, December 22, 2015 - link

    And it is 3.93 GB/s for 4 lanes of DMI 3.0
  • nunya112 - Friday, November 27, 2015 - link

    no. they will have their own. if there is no video card installed or if they only set the video card to 4X
    that would then dedicate all the lanes
    4 will be in the DMI links, then depending if they have video card or not there will be 12 lanes free. meaning it would go full speed at 4X EACH = 12 lanes then the 4 lanes for the video card. or even none if using onboard
  • phoenix_rizzen - Sunday, November 29, 2015 - link

    The video card slots get their lanes direct from the CPU.

    The M.2 slots get their lanes from the PCH, along with the x1 and x4 slots.

    Separate lanes; not shared.

    You can run SLI video setup using 16 lanes from the CPU, and still use 12 lanes for M.2.
  • extide - Monday, November 30, 2015 - link

    BUT evilspoons/eddieobscurant are correct. The DMI link is EQUIVALENT to a PCIe 3.0 x4 link. Regardless of the video card setup, the 3 M.2 ports are sharing essentially the B/W of a single PCIe 3.0 x4 link.

    WITH THAT BEING SAID, this is still my #1 Z170 mobo. The feature combo is PERFECT. DO WANT
  • phoenix_rizzen - Monday, November 30, 2015 - link

    Only if the storage controller is on the CPU, no? The DMI link is between the CPU and the PCH, correct? And the storage controller won't be sending all storage traffic through to the CPU (this isn't IDE), so that shouldn't be a bottleneck.
  • extide - Thursday, December 3, 2015 - link

    Where else is the data going to go? I mean if you are copying from one drive to another that are both hanging off the PCH ... I am not sure if it needs to go through the CPU or if it has something like DMA.
  • Ian Cutress - Friday, November 27, 2015 - link

    Something seems to have gone awry with our content engine. It should be fixed now.
  • bug77 - Friday, November 27, 2015 - link

    One of these babies should be delivered to my door tomorrow. Needless to say, I can't wait.
  • Idrathernotsay - Friday, November 27, 2015 - link

    Would it be possible to see a review of one of the lower-end z170 motherboards, even a quick test? I realize that they don't provide much in the way of interesting or remarkable features, but I'd be really interested in seeing what, if anything, you actually lose with a mobo under $100 in performance if you don't care enough about the "crazy" stuff to fork over twice or three times as much cash.
  • Impulses - Friday, November 27, 2015 - link

    Didn't they review the ASUS Z170-A? It's a little over your budget but it's one of the better "cover all bases" budget options (2 3.0 headers, Type C in back, M.2, etc). I think when you start going closer to $100 than $150 you do sacrifice a lot more than it's worth, starting with build quality.
  • Idrathernotsay - Friday, November 27, 2015 - link

    Dunno about the actual quality of the motherboard, but Newegg has, for example, an MSI board for $89, with USB 3.1, m.2 slot (PCIe 3.0 x4), and Realtek audio/network. I'd just like to know if a fancy network card can justify 2x the price, or if the bottom-of-the-barrel motherboards are made of literal garbage and will burst in flames if you look at them the wrong way, because, -for me specifically-, a network card just has to be there and the audio is useless (as I've a xonar for my headphones).

    Are there actually any stability issues with the cheaper motherboards? Weaker power delivery circuitry that limits the possible overclocks— if so, by how much? $80 dollars can be quite a bit, and either saving or spending them somewhere else on the machine might make a bigger difference than going for a pricier mobo...
  • Aspiring Techie - Friday, November 27, 2015 - link

    The cheaper motherboards probably can't handle much of an i5 overclock. However, I own a $60 Asrock H81m DGS board, and I can maintain a stable 4.0 GHz at low temps with the stock cooler. Cheaper motherboards don't have as elaborate vrm cooling solutions, so high overclocks can bust the board. However, cheaper boards will do roughly the same as a more expensive board as long as you aren't running a PCIe SSD or a very high end graphics card with an overclocked i7.
  • Idrathernotsay - Friday, November 27, 2015 - link

    Fine, but that's a bit vague tho. The "problem" with the cheaper end of the spectrum is that you have to rely on word of mouth, or occasional reviews on stores, that might or might not actually apply to the motherboard you're considering. How does "probably can't handle much of an overclock" translate in numbers? It's obviously gonna run stock (or they wouldn't sell it), and it —does— support overclock, but how would that be affected by the power delivery with the specific Skylake architecture? If it runs 4.6 fine, I (and a bajillion other people) would be perfectly fine with that while putting $80 more onto the GPU (or back in my wallet).

    I'm not asking for an in-depth review (as there's not much point in a feature-light product, when the general architecture of the platform has already been analyzed in details), but I just wonder how useful reviewing pieces of hardware that are full of "gimmicks" and targeted to the smallest of niches (as the article itself states in the conclusion), and not even give a quick glance to the bottom-tier that actually moves the most units.

    I'm not saying that any of the writers here —have— to, or that the reviews that are currently published aren't interesting, just that I'd like to see some Anandtech-quality numbers of shit I might actually buy. They do test cheaper SSDs and the less expensive GPUs, less pricey mobos wouldn't be that far out.
  • alexdi - Saturday, November 28, 2015 - link

    There isn't any significant difference. You'll have fewer PCIe lanes, fewer third-party chips, and somewhat simplified voltage regulation. That stuff will matter for some edge cases and not at all for everyone else. Otherwise, IME, the manufacturer of the board is considerably more important than what it has on it. They apply the same QC, good or bad, to L the boards.
  • JlHADJOE - Friday, November 27, 2015 - link

    OC software is getting really good. In this case it looks like it actually outdid your manual OC, getting 4.7GHz stable at lower voltage and temps.
  • extide - Monday, November 30, 2015 - link

    x2, I was surprised that this was not mentioned in the article!!
  • rallyhard - Friday, November 27, 2015 - link

    The "Conclusions" link, under "Quick Links to Other Pages" actually links to page 1.
  • Ian Cutress - Friday, November 27, 2015 - link

    I'm not sure why that is. The link should work as it was.
    I've updated it now to include the slug, and it seems to be fixed.
    Thanks for letting us know :)
  • ghostmuse - Friday, November 27, 2015 - link

    Man, that is a bummer about the DPC latency. As someone who does a lot of pro audio work, which requires the lowest possible DPC latency, and someone who wants to build an enthusiast system for 4k gaming, Skylake thusfar is not letting me have my cake and eat it too. I was looking at this board because it looked like you'd be able to run Dual SLI and still use one of the M.2 ports, AND was hoping the DPC latency would be low, but that looks like that's not the case. My first choice, ASUS' Z170 Deluxe scored well on the DPC latency tests but if you try to use SLI you block the M.2 port. I guess I'll just save my cash until some next gen boards pop up, or hope that ASROCK manage to get that DPC latency figure down with bios updates.
  • Byte - Saturday, November 28, 2015 - link

    I just got one of these puppies running the past week and its awesome. I just got the latest BIOS update, is there an accurate way to measure DPC latency in Win10?
  • ghostmuse - Monday, November 30, 2015 - link

    I haven't used this particular piece of software before, but it says it has been updated for Windows 10. Hopefully that = accurate readings. http://www.resplendence.com/latencymon
  • hubick - Friday, November 27, 2015 - link

    No 10GBASE-T? :(
  • extide - Monday, November 30, 2015 - link

    Not Extreme 11
  • msrt - Friday, November 27, 2015 - link

    Another pro audio user here.
    Are you sure those DPC values are caused by the BIOS and not the wireless drivers?

    On the DPC slide both boards above 200us have WIFI onboard, the others don't.

    With my quite old system I have to use streaming mode through WLAN optimizer or disable WIFI altogether to get rid of DPC spikes. Steady 60-80us vs. the frequent 300us spike.

    Maybe it's the same issue with these (The MSI G1 and this one).
    Ian, could you post an average DPC number or have a quick look with WIFI disabled?
  • Byte - Saturday, November 28, 2015 - link

    This board does not come with Wifi, but has a mPCIe for it and some holes
  • alyarb - Friday, November 27, 2015 - link

    I clicked on this because I wanted to see 3 nvme drives in RAID-0. It's in the title and you post a picture of such a setup on page 1, but I don't see any storage testing besides USB file copy?
  • OFelix - Friday, November 27, 2015 - link

    "The UEFI Tech Service is this weird idea for ASRock to receive submissions about BIOS errors and issues. I’m not sure if ASRock actually answers any of these, or if it is just used as an internal metric to move certain issues up the priority chain."

    I reported a problem last week and got a response from Taiwan in a couple of days.
  • DanNeely - Friday, November 27, 2015 - link

    Is the USB 3.1 front panel adapter something that could be used to salvage the worthless SataExpress ports on any motherboard; or did asrock have to do anything special on the board to make it work?
  • pedjache - Saturday, November 28, 2015 - link

    As pictured, speical stuff is on the board of the panel, not the mobo. You use 'normal' SATA express, usb2.0 and molex. Wonder if(when) they or anyone else will put up just the front panel on sale.
  • Byte - Saturday, November 28, 2015 - link

    Newegg has the front panels listed, but OOS. $60 though.
  • DanNeely - Monday, November 30, 2015 - link

    Immediate availability isn't a concern, and I'd rather have a 2x type C model anyway (Asus has announced one, but it's a 5.25" bay enclosure for a full height expansion card making it a bit of a WTF); I probably won't have a USB-C device until late summer/early fall of next year.

    If these can be made usable with Haswell at all is a bigger concern for me. I know I'd be bandwidth limited and only able to hit max speed on one port if the second is idle; but that's a limitation I could live with.
  • pedjache - Monday, November 30, 2015 - link

    It's actually ASmedia controller that drives the front panel, so apart from the fact that using SATAexpress will probably disable 2 of your SATA ports, I see no other concern on usability with haswell.
  • nunya112 - Friday, November 27, 2015 - link

    how many layers in the motherboard? is there copper etc.??
    ASROCK interests me, as they are better priced. if they start to have better components etc. and now the only issue I have is the motherboards are usually thin, and quite flexible
  • extide - Monday, November 30, 2015 - link

    Why do you care about the # of layers? Do you mistakenly think that more is better?
  • tuxRoller - Saturday, November 28, 2015 - link

    How many samples did you use to determine that dpc time?
  • careyd - Saturday, November 28, 2015 - link

    Ian, love the article, read every word, studied every chart. Desperate to see the headline feature of triple M2 RAID benchmarked. Preferably with Samsung 950's, since I think adapting to the Intel 750's is a lot less elegant despite the possibility of greater performance. This type of arrangement should saturate anyway. But want to see throughput and IOPS numbers. Also, if this beggar can be a chooser, I'd love to see benchmarkes of two of the M.2's striped, with a third running solo. I could envision a number of reasons why this configuration might be a preferable, having a solo M.2 for system drive and the striped pair for other data intensive tasks/media/etc. I think it would make a heck of a cache drive for Adobe After Effects.
  • TesseractOrion - Sunday, November 29, 2015 - link

    I think I'd choose striped for OS, solo for redundancy instead, I'd want the OS to be as responsive as humanly (computerly) possible.
  • jasonelmore - Sunday, November 29, 2015 - link

    what a waste. the OS is never going to need 2.1GB Per second write and reads. Only a scratch disk or video editing app would use it, then you might as well use careyd configuration
  • murak01 - Sunday, November 29, 2015 - link

    A few tests with three 950 Pro in RAID0 would be interesting. Thanks for a good review!
  • Reflex - Sunday, November 29, 2015 - link

    Right now you can pick up 250GB Samsung 850 EVO M.2 SSD's for only $79 at Newegg and Amazon. Given that this board is $220, dropping $240 on storage is not ridiculous and should perform exceptionally well.
  • Kutark - Sunday, November 29, 2015 - link

    Did i miss the HUGE opportunity to nerd out and test triple m.2 sata express RAID array transfer speeds? I see no benchmarks...
  • scottjames_12 - Sunday, November 29, 2015 - link

    Correct me if I am wrong, but with all 3 M.2 ports hanging off the Z170 PCH (which they must be if they are RAID capable), isn't the bottleneck going to be the DMI link from the PCH to the CPU? Which is essentially a PCIe 3.0 x4 connection, with maximum theoretical bandwidth of 3938.4 MBps. Seems to be confirmed by the numbers ASRock quoted, as 3 SSD750's should be able to hit over 8000 MBps sequential, but they only managed 3500.
  • extide - Monday, November 30, 2015 - link

    YEP
  • Max Hereticus - Sunday, November 29, 2015 - link

    Bought the ASRock Extreme 6 for my I7 6700k. What a piece of shit.. still in the middle of the RMA process. ASRock tech support and quality leave a lot to be desired. Nice Motherboards except they last less than 2 months.
  • extide - Monday, November 30, 2015 - link

    Your anecdotal references/experiences are basically worth nothing in the grand scheme of things, ya know.
  • fackamato - Sunday, November 29, 2015 - link

    No benchmarks from the x3 M2 in RAID?
  • firstmithrandir - Monday, November 30, 2015 - link

    Is there any possibility to test it with more than one SSD disks like it is claimed in the tittle: "The ASRock Z170 Extreme7+ Review: When You Need Triple M.2 x4 in RAID article"?

    At the moment the tittle of article is misleading.
  • asgallant - Monday, November 30, 2015 - link

    Even if z170 breaks out enough connections for 3 M.2 PCIe x4 slots, aren't you still limited to the aggregate bandwidth across the DMI? The 4 DMI lanes between the chipset and CPU are roughly (exactly?) the same bandwidth as 4 PCIe 3.0 lanes. Maybe RAID could provide some improvement on drives that cannot completely use the ~3.85GB/s of bandwidth that 4 PCIe 3.0 lanes provides, but you're not likely to see much, I think.
  • fackamato - Monday, November 30, 2015 - link

    Still, that's >1GB/s per SSD.

    Plus, triple RAID0 would increase random IOPS as well right.
  • Ethos Evoss - Wednesday, December 2, 2015 - link

    Sooo useless and pointless and ''BLACK HOLE'' expensive !!
  • Ethos Evoss - Wednesday, December 2, 2015 - link

    your gpu will be struggling as u obviously sharing PCi-e lanes !!
  • juliabrown943 - Thursday, December 3, 2015 - link

    what Jeffery said I am impressed that some one able to make $8960 in one month on the computer . you could try this out.....>>>>>>>............. .­­earni8­­­ dot ℭom
  • juliabrown943 - Thursday, December 3, 2015 - link

    what Jeffery said I am impressed that some one able to make $8960 in one month on the computer . you could try this out.....>>>>>>>............. .­­earni8­­­ dot ℭom

Log in

Don't have an account? Sign up now