Performance
Test Setup:
Since the X99 platform is quite young then we can expect many issues with the UEFI BIOS. I have noticed that some X99 motherboards have problems with DDR4 performance. The MSI X99S SLI Plus motherboard was working great only on BIOS 1.1. Versions 1.0 and 1.2 were performing worse, up to 10GB/s difference in benchmarks like AIDA64. Considering that fact, the used BIOS version was 1.1.
The CPU used in this review is a retail version of Intel i7 5820K, which is probably the most popular 2011-3 processor due to the lower price, good performance and 6 cores. The CPU was overclocked to 4.2GHz and the cache frequency to 3.5GHz to be sure it is not slowing down the memory.
Additionally for the 3D graphics tests we used the Gigabyte GTX970 Gaming G1 graphics card.
There are still not many DDR4 memory kits on the market and manufacturers still have problems with samples so we do not have many other kits for comparison. As a second kit for comparison, I have used the Corsair Vengeance LPX 4x4GB DDR4-2666 CL15-17-17-35 memory and also added the Crucial DDR4-2133 results after overclocking at fully stable settings of 2666 13-13-14-34 1N 1.35V.
I do not want to compare DDR3 to DDR4 results because there are too many factors that are affecting the memory bandwidth, so the results will not be accurate. Simply put, it is not possible to directly compare DDR3 to DDR4 speed and a comparison like that would be misleading.
AIDA64 Cache and Memory Benchmark
As usual I am starting with AIDA64 which is probably the best memory bandwidth benchmark and also shows the speed of various cache transfers.
The much lower clocked Crucial memory at standard settings is mainly slower in the memory read and latency tests. In other tests the difference is not so big and should not affect daily performance.
HyperPi 32M
HyperPi 32M which is an overclockers benchmark, is perfect to test the memory stability and performance. This test is using multithreading for the CPU and memory calculations and really likes fast memory.
As we can see in the results comparison, even though the Crucial DDR4-2133 has a lower clock speed then general performance is not bad. It is losing 10 seconds to next in line Corsair 2666 memory but after overclocking at the same clock of 2666 but tighter timings is able to run over 25 seconds faster.
Futuremark PCMark8
PCMark8 has been tested in two presets using accelerated version of benchmarks – Home and Creative. Both presets test the whole system so we can check how much the memory is affecting daily work.
In both presets we can see some differences but they are so low that we should not see it in daily usage. While both memory kits are set to 2666 frequency, we cannot see any difference so clearly the benchmark is based more on memory clock than faster access time.
For most users, the most important question is how much performance gain we can see in daily work? The answer is almost none and it looks like even DDR4-2133 memory performance is high enough to work with even more complicated applications.
Cinebench
Cinebench has two rendering benchmarks. In this case we will focus on the CPU presets that are using more memory bandwidth.
Similar to PCMark8, also here the difference between memory the kits is not really visible. There is couple of points more on higher clocked memory but it is not really affecting the general performance of the computer.
Both benchmarks, Cinebench 11.5 and 15, are acting almost the same even though there is couple of year’s difference between them.
Futuremark 3DMark – Fire Strike Extreme
Now something for the gamers, 3DMark in the latest and most demanding test, Fire Strike Extreme.
This test is testing computer’s performance in games and is mainly based on graphics card performance but is also using the CPU to calculate physics. Memory performance is highly affecting the physics score and also total score.
As we see in the table below, There is big difference in the physics score but final score is not changing so much as general performance is already high. Memory performance is high enough to keep similar results in the graphics tests, so we are sure that the lower clock will not slow down games even though there is big difference in frequency between standard settings of the Crucial and Corsair memory kits. If we need more performance, then the overclocked Crucial will be even faster.
Now a couple of words about the performance and overclocking results. Some readers may say that Corsair memory after overclocking will be faster what wouldn’t be really true. Crucial after overclocking is faster than Corsair memory at declared XMP settings but the same Corsair memory after overclocking is reaching about the same performance as Crucial DDR4-2133 so in both cases optimal settings after overclocking or timing adjustments are 2666 CL13-13-14 1.35V. Even though both can run at higher clock then effective performance is not raising and using MSI X99S SLI Plus motherboard was even dropping.
Overclocking
To be honest, being used to DDR3 modules based on Micron IC, I was not expecting really good overclocking but I was wrong. The Micron D9RGQ is overclocking really good. I had a chance to see more memory kits based on this IC and we can count on clocks between 2600 and 3200 depends from voltage and timings.
Now a couple of words about voltage. As we know, too high a voltage can damage memory, if not instantly then after some time. In official JEDEC specification we can read that standard voltage for DDR4 is 1.2V +/- 0.06V. However in the Micron specification for our review kit we can find absolute maximum of 1.50V at which this memory should work. It is hard to say if it is safe as it is still a new product but I would not recommend using such a high voltage for daily work.
Considering the standard and maximum values I stopped at 1.35V for daily overclocking and 1.5V for short benchmarks.
Our memory kit was able to run fully stable at 1333MHz ( DDR4-2666 ) 13-12-14-34 and command rate 1N at about 1.35V. It is a great result considering much the higher memory frequency and tighter timings, while the voltage is not much higher.
The maximum clock achieved using 17-17-17-36 main timings and 1.5V memory voltage was 1572MHz (DDR4-3144). This is simply an awesome result.
This clock was not fully stable, but it says us a lot about the overclocking potential of the new Crucial memory. Many users can reach DDR4-3000 clock at lower voltage but it is never guaranteed. We will have to wait for factory tested kits at higher clocks from Crucial.
Below is a part of the window from the CPU-Z validation of Crucial 32GB DDR4-2133 CL15 @ DDR4-3144 CL17.
