Recently we received a new memory kit from Crucial that is part of their Ballistix line. We decided to test this kit on the AMD platform to see how they would fare. Considering that AMD can be picky with certain modules, we found this particular set quite interesting. Let’s take a closer look shall we?
Based on the target consumer and price range of these modules, we decided against testing with SuperPi 32m and instead opted to test ultimate stability using Prime 95 (Blend) for a minimum of 3 hours per setting. We also opted to use a Deneb quad core CPU versus Thuban so aso to cover the bases for all users. Those using Thuban may see larger gains due to an improved IMC, and we felt that may impact/skew results for the majority.
Let’s take a look at the test rig briefly and go over a couple of things.
As we can see here, commonly found with many of today’s performance memory modules incorporating tall heat spreaders, DIMM interference with larger heatsinks can be a problem. Everyone has their own opinions on this, and mine is that the heatsink manufacturers are intruding into a “no fly zone”.
Personally, I purchased a heatsink that took this issue into consideration with the Prolimatech Megahalems, and clearance is a non-issue with it. So you may want to keep this in mind when choosing a heatsink.
Here we can see that the system is all set and ready to go on the test bed.
My particular methodology for testing memory geared towards the average user on the AMD platform, is to use Prime 95 in Blend test mode. This particular mode stresses both the CPU and Memory combined and will usually fail with even the slightest hint of memory instability. Especially around the 2 1/2 hour point in the test, which is why we opted for 3 hours minimum.
Prior to Prime 95 stability testing, Memtest86+ was used to set some of the sub-timings for optimal stability. tRRD-5, tWR-10, tWTR -5 and tRTP -5 were found to be optimal for this particular set. We have enclosed CPU Tweaker 1.3 in all of the screenshots for anyone that would like to match our timings for testing.
To start, we set the baseline testing at CAS 8, 1600MHz with 1.65v and the CPU running at stock speeds.
Next up is baseline testing at 1600MHz, CAS 8 and 1.65V as before, but this time we overclocked the CPU and Northbridge. The purpose of this is to put more stress on the memory and the system in general, as many times memory is known to pass stability at artificially higher speeds until you introduce a total system overclock.
Now for 1700MHz, CAS 8 and 1.65v. At this point the memory is scaling without the need for a voltage bump.
1720MHz, CAS8 and 1.65v is still scaling without the need for more voltage.
1760MHz, CAS 8 and 1.65v was the maximum that we could achieve without scaling voltage higher with the memory. We’ll need to bump the voltage slightly to continue increasing the clock speed.
Now we’re at 1800MHz, CAS 8 with 1.675v this time. At this point we are starting to see scaling without increasing voltage beginning to taper off.
At 1840MHz, CAS 8 with 1.70v we’ve reached the point where adding voltage is becoming fruitless and will just create more heat.
That concludes our CAS 8 testing. We settled at 1846MHz with 1.70v, which was more than likely the limit of the IMC, rather than the RAM itself. So at this point we decided to deviate from spec. AMD is well known for performing best with lower latencies in the CAS 6 range, so I wanted to feel the memory out and see how it liked the tighter timings. In the past, I have found that just because a particular set of ram has a higher speed/timing spec, this does not mean that it has the ability to be tightened up at lower speeds. This particular set of ram had no problem doing so however.
Here’s 1333MHz at CAS 6 with 1.70v.
Impressively, 1600MHz at CAS 7 can be achieved with only a slight voltage increase over the stock rating. This would likely be what I would choose for 24/7 usage based on the balance performance, heat and voltage.
While attempting to max out our memory, we managed to squeeze out 1880MHz at CAS 9 with just 1.725v. Although this is interesting in the fact that the Crucial Ballistix modules were able to exceed what many sets of ram can not do on AMD quad cores with prime stability, the higher latency required to achieve such speeds does not offer great performance. As mentioned, we recommend sticking to tighter timings on the AMD platform whenever possible to maintain optimal performance.
Let’s take a look at an interesting feature that these particular modules offer. Here we can see that we can actually monitor the temperatures of the modules, thanks to Crucial’s new addition of thermal sensors within the heatspreaders. A practical use for this feature could be to determine if there are dead spots in case airflow or possibly discover a faulty DIMM that’s overheating.
Here we can see the SPD data of the sticks, including the speed, default timings and voltage. This can also be viewed using CPU-Z’s SPD tab, but it’s always nice to have this information available.
I should mention that to install this utility, you will need to have the .NET Framework v4.0 installed.
Unfortunately, at this time this software is not compatible with these modules on the AMD platform. Crucial assures us that this is a known issue and they are working on a release that will resolve this issue as we speak.
This particular Crucial Ballistix kit works extremely well on AMD-based platforms and they are quite versatile in regards to the speeds and timings that they can run. We found that with little effort we were able to get these sticks clocking well without having to invest hours of time in to tweaking. I am however disappointed by the 2T command rate that was needed for these sticks as the user will lose some performance from this timing. The lack of the software support for the AMD platform regarding the temperature utility was also unfortunate, but again, Crucial did mention that they would be rectifying this issue. I might also mention that no where on the packaging does Crucial mention anything about the software supporting there sticks, which I find odd considering it’s the only way to get the temperature readings from their new thermal sensors.
Minor issues aside, for users looking for a set of memory that just plain works on AMD without too much fiddling around, I think Crucial has really a winner here! Add to that the fact that they are very attractively priced and can be had for just over $110USD at various e-tailers and you can consider us sold!
The Crucial Ballistix DDR3 1600MHz kit receives the TechREACTION.net Silver Silicon Award!