Puget Systems print logo
Read this article at https://www.pugetsystems.com/guides/150
Article Thumbnail

Technology Primer: Low Voltage RAM

Written on July 16, 2012 by Matt Bach


Computer RAM is generally chosen based on three key factors: Speed, Size and Price. As green computing increases in popularity, power draw (reported as voltage) is becoming an increasingly important specification.

The majority of DDR3 RAM runs at 1.5V, but many performance models can run at voltages as high as 1.65V. Low voltage RAM, on the other hand, can run at voltages as low as 1.25V. In addition to simply consuming less power, RAM that runs at a lower voltage is also typically more stable since less voltage typically means less heat. When it comes to low voltage RAM, there are three main questions that often come up:

  1. Is low voltage RAM slower than standard RAM? 
  2. How much overall power savings does low voltage RAM translate into? 
  3. How much cooler does low voltage RAM run and does this affect overall system temperatures? 

To answer these questions, we acquired two different models of low voltage RAM and compared the performance, power draw, and thermal properties against our current performance RAM. For our testing, we will be using four of each of the following models:

Model Patriot Viper Xtreme Kingston HyperX Lovo G.SKILL Sniper Low Voltage Series
Size 4GB 4GB 4GB
Frequency DDR3-1600 DDR3-1600 DDR3-1600
Timings 8-9-8 (tested at 9-9-9) 9-9-9 9-9-9
Voltage 1.65V 1.35V 1.25V

Note that even though the Patriot Viper Xtreme is rated to run at 8-9-8 timings, we will be reducing it to 9-9-9 to match our low voltage RAM. This will reduce the performance, but we want to test how low voltage RAM performs compared to standard RAM of the same specification. 

Test Setup

To perform our testing, the following hardware was used:

To thoroughly test for any thermal or power draw differences between each model, we took power and thermal measurements at both idle and full load. Power draw readings were taken using a Power Angel power meter.

To fully load the system, we ran a combination of Prime95 and Furmark for ten minutes. This will cause much higher temperatures than any normal user is likely to see, but will give us a clear worse-case scenario for cooling and power draw. Temperature readings were taken from the hottest core of the CPU (Core 0 on our test CPU) as well as from the video card with a margin of error of 1°C. In addition to the thermal readings, we also used a thermal camera to look at the overall system temperatures at both idle and load. Performance benchmarks were taken using our standard set of benchmarks which includes Unigine Heaven Pro, DiRT 2, X3: Terran Conflict, Geekbench, CineBench R11 and various other benchmarks to test a wide variety of the system's performance. 

RAM Performance

After running our full gambit of benchmarks, we can say pretty conclusively that low voltage RAM performs exactly the same as standard voltage RAM. But since leaving it at that would make this section awfully short, let's take a look at a small sample of the benchmarks we performed:

Geekbench is a processor and memory benchmarking utility that is one of the more recent additions to our benchmarking lineup. The sequential read and write portion is just one of many tests, but it accurately reflects the overall results. Although the Kingston Lovo is technically to top performer, the results are all well within our margin of error indicating that all three of our test models are effectively identical for this benchmark.

Handbrake is a video encoding benchmark that is a very good indicator of real world encoding performance. In this benchmark, the Kingston Lovo has a very slight lead, but it is once again within our margin of error.

CineBench is a great benchmark to measure CPU performance, but is also decently affected by RAM performance. Once again, our results are all within our margin of error.

Unigine Heaven Pro 3.0 is a DirectX 11 benchmark that we ran with 8x AA and 16x AF to try to accentuate any performance variance. Unsurprisingly at this point, the results are all within our margin of error with less than a .5 FPS difference between the highest and lowest results.

Power Draw

We expected to see a little variance in power draw, but with the system pulling nearly 300 watts at load we expected the results to be somewhat minimal. We were pleasantly surprised to see that the low voltage RAM reduced the overall system power draw by a decent amount when under full load. The 11 and 13 wattage reduction from the Kingston and G.SKILL sticks works out to be about 2-3 Watts per sticks, for a total system power reduction of 3.65% and 4.32% respectively. For a component like RAM that only uses 1.25-1.65V of power, this is a much higher number than we anticipated, and great news for low voltage RAM.

Thermal Performance

Since we now know that low voltage RAM has a measurable impact of system power draw, it stands to reason that low voltage RAM should also run at a lower temperature. The question now is if the lower voltage will also cause lower CPU or GPU temperatures, or if it will only affect the RAM sticks themselves. First, let's look at the CPU and GPU temperature readings we recorded during our 10 minute stresstest:

Since the GPU is positioned below the RAM, it is not surprising that the GPU temperatures were completely unaffected. What was surprising was that the low voltage RAM did result in a 1-2 °C drop in the average and maximum CPU temperature. While this drop is not huge by any means, it does clearly show that using lower voltage RAM does slightly improve system temperatures. What this data does not show is how hot the RAM itself is running and how the voltage affects the various motherboard components surrounding the RAM. To find this out, let's take a look at our thermal images:

  Patriot Viper Xtreme Kingston Lovo G.SKILL Sniper Low Voltage

It's a bit hard to see in the thumbnails, but the low voltage ram definitely runs cooler than the Patriot Viper Xtreme. Specifically, if you look at the bottom portion of the RAM and the motherboard to the right of the RAM you can see that the low voltage RAM is running about 5 °C cooler. Between the Kingston Lovo and the G.SKill Sniper Low Voltage, the G.SKill is does run slightly cooler, but the difference is much less noticeable than the difference between the low voltage and the standard RAM.


We set out at the start of this article to see how low voltage RAM compared to standard RAM. We can now confidently say that low voltage RAM fares very well against standard RAM. Versus RAM of the same frequency and timings, it performs just as well as standard RAM while reducing the overall power requirements by 2-3 watts per stick. This power reduction also allows to the RAM to run much cooler than standard RAM. While this did not translate into significantly lower CPU or GPU temperatures, anything that reduces heat without sacrificing performance is a good thing in our view.

So what are the drawbacks of low voltage RAM? Certainly it cannot be all around better, or no one would ever buy anything other than low voltage RAM. First, low voltage RAM is slightly more expensive, with roughly a 10% premium (or about $4-5 per stick) at retail. The other thing about low voltage RAM is that there are reports that low voltage RAM does not overclock as well as standard RAM. If you can afford the extra couple of bucks and are not planning on overclocking your RAM, however, low voltage RAM is a great product and is something that almost all users should consider using in their computer system.

Tags: Technology, RAM, Performance, Power, Laptops

some good info here

But might be worth testing the OC's and testing playing with tightening the LoVo's timings...
another tester got a set of the older 1866 to 1908MHz at 7-7-7-20

Am trying to see what this set you reviewed can do... will find out soon enough, last night I ordered 2 sets of it ^^

Many people seem to be able to get it over 2GHz

Anyways, thanks for the review
(people seem to get mixed results with the Sniper set, seems more of a gamble than the Kingston LoVo set)

Posted on 2012-08-30 14:20:17

Yea, we thought about doing some overclock testing, but we tend to shy away from overclocking RAM. CPU overclocking has advanced to the point that even overclocked CPUs are stable and reliable (as long as you don't go too far) but RAM is much more sensitive to increased voltage. Of course, advanced users like yourself are willing to spend the effort to get that little bit of performance increase, but its not something we usually do for our customers.

As for the G.Skill versus Kingston, we haven't used too much G.SKill RAM, so I can't really comment on their reliability; but I can say that Kingston RAM is historically very reliable RAM. Over the last year we've only seen a .66% failure rate with Kingston RAM, which is ridiculously low for any PC component.

Posted on 2012-08-30 16:32:14
Andy Shieldsmith

For a 16Gb system, which would use less power and produce less heat - four 4Gb sticks of 1.35V RAM vs two 8Gb sticks of 1.5V Kingston RAM?

Posted on 2012-10-04 16:00:58

That... is a fantastic question.  I'll see what I can find out!

Posted on 2012-10-04 16:08:29

Those are good questions, and should be directly related. The more power a computer component uses, the more heat it will generate. I had a test bed already built up for another project (very similar to the system from this article but with a 1000W PSU and a AMD FirePro W5000), so I took some power reading with four sticks of the 1.35V Kingston Lovo as well as two sticks of the Kingston 1.5V. I was surprised at the result, but they each pulled exactly 180W under load.

Unfortunately I wasn't able to find any good thermal pictures showing the difference thermally, but in my experience the 4x 1.35V sticks should run just a bit hotter. They don't consume any more power so they won't contribute significantly more heat to the system overall, but the four 1.35V sticks are stacked close together which hinders ventilation. Of course, the 1.35V Lovo sticks have headspreaders, so that does help dissipate the heat somewhat.

So overall, power draw is identical. Heat-wise, the 4x 1.35V sticks will likely run a bit hotter, but not by any amount I would call significant. So I wouldn't let that be a determining factor when deciding between two 1.5V sticks or four 1.35V sticks. Basically, the 2x 1.5V give you more room for expansion in the future, while the 1.35V sticks will give you slightly better performance due to the tighter timings.

Posted on 2012-10-04 19:23:58

Would there be a significant electric bill savings for a gamer that used their system an average of 20 hours per week? I'm wondering if the extra cost of low voltage RAM in 2 sticks would pay fir itself.

Posted on 2013-03-06 17:37:05

If you are talking about using the 1.35V memory (Kingston LoVo, which we now carry) vs the normal 1.5V Kingston Value memory - then I think you are looking at about 5-6W of power savings.

5W x 20 hours = 100 watt-hours, or 0.1 kWh, per week. Over a year that is ~5 kWh, which around here is only ~$0.50. It could be more, even exceeding a dollar, depending on where you live... and the cost difference increases the more you use the computer as well. Still, you would have to use it for many years or for a lot more time per week before it made sense on the basis of cost savings alone.

Remember, though, that it also puts off less heat - which can help keep a system quieter, and that performance is improved slightly as well. Taken as a whole these reasons might plenty for someone to make the decision to go with LoVo RAM :)

Posted on 2013-03-06 19:23:14

does anyone know the voltage range an i7 3770k ivy bridge processor runs at, is there a specific point it runs best? i looked at different websites and they say it runs as low as 1.05 volts or something about that range which is quite low since most ram sticks run at 1.5, if i were to install ddr1333 1.5 volt ram sticks paired with the i7 3770k with an aftermarket cooler (no overclocking), would it run fairly cool? the best ram i found was the gskill 8gb sniper 1.25 volts ddr1600, would that be a better choice in terms of running cooler? i dont really care much on the performance, as long as that gskill sniper ram is more stable i would go with it, o and how difficult is it adjust the ram speed and volt? im a noob at computers in technically specs and im picking out a build for my friend, though he is gonna have another friend build him the pc for him, we might ask his friend if he can do that manual adjustment with the bios or whatever to get that 1.25 volts and ddr1600 settings with that gskill ram.

Posted on 2012-11-01 20:56:01

CPUs and RAM use separate voltage settings, so you do not need to try to match the CPU and RAM voltages. If you are not overclocking, your RAM and CPU temperatures will be largely independent of each other. For RAM, the biggest concern is general chassis airflow. For CPU, you want to make sure you are using a good heatsink to keep it running at a good temperature.

Typically, the lower voltage the RAM is rated at, the cooler it will be. So if you want the coolest running RAM possible, go with whatever RAM has the lowest voltage. That doesn't say anything about reliability, though. What I can tell you is that we use Kingston RAM because we have found it to be extremely reliable. The Kingston Lovo we used in this article is currently at a 0% failure rate, and the standard Kingston DDR3-1600 RAM we sell is sitting at a .2% failure rate. Those numbers are insanely good, so if you want reliability I can't recommend Kingston enough. This isn't to say that G.Skill or Patriot or any other brand is not reliable; just that I can say for a fact that Kingston is.

As for manually setting RAM timing and voltage, that is going to depend on your motherboard. The manual should have the information you need.

Posted on 2012-11-01 22:04:11

How many sticks of RAM were used? It doesn't say? I'm assuming 4 because it says it saves 2-3W per stick at load and there's a 12W difference with the 1.25V RAM.

Posted on 2013-07-11 09:57:07

I'm really surprised this hasn't come up before now, but you are completely right, I forgot to say how many sticks of RAM were used in our testing. You are also right about the number of sticks, it was four sticks of each of the three models. I went ahead and updated the "Test Setup" section to include this information.

Thanks for letting us know about the missing information!

Posted on 2013-07-11 18:55:31