New Haswell microcode to block overclocking on non-Z87 mobos

Officially, Haswell overclocking requires two things: one, a K-series variant of the processor, and two, a motherboard with a Z87 chipset. Several motherboard makers have recently circumvented the second requirement, enabling Haswell overclocking on H87, B85, and H81 mobos via a firmware update. Unfortunately, it looks like that workaround may not be available for much longer. The folks at Hardware.fr asked Intel about the workaround last week, and they received the following statement.

Intel plans to release a microcode update that limits processor core overclocking to Intel Z87 based platforms WW30’13.

"WW30’13" seems to be shorthand for week 30 of this year, which is this week. It’s not clear how quickly the update will be rolled out in the form of firmware updates for motherboards, but I assume we’ll eventually see Haswell processors with the new microcode trickle into e-tail stocks.

In other words, if you’re planning to build an overclockable Haswell system, you’re probably better off going the Z87 route. It’s possible that motherboard vendors will find another workaround to the limitation, but I wouldn’t bet any money on it.

Comments closed
    • BaronMatrix
    • 6 years ago

    Umm, when has Intel done ANYTHING that didn’t involve monopoly profits…?

    • SylviaHouston7
    • 6 years ago
    • Krogoth
    • 6 years ago

    This entire thing is a non-issue. The vast majority of K series buyers are going to get a motherboard with a Z68, Z78 and Z88 chipset. The people who get H and Q chipsets are going to pair it with a non-K series chips with no-overclocking.

      • travbrad
      • 6 years ago

      You’re right it is mostly a non-issue for the reasons you mentioned. It’s still sort of an arbitrary dick move by Intel though.

    • CathyRolphe5
    • 6 years ago
    • brothergc
    • 6 years ago

    or one could kick intel to the curb and get a AMD system , they do not pull any of this BS

      • Klimax
      • 6 years ago

      Because they can’t, not because they wouldn’t. They would if they had 80% of market…

      (the->they)

    • Meadows
    • 6 years ago

    Why is intel being such a jerk about this?

      • NeelyCam
      • 6 years ago

      Because Intel likes money

        • mcnabney
        • 6 years ago

        It isn’t just that. The K series can overclock, but are crippled by having VT-d and other new instructions disabled. Non-K are current, but can’t be overclocked. You can’t have both no matter how much money you have.

          • NeelyCam
          • 6 years ago

          If you want overclocking and VT-d, you take some E

            • Deanjo
            • 6 years ago

            Or buy AMD.

            • MadManOriginal
            • 6 years ago

            So you can overclock just to get the same performance as a stock Intel CPU? MAKES SENSE TO ME!

            • Deanjo
            • 6 years ago

            1) it overclocks
            2) it does not castrate virtualization

            If you are going to virtualize, chances are you are not going to have an aggressive OC, if any. AMD is still the better choice for virtualizing without having to spend some serious money on high margin Xeons or Opterons.

            • MadManOriginal
            • 6 years ago

            Two posts ago you said ‘buy AMD’ if someone ‘wants overclocking and VT-d’…heck, you even say in this own post ‘1) it overclocks’ then you say someone is not going to overclock when virtualizing. Make up your mind.

            Also, almost all of the non-K Intel Haswell CPUs have VT-d, so no Xeons required. You’re going to have to explain more clearly how an AMD CPU, not overclocked because someone is not going to when virtualizing, is ‘still the better choice’?

            • Deanjo
            • 6 years ago

            They still have that option if they want to. With intel you don’t. What is so hard to understand about that?

            With AMD you can overclock, with AMD you do not get castrated virtualization, do with it what you please. Can you say the same with intel without paying 3x as more? I don’t think so.

            I an guarantee you I can take any 8350 off the shelf, load it up with 6-8 VM’s, encrypt the drives and run it on ESX cheaper then you can do with any intel. With ECC to boot.

            • Klimax
            • 6 years ago

            Slowly…

            • Deanjo
            • 6 years ago

            I guarantee you that if you take a $200 intel and $200 AMD throw ESX on them and slap 4 virtual instances on them that the AMD whoops the intel.

            • Klimax
            • 6 years ago

            200? Like this one? [url<]http://ark.intel.com/products/75043/Intel-Core-i5-4570-Processor-6M-Cache-up-to-3_60-GHz[/url<] 4 instances on four cores...

            • Deanjo
            • 6 years ago

            Keep going higher… Sorry but the 4570 doesn’t even come close to the 8350’s running 8 instances and on IO thanks to intels lack of VT-d. The 4570 doesn’t even support ECC.

            • Klimax
            • 6 years ago

            Goalpost moving detected. Original claim was for 4 instances.
            And linked CPU has VT-D… (read link before commenting)

            Also there is missing piece of targeted workload (latency vs. throughput)
            And finally not even 8 “cores” might help, when bottleneck is often elsewhere, but this kind of test is so far missing.

            ETA: That bloody original post was edited and number of VMs changed. Should learn to quote here to prevent this…

            • clone
            • 6 years ago

            have you seen the benches or are you just focused on the last 5% in certain scenarios regardless of expense?

            AMD cpu’s aren’t nearly as slow as implied and in some cases they are as fast as Intel cpu’s that cost 3X as much.

            • Klimax
            • 6 years ago

            First, you’d need to define yourself problem area and second define which CPUs are compared.

            Anyway:
            [url<]http://extrahardware.cnews.cz/testy/haswell-mezi-nami-recenze-intel-core-i5-4670k-i7-4770k/strana/0/5[/url<] Virtualization by Virtualbox. Note that they had k version of chip.

            • MadManOriginal
            • 6 years ago

            Ahmegerd, facts!

            • Bensam123
            • 6 years ago

            Good thing running one virtual box and looking at the performance in it is the same as running 4-8 virtual boxes and looking at the performance across all of them.

            Facts indeed… cherry picked.

            • MadManOriginal
            • 6 years ago

            Feel free to find some benchmarks or run some yourself then. If what you’re implying were true, AMD would be cleaning up in the server market, but they aren’t.

            • clone
            • 6 years ago

            AMD’s failing to “clean up in server” has as much to do with the closure of it’s server business as it does their chips power consumption levels while crunching those numbers.

            something that matters far less in small scenarios where cooling a single system vs a farm isn’t an issue.

            • Klimax
            • 6 years ago

            Well number of cores required can be killing too as nearly all licensing moved to per-core model and either per-CPU has been killed or made very expensive. At that point AMD CPUs will lose their cost advantage when SW license will cost you more then “more” expensive Xeon…

            ETA:Removed note about ent. Linux, because I realized I don’t know exact arrangement for companies. (Base OS might be free, but support is definitely not and management SW might be paid too)

            • Deanjo
            • 6 years ago

            Who the hell does serious virtualization on windows?

            • MadManOriginal
            • 6 years ago

            The same people who do virtualization on overclocked systems? :p

            • Klimax
            • 6 years ago

            Those who understand and know. Hell, this is ridiculous and stupid question.
            Second, such licensing is getting everywhere…

            • hescominsoon
            • 6 years ago

            I do..hyper-v hypervisor is actually pretty good..the ability to not have to have a dedicated controller card for RAID is nice as well.

            • kc77
            • 6 years ago

            Most support in the land of Linux or Unix is by the socket not the number of cores. That being said, at the 2P level the price for Intel CPU’s is steep. So much so that you’ll spend half of what you would spend on an Intel 2P rig and still the AMD system will be faster because AMD is selling their server processors at incredibly cheap prices.

            Deanjo isn’t wrong when it comes to the best bang for the buck when it comes to visualization and 2P systems.

            • Bensam123
            • 6 years ago

            I’m using a 8350 because it IS extremely good at multitasking and my experience in a completely unrelated area to VMs, streaming, has lead me to choose it over a 3570k. No I don’t have any amazing benchmarks to back it up and despite asking TR pretty much every time they do a review, they haven’t decided to look at this growing area in technology.

            Perhaps there is a bit more to why AMD isn’t mopping up the server market… Like thermal envelopes, Intel always being preferred there regardless of performance, and the fact that Intel actually has chips with 12 actual cores on them if they really want to pack in some density. There it’s not so much about price as overall punch. Budget seekers generally aren’t found putting together clusters.

            It could also be that no one actually looks into this stuff and it’s why we’re stuck with a wopping one benchmark that doesn’t actually cover the scenarios we’re talking about. People simply just go ‘meh AMD can’d do nuttin’ gud’, look at all the reviews showing how awesome Intel is at playing games and loading windows which confirms what they think and call it good.

            • Klimax
            • 6 years ago

            So far no other tests known. thought Anandtech did, but can’t find anything there either, just standard perf tests…

            So not cherry picked, just the only known to me and actually existing…

            ETA1: Also depends, whether you want latency or throughput…
            ETA2: Found it finally on AT – didn’t previously remember name of virt.
            Note: Server class hardware in same price bracket per chip. Whole servers had quite bit differences.
            PD:
            [url<]http://www.anandtech.com/show/6508/the-new-opteron-6300-finally-tested/4[/url<] Orig. BD: [url<]http://www.anandtech.com/show/5553/the-xeon-e52600-dual-sandybridge-for-servers/6[/url<] Anyway, generally my statement stands - slowly as I didn't say how much... 😀

            • Lans
            • 6 years ago

            How about you defining your problem that your link addresses? 😉

            If you want to say older Core i7 chips like 4770K is other alternative for virtualization and overclocking then sure.

            But I have to agree with Bensam123 that your link seems valid for only one VirtualBox… Maybe you want multiple OSes on one machine rather than multiple users but if you wanted the latter, as previously mentioned, this is where AMD seem to do pretty well:

            [url<]http://www.phoronix.com/scan.php?page=article&item=ubuntu_1204_virt&num=8[/url<] Phoronix reports Core i7 3960x being 21% of "bare metal" with 7-zip with VirtualBox. And if I look at your link, for Core i7-3820 (no 3960x so I picked another "Sandy Bridge"), "bare metal" seems to be 22223 and VirtualBox is 16610 with 7-zip so that is why I don't believe there are concurrent VirtualBox running.

            • Klimax
            • 6 years ago

            First, it was one few tests I was aware of, so I linked it. (In later post I found some Anandtech benches too) There is a reason, why I noted they had k version of chip without VT-D.

            Second, you should have linked to result page, not conclusion page. (Makes it hard to hunt down results you refer to) And I still don’t see multiple VMs at once in it. Looks more like tests in my links. (Contention is about multiple VMs at once)

            I don’t think anybody tested yet Haswell non-k or Xeon versions in this area.

            Maybe I could provide 3930k Hyper-V tests (got some nice WS2012 lab keys for project at uni), but I don’t have BD/PD.

            • Pholostan
            • 6 years ago

            [quote<]I an guarantee you I can take any 8350 off the shelf, load it up with 6-8 VM's, encrypt the drives and run it on ESX cheaper then you can do with any intel. With ECC to boot.[/quote<] This exactly what we have done at work. I assembled a stupid cheap server for some random stuff, you know logging, Zabbix, puppet etc. The 8350 CPU was the most expensive part (motherboard was very cheap), the whole thing ending up costing us well under a $1000. Then some of our Vhosts went belly up (a couple of Dell blade servers crashed due to firmware problems, etc) and my little thing ended up taking over all critical VM:s, or about 15 VM:s or so. Slapped in another 16 GB of ECC RAM in it for 32 GB total and no problem really. It ran all our critical stuff no problem. Makes you think.

            • Bensam123
            • 6 years ago

            Depending on how much you pay, ‘stock’ used in that contenxt can mean a lot of different things and is a meaningless metric in and of itself, especially when price isn’t considered. SandyBridge-E is ‘stock’, server processors come ‘stock’.

            • MadManOriginal
            • 6 years ago

            Oh, so now TCO matters? How ’bout that electricity.

            • Bensam123
            • 6 years ago

            That wasn’t the point, you’re comparing apples to oranges, then exclaiming the apples are better.

            If you want to say a $240 processor is faster then a $180 one, then I would whole heartedly agree. But instead you chose to make it seem like they’re the same price and one ‘stock’ processor merely defeats another one of equal value.

            • MadManOriginal
            • 6 years ago

            And you’re being stupid by talking about SB-E and ‘server processors’ (Xeons and Opterons).

            Intel’s non-K DESKTOP (repeat it with me: desktop) CPUs have VT-d, which is what has Deanjo’s panties in a twist. No, they don’t overclock – at least Haswell doesn’t, Ivy Bridge non-K CPUs can overclock by +4 bins on Z and some H chipset motherboards – but Deanjo himself said that people who are doing virutalization shouldn’t overclock, then he says they could, so I really don’t know how to take what he says seriously. The non-K DESKTOP (there’s that word again) CPUs at $200 versus a $180 FX-8350 are close enough pricewise to not matter. $20, OH NOES!

            • Bensam123
            • 6 years ago

            Are you being stupid by thinking that I’m actually comparing a 8350 to server processors? I think we’re on two different wires here (including the silly insult at the beginning). I mentioned server chips because they are ‘stock’, not because that’s what we’re actually comparing here.

            The overclocking on non-k options is silly, I wouldn’t even consider that a option (when you find a motherboard that supports it). 4 Bins is also a clever way of saying it just operates at max turbo frequency, which no longer works with haswell… which is Intels new processors and what everyone buys. Making it seem like buying older processors is a relevant option now is just as silly as implying apples to oranges comparisons.

            Deanjo can say whatever she wants. Just because you have more options doesn’t mean you can’t overclock. This supports VT-D and OCing, so you have a processor that does BOTH, not one or the other. And you CAN OC while running VMs, even if Deanjo doesn’t want to do it or think it’s a good idea. Hell you could run VMs in a normal mode, then OC it when you plan on gaming while in whatever OS you want.

            Then there is the whole part about the 8350 doing some workloads as good as a 3960k, even if it doesn’t do them all nearly as well. Then the whole part about the 8350 doing multithreaded workloads well. Then the part about VMs, especially when you’re running a bunch of them, being extremely multi-threaded (because… you know, running 4-8 OS’s all at the same time is kinda a multithreaded thing), and you end up in a position where there really is no contest. Even if it’s a niche one.

      • Krogoth
      • 6 years ago

      Market segmentation FTL.

      Intel wants to justify the existence of their enthusiast tier platforms.

    • marvelous
    • 6 years ago

    pure evil

    • HisDivineOrder
    • 6 years ago

    It amazed me that ASRock did it in the first place. That it lasted this long was interesting and that it ended is unsurprising.

    Still, I can’t help thinking the whole thing is indicative of something a little more important, which is that Intel is needlessly segmenting its product lines to milk the enthusiast as hard as possible while year to year decreasing any reason for the enthusiast to continue to upgrade.

    This is going to lead inexorably to Intel raising prices and further segmenting the market to try and make enough money to continue in the market while also reducing performance improvements.

    Looks like a death spiral to me.

      • Voldenuit
      • 6 years ago

      [quote<]This is going to lead inexorably to Intel raising prices and further segmenting the market to try and make enough money to continue in the market while also reducing performance improvements. Looks like a death spiral to me.[/quote<] QFT. Intel is its own worst enemy, and alienating enthusiasts sounds like a bad idea when they are probably the only people interested in buying new PCs and PC parts in these times.

      • NeelyCam
      • 6 years ago

      [quote<]Looks like a death spiral to me.[/quote<] It's only a death spiral if the enthusiast market disappears altogether... and I don't think it ever will. The reason why this is happening is because AMD decided not to focus on high end consumer market anymore. No competition -> no limits to Intel milking the market.

        • Concupiscence
        • 6 years ago

        I don’t think it’s AMD’s decision to make. They’ve been out-maneuvered and out-engineered.

      • Klimax
      • 6 years ago

      No. It doesn’t indicate anything about Intel, just boardmakers willing to do anything for short-term gain. It’s not like those cheap-craptic boards are made to sustain OC anyway.

      Also you got several large gaps in reasoning in later part of posts. How you can infer anything from this about future prices?

      As for performance, you are still like many others missing reality – we are already near end of gains. We have no new general algorithms for base math operations, computers are still bound by Turing machine. There are limits for performance and Intel reached that point. Any attempt at workarounding issue will lead to significant complexity of chips and we all saw how well it works for GPUs.

      Reminder: ARM and AMD will hit same wall, question is when and what performance spot it’ll be, but no if.

        • HisDivineOrder
        • 6 years ago

        Intel hasn’t hit any wall. They deliberately CHOSE not to pursue performance because their competition now is performance per watt. They aren’t focusing on performance as much as the “per watt” part of that argument. They are designing their entire chip line around eventually having their high end in tablets and everything–from the low end to the high end–is based on a chip line that is primarily focused on getting power utilization down.

        That would be fine if that also meant they were making faster clocked versions of said chips that threw caution to the wind and gave people performance options that were faster than prior generations.

        Instead, the more they overtake the plumbing on the chip, the more they keep the clockspeed the same. Do you seriously think that Haswell is as fast as they could have made that chip? They didn’t want or try to make it faster. They didn’t have to.

        Why? Because their competition isn’t faster in performance than them. Not ARM, not AMD. The only competition that matters to them now is just downright smaller and cooler (temp-wise) than them, so they are ignoring performance gains entirely (though they have marginal ones).

        If you want to know why I make the leap on pricing, consider this. Every year, Intel raises the price of the enthusiast-segmented CPU’s. Sandy Bridge: 2500k ($220), 3570K ($230), 4670K ($240). Add on $100 for the 2600/2700K, 3770K, 4770K lines. The performance is not noticeably different between these CPU’s, but prices are inflating. When performance gains aren’t there for the enthusiast and prices continue to increase, enthusiasts will “sit this generation out.”

        This will lead to less enthusiasts buying into that generation. Meaning, Intel will make less money on said generation. Have you noticed, by chance, that there are a lot more sales a lot more quickly on Haswell K-based CPU’s now that a couple months have passed? How long did similar deals take for SB (six months)? Or IB (four)?

        The problem is Intel can’t rely on the enthusiast market to show up if they don’t give us gains and they can’t give us gains if they’re focused on performance per watt at the emphasis of the “per watt.” So in the end, they’ll abandon the mainstream enthusiast parts in favor of the E-line. The E-line is great, except that it’s incredibly expensive compared to the K-line built on the mainstream processes and chip lines.

        Pushing enthusiasts and hobbyists (and anyone who wants to install a CPU into their motherboard) to the E-line is going to reduce the number of people still doing it because of how expensive everything’ll become. When hobbyists one year have the option to buy $400 K-based CPU+enthusiast MB and $50 8GB memory and the next year have $500 CPU’s, $300 motherboards, and $150-200 for DDR4 equivalent, that’s going to naturally drive people away.

        Moreover, the E-line is always behind the times in technology, which tastes foul to the very people most concerned with the technical designs of the chips regardless of performance. Let’s face it. We’re all in this because we love technology and we want to think Intel is actually giving a damn about us rather than just throwing us table scraps a year or two later.

        The more Intel focuses on trying to beat ARM, the less concerned with performance they’ll be. The less they focus on performance, the fewer enthusiast chips they’ll sell year to year. The less chips they sell to enthusiasts the less meetings they’ll have where they have someone internally saying, “Enthusiasts and hobbyists are an important segment to us! They make us money and they do great things for advertising for us,” because there will be charts that show how enthusiasts are buying less chips and deserve less emphasis. Which will lead to higher prices and fewer chip sales being expected.

        Eventually, a tipping point will force them to the conclusion that the hobbyists and the enthusiasts are so few that having user-installed CPU’s in their main line is not worth the investment. That’s when they’ll make a push for those users to spend up and buy into the E-line. That’s why the picture about user-installed CPU’s post-Haswell is so murky.

        Intel suspects this will happen, but they don’t want to sound the death knell on the K-series CPU’s until they’re absolutely sure. Skipping Broadwell on the desktop side (since it’s just a refresh of the tech with a few minor improvements to CPU per watt without many gains to the performance (except the iGPU which is not essential to desktops), making Haswell look positively incredible in the performance gain realm) makes sense then and gives them time to consider how to approach Skylake.

        But notice. They aren’t even refreshing Haswell a year from when it released. They’re delaying that update. They’re stretching the generations out to try and make them last longer. They know now they’re going to have to spend longer in this gen than last gen, which as longer than the gen before that, to make up the same profit as they have in the past.

        That’s the death spiral. They give up on enthusiasts, who give up on them. Forcing them to give up on more of them and leading to more of the enthusiasts giving up on them. Until a critical mass is reached and they conclude the market isn’t there.

        Seems obvious to me.

          • Klimax
          • 6 years ago

          Very wrong post (even if long). Base algorithms used for math haven’t been improved much, limiting performance. Base theory underlying computing is still same, meaning there is limit to what can be achieved.
          There are some theoretical upgrades to architecture, but they increase severely complexity. Look at GPUs and their trouble with manufacturing and sizes of their dies.

          There are limits for everybody, Intel, AMD and ARM. Until those limits are breached, nobody gets another 30% increase after reaching SB-level of performance if even they get that far.

          As for -E, it uses server chips. Those get different treatment, because they have different requirements. (IIRC high performance Xeons were always later) Validation and larger die for all those cores, L3 and PCIe…

          Next massive change will be with quantum computing or another set of mathematical and data-processing algorithms not sooner.

          Stop clinging to notion that Intel holds CPU back for monetary gains, because it’s not true. IPC is more or less exhausted and ILP won’t last much longer. (Haswell was mostly about hyperthreading and finishing off integer based workloads)

    • brucethemoose
    • 6 years ago

    No big deal, you can always just roll back the firmware. If it is a big deal to some people… Well they aren’t the overclocking type anyway.

    Once there’s a sale/price drop on a 4670k, I’ll probablly save $20 and pick up a nice H87 motherboard with it.

    • Chrispy_
    • 6 years ago

    Most pointless product segmentation from Intel yet. It’s even dumber than removing VT-d from the K-series processors.

    Forcing the Z87 chipset for overclocking is like a restaurant offering a choice of wines to go with your steak, but then saying you can only actually [i<]eat[/i<] steak if you pick the 2008 Pape du Cote Cabernet Sauvignon at $60 a bottle; Choosing a different wine means you will be served steak and you may look at the steak, smell the steak, and maybe even lick the seasoning off the top of the steak, but you will be forcibly ejected from the restaurant and barred for life if you attempt to eat it.

      • deathBOB
      • 6 years ago

      Hmmm. I think it’s more like you go to a restaurant with a coupon (just like you get more than you paid for by overclocking) and the coupon says “Minimum Purchase $50.” You don’t have a right to the discount (just as you don’t have a right to overclocking), so requiring a certain minimum purchase to get it isn’t uncalled for.

      Alternatively, think of the price premium for overclocking as just being split into 2 parts: the premium for the K-series chips and the premium for the Z87 chipset (beyond the value of the added features). Maybe it’s a bit sneaky for Intel to split it up like that, but I don’t think it’s unfair.

      • MadManOriginal
      • 6 years ago

      No, it’s like a restaurant selling a steak, and you can choose to buy a red wine just because it’s red, or a red wine that specifically matches the side dishes that come with the steak.

        • tay
        • 6 years ago

        These fucking steak analogies aren’t making any sense to me.

          • BIF
          • 6 years ago

          Yeah, me neither. But they do make me hungry!

            • deathBOB
            • 6 years ago

            You guys sound poor. Try inserting the words “Hamburger Helper” instead of steak and “MD20/20” instead of wine and see if that makes sense. (JK of course)

          • MadManOriginal
          • 6 years ago

          Ok…so you buy a steak (CPU) with side certain dishes (CPU model features – K series for overclocking, VT, Hyperthreading, whatever) and get some wine too (motherboard.) You can get any old red wine (LGA1150 motherboard) simply because it’s red (compatible – B, H, or Z chipset) OR you get a red wine that specifically goes along with the side dishes (Z chipset). In this example, the side dish is overclocking, and the ‘matching’ red wine is a Z series chipset.

            • Voldenuit
            • 6 years ago

            Except in intel’s case they pour the wine out of the same barrel and charge more for one glass vs the other.

            • MadManOriginal
            • 6 years ago

            It’s the nature of silicon chip manufacturing… if singling out Intel makes you feel better, great, but everyone does it.

    • DPete27
    • 6 years ago

    [quote<] if you're planning to build an overclockable Haswell system, you're probably better off going the Z87 route.[/quote<] Or get your H87/B85/H81 mobo while the getting is good and don't update your UEFI with anything released after this week. Keep in mind, you can always roll back your UEFI firmware to a previous version. I doubt mobo makers will bother to go back through all their previous firmware revisions and make this microcode change.

      • MadManOriginal
      • 6 years ago

      Or don’t buy a product and expect undocumented features. Complaining about it later is stupid.

        • DPete27
        • 6 years ago

        Oh come on, if it werent for undocumented features we wouldn’t have great things like the Phenom II x2-550 and x3-720 (they had a chance to unlock cores/cache). I think this H87/B85/H81 workaround may be in that same category.

          • BIF
          • 6 years ago

          Right, the trunk of a Ford Fairlane would be great for carrying a body. Definitely undocumented!

          * I do not support the use of a Ford Fairlane or its trunk for this purpose.

          • MadManOriginal
          • 6 years ago

          Key word: ‘chance’.

      • Klimax
      • 6 years ago

      UEFI update is not the only way to get new microcode… OS can and will load it too.

        • Farting Bob
        • 6 years ago

        MS has only done so for critical security reasons i think though, and being able to overclock using a slightly cheaper motherboard is not a critical security risk for MS. There’s no advantage to them of doing it and getting all the backlash from users. It’s Intel’s problem, let them deal with this one.

          • Klimax
          • 6 years ago

          Microcode is sent to CPU by OS every boot. As for frequency of updates from Microsoft, least time is every service pack if not more often, because they push microcode as a whole including all fixes and workarounds as provided by Intel.

    • ronch
    • 6 years ago

    You just knew at the back of your head that this is bound to happen.

      • ClickClick5
      • 6 years ago

      Until Intel just ends the ability to OC…

      Time to break out the pencils ladies and gents.

    • Thorburn
    • 6 years ago

    Microcode is stored in the UEFI firmware and loaded at boot time, not stored in the CPU. Its like a CPU driver for the motherboard.

    Unless they revise the CPU stepping (therefore requiring a new microcode) a motherboard vendor could choose not to implement the updated microcode in to their firmware and keep non-K overclocking – but this means they’ll miss out on any other improvements and fixes implemented in to them as well.

      • Star Brood
      • 6 years ago

      And that’s how they get you.

        • Thorburn
        • 6 years ago

        A CPU stepping revision is unlikely on desktop (its expensive and people overclocking on non-Z87 boards will be pretty insignificant) and I don’t know of any serious issues that would force peoples hands on a microcode update. Remember most consumers won’t update their BIOS throughout a systems lifespan anyway.

        If you’re overclocking then you’re clearly not THAT concerned about 100% stability so I’d expect board manufacturers like Asrock who are known for pushing the boundaries a little further and exploiting these loop-holes to just stick with the older microcode.

          • Voldenuit
          • 6 years ago

          If I understand it correctly, intel can release microcode updates to their CPU lineup without any retooling.

          Expect new Haswell chips to ship with the updated microcode.

          They probably can’t *force* existing users to update their microcode, but I wouldn’t be surprised if they sneak in the update as a ‘security patch’ in a driver update.

            • Klimax
            • 6 years ago

            They don’t have to force users, just get OS to update it…

          • DPete27
          • 6 years ago

          [quote<]A CPU stepping revision is unlikely on desktop[/quote<] No? Not even to fix the USB issue in Haswell?

            • Deanjo
            • 6 years ago

            The USB issue was with the chipset not the cpu.

            • DPete27
            • 6 years ago

            That’s right, stupid mistake on my part.

      • WillBach
      • 6 years ago

      Intel microcode is in ROM on-die though [i<]updates[/i<] can be loaded with the UEFI.

        • Deanjo
        • 6 years ago

        Yup, pretty easy to update via a software update. Could appear via windows update or a driver update.

          • crabjokeman
          • 6 years ago

          echo “intel-microcode hold” | dpkg –set-selections
          😉

      • crabjokeman
      • 6 years ago

      I bet mobo makers are happy to close the loophole. When the it first got out, it was a competitive advantage to have it. Now, it’s a liability that lets people buy lower-margin boards instead of forcing them to shell out for Z87 boards. It’s still evil, but at least there are some reasonably priced Z87 boards..

    • allreadydead
    • 6 years ago

    This is what happens when there is no serious competition in the market :\

      • jihadjoe
      • 6 years ago

      Nearly the exact same thing happened with the early Radeon 6950s which were unlockable to 6970s. After a while, AMD plugged that hole, and there was some serious competition in the GPU market.

      Manufacturers want their product segmentation. Plain and simple.

        • bcronce
        • 6 years ago

        I got my 6950 unlocked. w00t. My ROP fill-rate isn’t as strong as the 6970, but my pixel shading power is.

        Interesting to note, not only was I able to unlock the extra shading units, but I was able to OC the GPU to near 900mhz and undervoltage the GPU.

        I was running my GPU at 890mhz and 1.05v(1.1v stock) for a few months. Eventually I flashed the second ROM to 870mhz and 1.08v, which ran about 5c cooler than 1.1v

      • ronch
      • 6 years ago

      And the company that provides ‘no competition’ to Intel has to resort to selling all their ‘enthusiast’ parts fully unlocked.

    • Star Brood
    • 6 years ago

    Hey, for once it looks like early adopters aren’t getting the short end of the stick.

      • windwalker
      • 6 years ago

      Early adopters didn’t know there would be such a trick found so they bought Z87.

Pin It on Pinterest

Share This