Comments Locked

61 Comments

Back to Article

  • thekimbobjones - Tuesday, February 11, 2014 - link

    Is this Correct? "These use sixteen 4 Gb memory packages on board to total up to the 8 GB number"
  • tipoo - Tuesday, February 11, 2014 - link

    Yeah. 4GB * 16 = 64Gb, which is 8GB. The big or little B may have thrown you.
  • tolis626 - Tuesday, February 11, 2014 - link

    Yes,it is.16x4Gb (gigabit,not GB which is gigabyte) means 64Gb.Divide that by 8 and you have 8GB of RAM.
  • Jammrock - Tuesday, February 11, 2014 - link

    The math is correct.

    There are 8 bits in a Byte. 4Gb x 16 = 64Gb / 8 = 8GB
  • TommyVolt - Wednesday, February 12, 2014 - link

    Just to clarify again
    A 8GB module uses sixteen pcs 4Gb chips (or 18 chips when having ECC)
    A 16GB module uses sixteen pcs 8Gb chips

    The problem is that Intels standard processors can not understand 8Gb chips, although they are part of the DDR3 specification standard. Thus 16GB modules do not work on Intels, but they do work on others
  • Lerianis - Friday, February 14, 2014 - link

    You're kidding? Jeez... Intel needs a swift kick in the keister for not including this in their chips.
  • 8steve8 - Tuesday, February 11, 2014 - link

    not working with intel processors like haswell seems to be a serious issue.. what about broadwell?
  • wlee15 - Tuesday, February 11, 2014 - link

    Not really 32 GB is probably sufficient for the mainstream/low-end server market until DDR4 comes along. Those who need more have the six core i7s and Xeons.
  • Gigaplex - Tuesday, February 11, 2014 - link

    It doesn't work with the six core i7s or Xeons either.
  • ZeDestructor - Wednesday, February 12, 2014 - link

    LGA2011 i7s have 8 slots, and Xeons have a much wider range of DIMMs available for high-memory use (RDIMM, LRDIMM).
  • Darkstone - Thursday, February 13, 2014 - link

    Unless you have a notebook.

    I know some folks with 4x8GB dimms in their notebook, that hope to upgrade that to 4x16GB some day. Some manufacturers such as clevo have gone so far to say that 64GB is officially supported. According to this article this isn't the case, and 64GB will never work on current mobile platforms.
  • Zero110 - Tuesday, February 11, 2014 - link

    Pretty sure I'm going to get slammed for missing something, but this also means no possibility of 32GB on an intel MiniITX setup, correct? Been wanting to go smaller than MicroATX, but 2 DIMM slots is too limiting currently.
  • TiGr1982 - Tuesday, February 11, 2014 - link

    Seems to be the case - if Intel platforms don't support 16 GB DDR3 modules, as written in this article, then Intel miniITX is stuck with 8 GB/module x 2 modules = 16 GB
  • Gigaplex - Tuesday, February 11, 2014 - link

    It'll work with Intel mini-ITX systems that use Avoton or Rangely CPUs.
  • Flunk - Tuesday, February 11, 2014 - link

    Which is only helpful if you need a lot of memory and a very slow processor. Pretty niche.
  • Drizzt321 - Wednesday, February 12, 2014 - link

    So awesome for memcache/redis/etc type servers you mean? Yea, pretty awesome I'd think for those kind of servers. Avoid having to move up to 2 socket boards to increase the memory substantially.
  • ricera10 - Wednesday, February 12, 2014 - link

    If you use an Avoton board, you can totally go to 64 GB!
  • Sahrin - Tuesday, February 11, 2014 - link

    Extremely disappointing that Intel didn't implement the DDR3 spec correctly.

    It's great that there's finally an option for 16 GB in a single DIMM, though. I guess I'll be able to buy an AMD system.
  • ddriver - Tuesday, February 11, 2014 - link

    Hope you will still device to keep on living :D

    This "product" aims to solve a non-existent issue, no one would purchase a consumer mobo with 2 memory slots only to put 32 gigs of memory in it.

    The systems that actually need plenty of ram support larger capacity modules, so it is all good.
  • Gigaplex - Tuesday, February 11, 2014 - link

    There's a fairly recent trend to build powerful mini-ITX systems lately which is generally limited to 2 memory slots.
  • madmilk - Wednesday, February 12, 2014 - link

    Which brings up the question: why are there no mini-ITX systems that support RDIMMs?
  • ZeDestructor - Wednesday, February 12, 2014 - link

    More interesting (to me at least) is why tehre aren't any mini-ITX boards with 4 SO-DIMM slots instead?
  • hardwickj - Wednesday, February 12, 2014 - link

    Lack of physical space. ASRock makes one, but it's expensive, not sold in the US, and is "Mini-ITX Extended" which would fall somewhere between Mini-ITX and Micro-ATX I'm guessing.
  • ddriver - Wednesday, February 12, 2014 - link

    Why would anyone ever need more than 16 gb of ram in a piny ITX system? And note that I say "NEED" not want, because the only limit to what one might want is his or her stupidity. I am a power user, I run a few workstations and a rendering farm and I can assure you even if you plan to make a gaming or workstation PC in ITX format, 16 gb is more than that system will need in its life span.
  • FunBunny2 - Wednesday, February 12, 2014 - link

    If you have a cube farm of actuaries, say, then giving them small machines with lots of memory can make sense.
  • Guspaz - Wednesday, February 12, 2014 - link

    While this is true (and awesome), even a powerful machine is hard pressed to benefit much from more than 16GB of RAM. That's not to say that I don't have a file server running 24GB of RAM sitting here in my apartment (using consumer hardware rather than server), but it's most assuredly not mini-ITX (with 15xHDD and 3xSSDs, I can't lift the damned thing).
  • MRFS - Thursday, March 12, 2015 - link

    We use a 12GB RamDisk Plus on a 16GB PCIe 2.0 machine, and pre-load into the ramdisk a 10GB database that is growing steadily. We'd prefer to leave the lower 4GB to the OS, so we'll need more RAM before too long. Searching and browsing the ramdisk is a real treat, and updating the COPERNIC index is also very fast, even though the RAM is a matched quad of Corsair DDR2-800. With die shrinks, single 32GB DIMMs are on the visible horizon. p.s. In the mid-1980s I predicted video on-demand, and people laughed at me.
  • mobutu - Wednesday, February 12, 2014 - link

    @ddriver "no one would purchase a consumer mobo with 2 memory slots only to put 32 gigs of memory in it."

    I'm a consumer and I'd buy 32GB for my powerful miniITX in a heartbeat, which is limited to 16GB atm.

    So don't speak nonsense
  • ddriver - Wednesday, February 12, 2014 - link

    Mind telling me what kind of an application you put that board into so that you NEED 32 gigs of ram? Or maybe just moar ram makes you feel moar like a man? If anyone speaks nonsense here, then it is you. The ONLY scenario where an ITX system will make any use of more than 16 gigs of ram is server, and there you already have ITX boards with 4 slots supporting up to 64 gigs of ecc ram.
  • damianrobertjones - Wednesday, February 12, 2014 - link

    VMWare
  • mobutu - Wednesday, February 12, 2014 - link

    virtual machines + ramdrives + gaming all in a tiny yet powerful system
    If you're not doing it then it doesnt mean nobody is doing it.
    reread my previous advice: don't speak nonsense

    Been waiting for some time for 16gig sticks, it looks now it is going to happen rather sooner than later. GREAT!
  • ddriver - Wednesday, February 12, 2014 - link

    Yes, because the world is full of gamers who run multiple virtual machines and ramdrives on living room media boxes... Take your own advice and apply it to yourself. If you need loads of ram, ITX is generally not the place to look, but even if you do need ITX, there are already boards which support up to 64 gigs of ram, plus all the enterprise features you will likely need in such a scenario that are not found on consumer ITX boards.
  • heffeque - Wednesday, February 12, 2014 - link

    It's not a matter of needing miniITX, it's a matter of choosing to have one. Not everybody wants a huge noisy machine as their working and/or home and/or gaming PC.
  • ddriver - Wednesday, February 12, 2014 - link

    Is there any particular reason you need the PC in the same room? I keep desktops/workstations/servers in a dedicated room - no noise no heat no clutter.

    BTW - huge machines are not necessarily noisy. Non-ITX machines are not necessarily huge. The rest is just drivel by people "amazed" by the possibility of putting an unnecessarily amount of memory in a machine constrained by its footprint., 99.99% of which will never really do it.

    When will kids learn that having lots of ram in your computer doesn't make you cooler, especially when you really have no use of it?
  • Guspaz - Wednesday, February 12, 2014 - link

    As the owner of an mITX machine (well, OK, a shuttle XPC that's got a larger-than-mITX motherboard but is still smaller than many mITX cases like the Prodigy), the thing is bloody loud under load. Why? Because tiny PCs tend to have smaller fans, and smaller fans need to spin much faster (and louder) to provide enough airflow.

    The fan in mine is a 92mm fan that acts as both the CPU fan and casefan, and it spins at 3850 RPM at full speed. That is *NOT* quiet. Sadly, I tried a Noctua 92mm fan, and it utterly failed to adequately cool the system.
  • hardwickj - Wednesday, February 12, 2014 - link

    macminicolo.net, macminivault.com….etc etc, which colocate thousands of mac minis alone. I could go on and on. The vast majority of people who colo their Mac Mini's are *eager* for 32GB of RAM. Unfortunately due to the Intel JEDEC bug, that will now never happen, or at least not until Broadwell based Mac Minis come out.

    I've seen you cry about this on other forums. Just because you don't have a need or are incapable of understanding others needs doesn't mean there *isn't* a need!
  • hardwickj - Wednesday, February 12, 2014 - link

    ddriver, let me give a simple example. Mac Mini quad-core w/ 32GB of RAM and ESXi installed making 4 VM's, each with a hyper threaded core and 8GB of RAM. Great little virtualized and isolated server environments. There are *tons* more potential applications, IF these things had the memory capacity.
  • LarsBars - Wednesday, February 12, 2014 - link

    +1
    Windows 8/8.1 Pro also have Hyper-V (what I use for a lab environment) and having 16GB of consumer memory per dimm is perfect for that. It also allows the rest of your PC to be cheaper when you don't need all the bells and whistles of server class hardware. I bought a mobo with 4 dimms / 64GB max ram for this purpose, and it was nice and cheap!
  • Bleakwise - Friday, September 11, 2015 - link

    Wrong.

    You'll pay a much higher premium for 16gb dimms than you would 8 dimm slotts. Your average 8 dimm slot board is about 50$ extra, while each 16gb dimm will be about 100$ more (400$ total) than it's 2x8gb kit counterpart.
  • TiGr1982 - Tuesday, February 11, 2014 - link

    It is indeed disappointing, that Intel does not support 16 GB unregistered DDR3 modules with the current CPUs. Maybe it will with Broadwell, but I think it definitely will for future DDR4 platforms.

    But just the single $320-350 16 GB module costs like twice the price of an AMD processor. And with only one module that's single channel RAM, so you need two of these modules to run in dual channel, making the RAM price 4 times the CPU price. IMHO, it's not worth it pricewise.
  • RoboJ1M - Wednesday, February 12, 2014 - link

    It also means you can build 32GB ITX Kaveri boxes where you're limited to 2 RAM slots.
  • ddriver - Wednesday, February 12, 2014 - link

    Hurray, it has been a while since I was able to do something completely pointless.
  • sligett - Thursday, February 13, 2014 - link

    I think that's what you are doing here by continuing to argue this point. Neither side will "win". And it would be a boring world if we all agreed on everything.
  • ddriver - Thursday, February 13, 2014 - link

    I don't do it for a purpose, I am just chilling in between work hours. :)

    And at any rate, KAVERI with 32 gigs of ram? I'd understand if it was at least some high performance CPU, but this utter joke, 32 gigs for kaveri is like buying a mansion for your dog to live in, when all it needs is a dog house.
  • hrga - Friday, April 18, 2014 - link

    It seems like you never heard of longly promised AMDs HSA. So at any rate these *superb APUs* could always find some use even with 128GBs of installed ram.
    I thou wondering how you never mention insanely high latency times could actually damage high end performance. And they probably become even worse when you install more than just single module per channel.
  • Bleakwise - Friday, September 11, 2015 - link

    Just get a 4 dimm slot board then. Instead of buying 2x16gb ram for 600$ just buy 4x8gb for 160$ and. 4 dimm slot boards cost the same as 2 slot boards, so no difference there. Sure, you could also spend 50 million dollars turning a 1960s Volkswagon Bus into a competitive offroad getaway car with a large cargo, or for half the money you could buy a cargo plane and get the damn job done proper.
  • Bleakwise - Friday, September 11, 2015 - link

    Yeah, because spending 1000$ for a kaveri system over an i7 with four dimm slots makes perfect sense!

    Yes, if you're paying 200$ extra for dimm slots that's what you're doing.
  • TommyVolt - Wednesday, February 12, 2014 - link

    "640k ought to be enough for everybody" (Bill Gates, 1981)
  • Primoz - Saturday, February 15, 2014 - link

    This. I was thinking of this quote the whole way through the thread.
  • hrga - Friday, April 18, 2014 - link

    Some find it priceless (While actually forgetting how many miliseconds has passed since this so
    famous quote was brainstormed)

    While others made trillions of profit based on these well marketed misconceptions
  • Lone Ranger - Wednesday, February 12, 2014 - link

    A few people here are implying that Intel screwed up by not implementing support for the larger DIMMs, even though it is in the JEDEC spec. It very well may be intentional, as to not cannibalize their low end Xeon sales (E3 series).
  • KAlmquist - Wednesday, February 12, 2014 - link

    The article indicates that the problem occurs on all Intel processors using DDR-3 except for the Atom models (which presumably use a different memory controller).

    The maximum memory sizes specified for various Intel processors seem to assume DDR-3 DIMMs no larger than 8GB. LGA1155 and LGA1150 models (including the E3 series) are specified as supporting a maximum memory of 32GB (2 channels times 2 DIMMs/per channel times 8GB per DIMM). Similarly, LGA2111 chips are specified as supporting a maximum memory of 64GB (4 channels instead of 2). So it would seem that the lack of support for 16GB DIMMs is at least partially intentional.
  • TiGr1982 - Thursday, February 13, 2014 - link

    That's a good point; as an argument in favor of that, Intel intentionally cuts ECC memory support for Core i7 to prevent using i7's instead of Xeons.
  • bobbozzo - Wednesday, February 12, 2014 - link

    Anyone know if these will work in the HP MicroServer Gen8?

    Someone tried some 16GB UDIMMs designed for an Apple workstation and they didn't get past POST.
  • TommyVolt - Thursday, February 13, 2014 - link

    HP Microserver Gen8 uses Intel CPU....won't work. But I recommend to write an email to HP telling them about these modules and asking if they would support in the future.
    I think HP has the power to make Intel move!
  • Lone Ranger - Friday, February 14, 2014 - link

    There is a key combo during POST that bypasses memory checks on some HP servers. Need to search around....
  • mattlach - Thursday, June 26, 2014 - link

    I am eagerly looking forward to buying 4 16GB ECC UDIMM's for my basement VMWare ESXi server, but I can't find these anywhere, and the distributors aren't responding to my emails (presumably because I am a small volume individual buyer)

    Has anyone found these for sale ANYWHERE?
  • kjboughton - Thursday, July 31, 2014 - link

    ASUS have just released a beta BIOS for the X79 DELUXE that is intended to address the general incompatability with Intel SBE/IVE.

    http://dlcdnet.asus.com/pub/ASUS/mb/LGA2011/X79-De...
  • q.epsilon.p - Saturday, April 18, 2015 - link

    Obviously these are going to be stupid expensive and unless you need 64GB on a 4 slot dual channel bored or 128 on a 8 shot quad channel bored it's completely not going to be worth it.

    But I'm not doing a total system upgrade to early 2017 when the Cannonlake successor will be launched, as well and AMD Zen will have been out for a bit. Although if the leaks are accurate it looks like AMD is taking the same dumb path they took with Piledriver and packing a s*** ton of cores on a single die rather than making more powerful cores.

    Which now we are talking about 16 cores on there APU's and probably 32 cores on there pure x86 processors... which is dumb and I hope the leaks are false as it makes NO sense to have so many cores with hyper threading, meaning with the pure x86 processor it would have 64 threads which is dumb as f***.

    We are still having problems with supporting 8 threads in a most applications and if the leak is accurate they want to jump to 64 threads... this is even more stupid than piledriver if the leak is accurate they should be working on IPC not increasing the number of cores sticking with 8 cores and doubling or increasing the intructions per cycle by a little more would be what they need to do.
  • Bleakwise - Friday, September 11, 2015 - link

    Completely disagree. For one, they are working on IPC. Second, 64 threads makes complete sense for many many workloas, especially on the SERVER, and the last time AMD was profitable was when they were competing with Intel in the server space. Guess what the #1 generator of revenue is for Intel? I can assure you it's not gamers buying their k model i5.

    While IPC will be boosted, even 32 or 64 piledriver threads would be AMAZING for things like Cinibench, content creation, compilation, level design, 3d modeling, sound production, etc etc. Look at say Cubase or Pro Tools, they need CORES not IPC. Or large Visual Studio projects, hundreds of threads get used in build clusters. Look at CryEngine or UE4 or Unity, hundreds of cores go to work on baking lighting.

    That said, even in gaming there are many workloads that are thread-bound. Physics for example is being held back by a lack of CPU cores as a complete Physics pipeline on GPU is simply not possible, just eye-candy, because rigid body and game state are too coupled.

    AMD is going to be putting large HSA enabled APUs on their Zen cores also, R9 graphics with an HBM cache acting something like the Xbox One's ESRAM. People will be able to buy an AMD APU and game on it, done and with DirectX 12 that is a bigger deal than it was, and being more powerful than consoles they'll be able to play every console port from this generation-- all on a budget.

    That's the hype anyway, and they could deliver what they're saying, we'll see if AMDs execs have remained completely out of touch with the market. TBH they could have delivered most of this stuff years ago, all they needed to do to take over the desktop/laptop market was put something like the Xbox One's ESRAM on their APUs and increase the GCN core count to PS4 levels, made them HSA enabled and sold them for 200-300 bucks, things they have already done but never made into a consumer product. Hell, they didn't even have a R7 gaming laptop APU until 2015 when even without an ESRAM cache that could have completely taken over low power gaming laptop market, boggles the mind.
  • Bleakwise - Friday, September 11, 2015 - link

    I mean, Intel basically jacked the Xbox One's ESRAM idea, stuck it on their HD 5000 series igpu and put it out there a generation ago with Broadwell, something AMD invented and still hasn't done.

    With HBM, AMD could even be doing it bigger and better. I'd love to see an L4 HBM CPU cache on AMD FX motherboards, that could have been a whole platform, and they'd have made a killing not only selling chips and motherboard chip-sets but stacks of their new memory when motherboard makers started offering not only say 256MB chips but 2-4GB L4 caches and such. Or hell, sell PS5 style GDDR5 motherboards! They have their own Memory R&D division, the tech done for the PS4, chipset, and a great relationship with motherboard makers so why the fuck not?

    It really is amazing to see AMD squander so much of their tech, 80% of the time it's their competitors who figure out how to use their own tech, such as Nvidia/IBM with unified memory in NVLINK. AMD literally invented GPGPU and unified memory, but it's Nvidia and IBM that took their technology and made a high end hpc/server/desktop product out of it, while AMD is just pushing a budget APU for playing indie games.

    AMD has world class R&D and engineers but their board of directors and execs are completely out of touch, you have to wonder if they're even technical people.

Log in

Don't have an account? Sign up now