Overvolted RAM could be a no-go on Core i7 PCs

For years now, memory vendors like Corsair, OCZ, and Kingston have been marketing “enthusiast” memory with ratings far beyond JEDEC specifications. But there’s a catch: most of those memory modules need a little extra voltage to run at their rated specs. That usually doesn’t pose a problem on decent motherboards, but it could become a serious issue once Intel’s Core i7 arrives.

As reported by Custom PC, Asus’ upcoming P6T Deluxe motherboard has a sticker covering its DIMM slots that says, “According to Intel CPU spec, DIMMs with voltage setting over 1.65V may damage the CPU permanently. We recommend you install the DIMMs with the voltage setting below 1.65V.” Asus spokesman Iain Bristow talked to Custom PC about the issue, and he added that “the sticker had been placed on the motherboard after Intel’s recommendation.”

The memory controllers inside Core i7 CPUs support DDR3 memory, and JEDEC specifies a standard voltage of 1.5V for that memory type, so this 1.65V limit would leave very little leeway for overvolting. It could force consumers to shop carefully for their high-end DDR3 RAM, too. A quick search for DDR3-1600 modules at Newegg today shows only one of the 20 cheapest offerings with a voltage rating below 1.65V. Many other modules need up to 1.9V to reach their rated speeds.

With that said, Core i7 users probably won’t run into too many bottlenecks. Not only does Intel’s Nehalem architecture promise significant clock-for-clock performance improvements over current processors, but the first Core i7 variants should support three channels of DDR3-1333 RAM—all on a low-latency, integrated memory controller.

Comments closed
    • PetMiceRnice
    • 12 years ago

    I’ve been using Kingston Value RAM for the past six years (yes even on my gaming computers), so I’m not worried at all about this. Hardware comes down in price so quickly that overclocking doesn’t mean anything to me. If anything, I’ve always experienced weird glitches anytime I overclocked in the past. Not worth it as far as I’m concerned.

      • MadManOriginal
      • 12 years ago

      I like to do the ‘easy oc’ route. Just keep stuff within reasonable specs and don’t fiddle with eeking out the last 5%. For CPUs for example, I’ll just try to get at least the speed of the highest member of the same CPU line, this can often be done with little effort.

    • Neutronbeam
    • 12 years ago

    sorry for the duplicate post!

    • Neutronbeam
    • 12 years ago
    • clone
    • 12 years ago

    all of the ram I buy including the overvolted comes with a lifetime warranty, the motherboards come with 3 years to lifetime warranties and the cpu also comes with 3 years.

    if motherboard makers offer overvolting for the ram I assume they have designed the motherboard to compensate for the increased stress…. it’s not going to affect the electricity bill in any significant way, if Intel’s cpu’s can’t handle high volted ram then so be it I’ll factor that limitation into the equation when choosing my next cpu.

    if say for instance AMD finally releases a CPU to compete directly with Intel’s cpu that doesn’t care about ram voltage then I’ll go AMD if Intel’s can overcome the ram limitation then I’ll go Intel and possibly save some coin on ram in the process….. likely not enough to make up for Intel’s absorbent pricing but whatever the goal is perf and a little saved here but spent there means nothing.

    as for the industry “practice” of overvolting ram…… bleh don’t care their is plenty of ram on the market running at spec’d voltage and if you don’t want overvolted ram then don’t buy it, if you are complaining about a lack of selection your arguement is like complaining that you hate any car that isn’t blue then whining about only being able to buy blue cars.

    I and others are looking for best performance if my ram requires 2.1 volts to operate compared to 1.65….. wow!!!! how could I possibly care so long as the surrounding components were designed to allow for it.

      • sigher
      • 12 years ago

      Good job of a long post showing no knowledge or understanding of what you are talking about.

      • LoneWolf15
      • 12 years ago

      And the point of your rambling post was what?

        • Meadows
        • 12 years ago

        He lost me too, pretty quick.

    • Forge
    • 12 years ago

    GOOD. Maybe this will serve as a wake up call to the hundreds of folks shipping PLAIN OLD DDR2-800 kits requiring 2.1 to 2.4V. It disgusts me that I have to make “1.8-1.9V” one of my primary criteria when shopping DDR2, and I get half the selection or less that others do.

    Sure, they’re just the guidelines, but when your product requires the maximum an enthusiast mobo provides OR MORE just to run at the official speeds, UR DOIN IT RONG.

      • MadManOriginal
      • 12 years ago

      Well if you’re just looking for JEDEC standard RAM you should be looking at lifetime warranty value RAM only anyway. There are a few things to consider about the rated overvoltage RAM. First, they are often rated for lower than standard latencies. Second, they often run at their rated speed and latency at lowered voltage but that requires testing and isn’t guaranteed, the flipside to that is they will often run at faster speeds and/or timings with their rated overvoltage.

      I agree though that’s it’s silly to see RAM with standard JEDEC speeds and timings to state it requires more than 1.8V. but there are plenty of sticks that fill that requirement so it’s not a huge deal imo.

      • Kent_dieGo
      • 12 years ago

      Overvolted RAM is fraud. Many motherboards cannot overvolt RAM. I only buy RAM that can operate at the specified frequency at stock voltage.

        • DrDillyBar
        • 12 years ago

        Exactly.

          • clone
          • 12 years ago

          moved on it’s own.

    • Corrado
    • 12 years ago

    meant as a reply

    • Krogoth
    • 12 years ago

    Could be simply be because part of the voltage circuit is on the CPU itself?

    It would give the CPU an additional area of hardware failure due to running out of spec.

    Anyhow, Intel and their motherboard partners are just covering their collective legal and warranty arses. They are saying basically “Do memory overvolting at your own risk, it is your fault if something breaks”.

    I do not see how this will stop the hardcore overclocking crowd. They do it anyway, know the risks quite well and can afford the “loss” for whatever reason.

      • Meadows
      • 12 years ago

      Everyone says that. AMD’s OverDrive application, for example, has a huge “at your own risk” notice coming up every time you start it.

      • sigher
      • 12 years ago

      The i7 is the first intel CPU with on-die QPI/memory controller, that means the RAM is directly connected to the CPU where it used to be connected to the chipset and then to the CPU, that’s the reason why it gets tricky now, if you don’t listen to intel you blow up your CPU.

      And seeing everybody forgot, AMD did in fact have the same issue and same warning when they started with an on-die memory interface, but they decided to fix it and make the interface more robust.

      Now as for the issues for the consumer, first off it becomes harder to OC the system, but already the RAM manufacturers have put 1.5v OC modules on the market so that’s not an issue that will be there long, but the other issue is that you might have bought expensive DDR3 RAM thinking you can use it for the i7 when it came around only to now hear you’ll have to run it at 1.5 volt and very slow, so you partly wasted that cash.

      There are obviously also advantages, since the RAM is directly connected to the CPU it drops latency and speeds things up (and can in fact make the chipset cheaper theoretically since it doesn’t have to handle that anymore), add the new triple channel thing and QPI to it and you will see a good boost in the speed the CPU gets its data, and all that makes the rumours of the top of the line i7 being about 50% faster than the top of the line predecessor at the same clock speed actually possible.

    • eitje
    • 12 years ago

    i agree with most of the folks saying that low latency is where the value is.

    • srg86
    • 12 years ago

    I’d only use Low latency DDR3-1333 RAM with a Core i7 machine either, I’ve never really seen the point of this fancy non standard RAM.

      • Usacomp2k3
      • 12 years ago

      Low Latency is over-rated anyway, at least for real-word usage (especially gaming).

    • bcronce
    • 12 years ago

    the problem is that 45nm silicon is NOT meant to deal with high voltages and being an integrated memory controller leaves the controller at the same processes as the cpu. Trying running a 45nm cpu at 1.9v and see what happens.

    edit:
    don’t forget, as memory cycles go up, their relative latency goes up up. Your bang per clock drops dramatically. Anyway, tri-memory channel is plenty fast for anything. One of the hardware review sites did a review of many different memory speeds and all that pre-caching cpus have now days left memory that was 33% ‘faster’ with under 5% improvement

      • srg86
      • 12 years ago

      Maybe this is one reason why Intel uses older processes for their chipsets.

        • Bigbloke
        • 12 years ago

        Maybe, but much more likely to be cost (older tech much cheaper), and design cycle time (older tech quicker to deisgn/layout and fabricate).

      • slaimus
      • 12 years ago

      Well AMD is going to be releasing DDR2 45nm CPUs in the form of Deneb. DDR2 is still 1.8V minimum, so Intel’s 1.65V claim is still questionable, unless SOI makes a big difference.

        • clone
        • 12 years ago

        that depends on how DDR2 works with the cpu compared to DDR3….. will the same AMD cpu work with both memory types or will their be 2 different cpu’s one for DDR 2 and one for DDR 3?

    • Bigbloke
    • 12 years ago

    This is a process issue. As the chips get smaller the voltages they use and can work with get smaller. Notice how the core voltage gets smaller with every process shrink? At 45nm you don’t need many volts to cause punch through of the silicon unless you put lots of protection diodes etc on each pad of the die. These protection measures tend to slow down the signals, so to keep the memory controller as fast as possible they need to be minimised.

    Intel is just keeping things very safe by this 1.65V limit. You can go higher but it will be easy to latch-up and even damage these signals.

    And I’m sure the 45nm AMD devices will have similar issues/restrictions. Just Intel got there first!

      • Meadows
      • 12 years ago

      Doesn’t matter who gets there first, in my opinion.

        • Bigbloke
        • 12 years ago

        Yeah whatever :rolleyes:

    • Faiakes
    • 12 years ago

    Isn’t this article (on the same page) taken into account?

    §[<http://www.fudzilla.com/index.php?option=com_content&task=view&id=9824&Itemid=1<]§

    • Corrado
    • 12 years ago

    The key words there are ‘MAY’ and ‘RECOMMEND’. Intel and AMD also RECOMMEND that you don’t over clock your cpu, they recommend you don’t over-volt your CPU, but since AMD sells a CPU that is mult-unlocked, you can see they don’t forbid it. Its just not supported or encouraged.

      • Usacomp2k3
      • 12 years ago

      The Extreme chips are unlocked too, aren’t they?

        • Corrado
        • 12 years ago

        Not sure. I don’t pay attention to single chips that cost more than my whole PC, and I’ll never own. Hehe.

        • Veerappan
        • 12 years ago

        Yeah, fairly sure the extreme chips are unlocked… but you pay dearly for them.

    • maroon1
    • 12 years ago

    Who cares ?

    Core i7 use tri-channel ram, so DDR3 1333 should give you the same memory bandwidth as DDR3 2000 in dual channel

      • Meadows
      • 12 years ago

      Who cares? The last time I was RAM-bound in a task was a RAM benchmark.

        • MadManOriginal
        • 12 years ago

        People looking for a performance jump through overclocking will care if this recommendation keeps them from running RAM fast enough to get a decent CPU overclock.

        I need to read up more on Nehalem architecture and speicifcally bus architecture though to see if it’s something like AMD’s integrated memory controller in that CPU speed and memory speed are more independent.

          • Meadows
          • 12 years ago

          The point is, you shouldn’t need to overclock memory at all, just keep reducing its multiplier or whatever works for Core i7 and keep it within specs. It’ll be “fast enough” anyway, especially if you had the money to go tre-channel with RAM.

            • MadManOriginal
            • 12 years ago

            Yes that’s what may be possible with Nehalem’s bus architecture and what I’m not certain on and won’t worry about unttil reviews are out, I won’t be buying Nehalem until at least the middle of next year anyway. But remember that lower than 1:1 ratio is not possible with the last few Intel chipsets so this could put a damper on oc’ing if the ratios with <1.65V RAM don’t work out in oc’ers favor.

            • CheetoPet
            • 12 years ago

            Since when is anything ‘fast enough’ for overclockers? That misses the entire point of overclocking.

            • MadManOriginal
            • 12 years ago

            He probably means the memory part of the system being fast enough. Memory oc’ing above a 1:1 ratio with the FSB does tend to yield very small performance advantages if at all. But it’s anyone’s guess how that will play out on Nehalem.

    • Usacomp2k3
    • 12 years ago

    I’d say a good portion of motherboard or RAM RMA’s are probably due to non-1.8V dimms trying to be booted and not working. This move makes sense.

      • Meadows
      • 12 years ago

      No, they’ve tipped the scale far too much here. While “enthusiast” gear was the source of some issues previously, now they’ll effectively prohibit such advances altogether quite harshly (by giving a CPU that explodes if you overvolt the RAM).

        • Usacomp2k3
        • 12 years ago

        Which is your opinion. I personally think this is a fine move. We don’t need much faster than what we have anytime soon. I’d rather not have to worry about it, and not have other gerbils that suffer from this.

        • Firestarter
        • 12 years ago

        I don’t think that overvolting to get higher speeds counts as an advancement. It has its merits, but its certainly not something that should limit Intel in the design of their memory controllers.

        Intel probably has a good reason to warn people not to use out of spec memory modules.

Pin It on Pinterest

Share This