Power requirements in PCs have grown enormously in the last few years. Whereas once we could survive with 300 Watt power supplies, and if you had a really beefy setup maybe 500 Watts. Now most of us have power supplies either at or near to the 1000 Watt mark in our ever greater need to satisfy our lust for more performance and feed our behemoths that give us the eye-candy we love so much.
However you'd have to live under a rock not to have noticed the global shift in recent years away from hedonism and towards a more conservative approach that is designed to retain what little is left of our planet after the industrial revolution.
To this end manufacturers are starting to introduce components that place a much lower drain upon the energy requirements of your house or office. One such company is Kingston who have been in the memory marketplace for as long as it's existed. Today we're looking at their low-voltage DDR3 kit called, in a moment of inspiration, the LoVo.
So what have we in our metaphorically sweaty mitts today? A nip to the Kingston website to download the datasheet finds the following.
JEDEC standard 1.5v
16 128M x 8-bit DDR FBGA per module
Programmable CAS latency of 6,7,8 or 9
Programmable Additive latency of 0, -2 or -1
8 independent internal bank
Bi-directional Differential Data Strobe
667mHz fCK for 1333Mb/sec/pin
XMP 1 : DDR3-1866 CL9-9-9-27-2T @ 1.35v
XMP 2 : DDR3-1600 CL9-9-9-27-2T @ 1.25v
The main thing we're interested in is the XMP profiles. We've tested many kits through here that required 1.65v to reach 1600MHz, so to see 1866 at reasonable timings only using 1.35v is very impressive.
As you can see we've got four sticks available to us today. We'll be running all four in our tests and at the end will give our thoughts on how 8GB fares compared to 4GB. Packaging is standard for RAM sticks with lots of big labelling to show you what they are rated at, and otherwise not much.
The sticks themselves are great to look at. The main three "default" modding colours have nearly always been red, green and blue, but over recent times green has taken a back-seat with blue being the mainstream colour and red being used for more enthusiast parts. Meanwhile poor old green has just sat idle. With this being a low-volt, or I guess ecological, kit, Kingston have brought the green out and it's very swish.
The heatsinks themselves are standard height so those with huge coolers don't need to worry. When you're running at 1.35v then isn't much heat to dissipate.
Test Setup and Overclocking
Intel Core i7 870 @ 3.6GHz
ASUS Maximum Extreme III
Windows 7 Ultimate 64
8GB Kingston LoVo 1866 (4x2GB)
For our tests we ran the LoVo at the two XMPs, being 1600MHz @ 1.25v and 1866MHz @ 1.35v.
With something that comes so low-voltage out the box it's very tempting to run it hard. After all it has to remain within ATX specifications and it's more likely that you need to use good chips to get such low voltages whilst remaining stable, than you can use ones out the parts bin that wont do higher speeds.
However gains with RAM don't only come from higher clock speeds but also lower timings. Striking the balance between the two is the key. As we've hopefully got a lot of voltage overhead we upped from the 1.25v XMP setting to 1.45v and tried to see if we could get the timings lower than the 9-9-9-27 they run at stock. As it was we easily hit 7-8-7-24 which is a huge improvement. These definitely are some good chips.
Having lowered the timings we clearly have a lot to play with here. So it's obviously worth putting them back to the 9-9-9-27 and seeing how far we can push the speed.
Naturally with the Kingston LoVo rated at 1866 @ 1.35v we were careful when overclocking to go up in small steps and constantly test stability and temperatures. To save you reading a ton of results though, here are our final results. 2202MHz 9-9-9-27 @ 1.65v, completely stable. Fantastic results.
Benchmarking Part One
Everest Ultimate Edition
Everest gives a clear demonstration how certain aspects are better than others when it comes to memory. Clearly the extra speed of the LoVo at 1866MHz wins the day, but between the G.Skill ECO with CAS7 and the LoVo with CAS9, both at 1600MHz, the tests vary, although still near to each other. The higher read and write speeds of the G.Skill is confusingly beaten by the copy speed of the LoVo.
Sandras bandwidth testing relies far more on pure speed, and the graph clearly shows the pecking order we'll see throughout most of the rest of our testing. At its rated speed the LoVo clearly wins, and at the lower voltage 1600MHz setting it's still got enough to just edge the G.Skill.
Similar to the Sandra results, but much less dramatic, POV-Ray gains a few rendering PPS as the speed grows, and even at the same clock speed the Kingston bests the G.Skill.
Without wanting to end up typing the same thing all the time, for the CineBench results, see the POV-Ray ones.
Whilst insanely consistent results are perfect in a review, they don't make it easy to say anything different. The LoVo takes first and second, with the G.Skill just a hair behind the Kingston when at the same speed.
Benchmarks Part Two
Let's move on to the 3D benchmarks and hopefully get at least something to talk about.
PC Mark Vantage
PC Mark Vantage definitely shows the consistency of this kit, with the graphs following a very predictable pattern.
3D Mark Vantage
Moving to 3D tests we'll hopefully see a little variance if only because bringing graphics into the equation always gives us more chance for the odd fluctuation.
Not here though. This Kingston is bullet-proof.
Still able to find the flaws in any setup, Warhead could be expected to show up the odd problem, but although there is a small variance in the average fps, which is what matters, the Kingston still sits atop. 2FPS from such a small speed increase is excellent.
Dirt 2, if we allow that 1fps is within tolerance, gives identical results across the board. Although it has many tweaks for the DirectX 11 API, the fact the bulk of it is designed for a console shines through.
Shift is much more pliant. Even if you did a hundred back-to-back runs with identical hardware you'd never get the same results twice. That is clear to see here with the 1600MHz LoVo having the highest minimum fps like it did in Dirt, when at 1866MHz it clearly wins, and yet when we get to the average results we have the same pattern we've seen throughout the last two pages.
Consistent I think best describes our results today.
Very little, if anything, has given us such a comprehensively predictable set of results across our testing. Normally a product will be good in certain areas and possibly not quite so good in others. It's why we test using the software we use.
However the Kingston LoVo stood proudly astride our graphs no matter what we tested. The strangest results actually came when it was using the XMP1 profile to go up against the only other low voltage kit we've tested, the G.Skill ECO. The G.Skill is CAS7, whereas the Hyper-X LoVo is CAS9, yet the Kingston was still just ahead by a nose in nearly all our tests.
Overclocking performance was nothing short of brilliant. It would be tempting to assume that just because this is low-voltage it might not be quite up to the task of running as hard as its more performance orientated brethren. However this was miles from the case as we got it up to 2200MHz whilst remaining within ATX specifications.
Before we wrap up we need to say a little something about the 8GB goodness compared to the 4GB that we normally test on our P55 platform. Memory is, to a small degree, one of those things that has diminishing returns. If you've only got 1GB then 2GB will be infinitely smoother. Equally the leap from 2GB to 4GB has noticeable performance improvements too. Once you cross the 4GB thresh-hold it's harder to spot those differences unless you're working with immense file-sizes and six or seven things at once.
Thankfully we often do that here at OC3D. When we're collating all our results and things it's not a shock to find four or five explorer windows, our graphing program, Excel and Firefox open, along with Photoshop and about a dozen massive images. Whereas we normally do this on a 6GB equipped LGA1366 rig, but for this review we used the rig we'd benched on. With 4GB there was definite performance reduction with everything open. However with 8GB it was a breeze. Not a lot more than with the 6GB 1366, but definitely more than with 4GB on a P55.
The pricing makes the conclusion a little stickier and less clear cut. 4GB of the LoVo will set you back £150. This is compared to 4GB of the G.Skill at around £110. If you're running at the XMP 1 profile (1600 @ 1.25v) then the small performance increase it has, coupled to the very minor performance improvements, make it a very close call.
However, if you run at 1866MHz, or plan to overclock, then this is bullet-proof, and consistent in all tasks. We're happy to award it the OC3D Recommended award for performing so well under such low voltage, whilst still being an able overclocker.
Thanks to Kingston for providing the LoVo for review. Discuss in our forums.