We recently reviewed the Mushkin Redline DDR3 kit and found that not only was it a thing of beauty and performed well at stock settings, but it also was an overclocking beast. The Ridgeback heatsink being both low-profile to fit under our test Noctua NH-D14 cooler and also high-performing with the RAM cool even highly overclocked.
This got us to thinking. If their high-performance RAM is exceptional, then maybe their low voltage RAM will be good too. After all we've seen a few low voltage kits pass through our doors and all of them were surprising in their level of performance and ability to overclock.
Mushkin have a very low latency, low voltage 6GB kit available so we just had to see how it faired.
As you can see with an operating voltage of only 1.5v there is a cut down in speed when using the primary XMP profile with the RAM being rated as [email protected]
Low latency timings and some voltage headroom should at least see us able to push onwards a bit, but how far?
|Kit Quantity:||Triple Kit|
Onwards to have a look at the RAM itself.
Just like the Redline we reviewed recently the low voltage kit comes in a standard DIMM package with a nice green that accentuates the black Ridgeback heatsink.
Here they are in all their understated glory. A very nice matte black finish and a simple Mushkin logo are more than enough. So often companies put lairy graphics on their RAM when the reality is that once they are installed you never see it.
The standard RAM timings sticker shows that this indeed is CAS7 1.5v RAM.
We much prefer the stealthy appearance of the Ridgeback to some of the more "shouty" designs we've seen. It's a nice tough to see a Mushkin logo on the top face of the DIMMs as this is where it will actually be visible.
Intel Core i7 950 @ 3.2GHz and 4GHz
6GB Mushkin Blackline PC10666 1.5v Ridgeback
ASUS Rampage 3 Extreme
Windows 7 Ultimate 64 Bit
Prolimatech Armageddon with Arctic MX-3
We will be testing the Mushkin XMP1 against a 1600MHz Corsair Dominator kit and against 2000MHz Corsair Dominator GT when using XMP2.
Available XMP Profiles
XMP1 is, as to be expected, the default advertised speed of the Mushkin being 1.5v, 1333MHz and 7-7-7-20 timings.
XMP2 is a hell of surprise. If we remember that this is 1333MHz 'out of the box', the second XMP profile is 1.65v, 2000MHz @ 8-10-8-27 timings with a 143BCLK. There are many kits labelled as high end that have this performance at these kind of timings so we can't wait to see how it fares.
The great benefit about this Mushkin DDR3 Kit being able to run at 2000MHz is the ability therefore to run with a 200BCLK and so have the processor at the 4GHz that all Core i7 Processors are easily capable of.
Everest Ultimate Edition
The grand old Dame that is Lavalys' Everest is heading into the sunset, but for the time being we still have a wealth of results to hand and so let's crack on.
Straight away we can see the balance between CAS8 and 1600MHz of the Dominator kit is right at the sweet spot of the CAS7 1333MHz Mushkin on test today. Both of them giving similar results although the Mushkin edges ahead.
Up at 2000MHz the Dominator GT just edges ahead by virtue of being 8-8-8-24 instead of 8-10-8-27. Of course when we bump the processor up to the logical 200BCLK the Mushkin goes back ahead as we'd expect. Although it does demonstrate how much pure performance is available.
AIDA64 is the replacement for Everest, designed and built by exactly the same team. We'll be gradually replacing Everest with AIDA64 as we gain more benchmarks.
For now the results nicely follow those we saw in Everest with the increased bandwidth available from XMP2 really allowing the Mushkin to stretch its legs and the CPU overclock pushes it still further.
The Memory Bandwidth test of SiSoft Sandra is much more reliant upon the rated MHz of each RAM kit and so is probably the purest synthetic benchmark we run here at OC3D. As you can see the results are almost exactly where you'd expect them to be compared to the speed. The 1333MHz Blackline Ridgeback at the bottom, then the 1600MHz Dominator. The main difference between the 3 2000MHz kits are that the Mushkin needs slacker timings to achieve that speed than the Corsair Dominator GT but regains the loss with the CPU overclock.
PC Mark Vantage
Speaking of linear graphs, PC Mark Vantage gives us almost exactly that. When we're creating the graphs we try and order them in the order we'd expect the results to be and PC Mark Vantage with its suite of in-built utilities gives us nearly perfect results.
With the extra oomph of the 200BCLK test completely dominating the chart, the other tests are surprisingly close in the overall PC Mark score.
CineBench is almost the antithesis of our Sandra Memory Bandwidth in that it is hugely CPU dependant. Although our testing confirms this with the memory speed not making as much of a difference there is definitely something curious going on with the XMP2 profile on the Mushkin. As the CPU speed is similar to the Dominator GT 2000MHz kit the results should be on a par but it lags far behind both the 200BCLK and Dominator GT. Curious.
Curious as to whether this is an issue solely with CineBench we tested POV-Ray and although the drop-off wasn't as dramatic as we see above there is nonetheless some performance issue with rendering using an un-tweaked XMP2.
The 32 Million place calculation actually shows a difference between the memory on test which is surprising considering how quickly its over. Once again we're seeing something strange from XMP2 as it's slower than even the XMP1.
The 1 Billion place test naturally widens that gap even further. The low CAS that we have in the 1333MHz Mushkin at default is obviously the reason for it undercutting the Dominator 1600MHz, but the XMP2 profile isn't so lax in its timings compared to the Dominator GT that it explains this huge difference.
3D Mark Vantage
Finally the ever popular 3D Mark Vantage. Whilst we await the latest version there is still a little life in the old dog. The results follow the pattern we're used to now with the Mushkin XMP1 holding its end up very well, XMP2 being surprisingly under-performing and the the 200BCLK overclock rocking our world.
The Mushkin PC10666 Low Voltage kit is definitely a little schitzophrenic thanks to two polar-opposite XMP profiles.
The primary profile, the one that you're buying the kit for, XMP1 1333MHz @ 1.5v 7-7-7-20 timings is fantastic. Don't let the 1333MHz aspect of it fool you as the low timings mean it beats the very popular Dominator 1600MHz in pretty much every test.
There is a tendancy with PC hardware to always feel that faster must equal better. We've tested low latency vs speed a few times before and normally the pure speed wins out thanks to the enormous bandwidth available, but this Mushkin kit flies in the face of that convention and at only 1.5v.
Curiously though that's not the only convention it flies in the face of. Rather than having a second XMP profile that is around 1600MHz and letting us overclock it, Mushkin have kindly got it all done for us by providing a CAS8 2000MHz profile for us. Initially this seemed to be a great thing as our synthetic benchmarks showed that this could be a kit that provided all things to all men.
Once we'd got out of the synthetic world and into some real-world benchmarking though it quickly crumbled. Sometimes even being out-performed by the XMP1 profile. Rendering tests in particular were woeful.
As I said above, this is schitzophrenic though because once the CPU is overclocked, but with the Mushkin at the same XMP2 settings, it suddenly became a beast, dominating every chart comfortably.
So what do we conclude from all this? Well if you want a nice low voltage kit, you can't do much better than this Mushkin Blackline Ridgeback kit. If you think the opportunity to get a 2000MHz kit on the cheap is too much to pass up, you need to be willing to tweak your CPU and raise your BCLK to get the most from it. You just need to decide what you want it for and judge accordingly.
Thanks to Mushkin for providing the Blackline Ridgeback for review. Discuss in our forums.