I'm looking at some Mushkin Blackline DDR 1600 for a new build. Mushkin offers two identically spec'd 8GB (2x4GB) kits with the only difference being that one is spec'd at 1.5V (p/n 996995) and one is 1.35V (p/n 996998). Is there any real-world advantage to the 1.35V ram for an extra $5?

http://www.mushkin.com/Memory/Blackline/996995.aspx

http://www.mushkin.com/Memory/Blackline/996988.aspx

Runs cooler, don't have to deal with gnarly tall heatspreaders, more likely to be compatible with Haswell next year.

I though haswell was going to use DDR4.

Both of them use the same low-profile heatsink. I get the impression that the silly looking 'gnarly tall heatspreaders' are 90% for show, anyway. Like neon case lighting and glowing cables.

The lower heat and energy consumption is actually significant?

In a datacenter with thousands of sticks of memory, the power savings can add up. For a home user, however, I would imagine the power savings would be less than a few watts.

Haswell-E will. Haswell is set for DDR3 and DDR3L.

They are a 100% for show. Or for factory margin. To have those rams to work on lower voltage, your mobo should support running them at lower voltages.

I believe (not having done the math) that you will not see a significant difference. Yes, they use less power. No, i doubt that a kill-a-watt is precise enough to catch any difference.

€: On second thought: Are the lower voltage ones only supposed to give more oc headroom in case you up the frequency AND the ram voltage?

My personal preference is low-voltage, both for the principle (power efficiency enables silence) and the perceived quality (efficiency comes from better design). The cost difference has been either negligible or absent.

Practical, noticeable advantages are few, although that point about Haswell is a new one to me (thanks CA_Steve). DDR3 is so low power to begin with, and I wouldn't know if they're much better for overclocking - but I have gotten the impression from reviews and discussions over the years that it takes better design (from wafer to chips to control and power components) to get to low voltage rather than high. Case in point is memory that only achieves "regular" timings at "regular" frequency while overvolted (say, DDR3 CL9-9-9-24 1600 MHz at 1.65 V); one reason why I started steering away from Kingston and Corsair.

So the question to me really is: why pay for inefficient design - maybe even inferior components in case of overvolted kits - when you can get efficient design for essentially the same money. Why indeed - go for the L or U kit if your motherboard supports it!

If installing 16GB of DDR3 1600, which would draw less power - 4x4GB @1.35V or 2x8GB @1.5V?

It is module-dependent, but to get you in the ballpark we're talking here: http://www.xbitlabs.com/articles/memory ... html#sect1. Pappnaas' hunch was pretty much right, as the difference is there in that test with 2x2, but it was barely measurable.

As for this case, more sticks to me suggests more draw. The difference in mW per stick should not be dramatic, certainly not double, between 4 GB sticks and 8 GB from what few specs I've seen. Motherboards have in the past been more reliable with fewer slots loaded, and it's arguably easier to install 2 rather than 4 modules, as coolers overlapping memory slots is common (big, quiet coolers and compact ones being the norm here), so I would argue 2x8 is the way to go for convenience's sake, even if you can't go DDR3L or DDR3U in that case.

PS. If you REALLY want to find out what's what, Micron is one place where you can find actual spec sheets of modules with readings down to the µA. In the meantime I'd just google for more system power tests.