Corsair CMK8GX4M2A2133C13 Review

The time has now come for us to perform our usual overclock-testing of the test sample. Even if you are not planning to overclock your memory, knowing its spec headroom and overclocking capabilities can serve as a good assessment of manufacturer's quality control.

We conduct our experiments using a test platform based on Intel's latest generation of CPUs, which is commonly known as Skylake.

 CPU  Intel Core i7-6700K (ES)
 Motherboard  ASUS ROG Maximus VIII Hero (BIOS 0040)
 Memory  Corsair Vengeance LPX CMK8GX4M2A2133C13
 Graphics card  Zotac AMP GTX280
 Storage  Samsung SSD 850 Pro 256GB
 Power supply  Seasonic Platinum 1200W
 Operating System  Windows 7 SP1, 64-bit

There are many ways to test stability of the memory system and our procedure is as follows. We first lay a baseline with Memtest86+, which works independent of OS and is highly reliable at detecting errors. Then, we verify the stability of each setting combination in Windows 7 by using a series of memory-dependent benchmarks such as SuperPi 32M, Intel Burn Test and XTU. Sometimes, we also double check with HCI Memtest or Prime for daily stable settings.

Our test platform has a very wide range of memory clockspeeds to choose from, but to keep things simple we will only concentrate on the common memory dividers from DDR4-2133 and up, leaving the lower and some of the intermediate options out as potentially irrelevant. Once we select the memory frequency, we seek for optimal combinations of primary timings, for which we minimise the stable voltage in 0.01V increments.

The Spectek ICs we found on the Corsair were new to us, so we had to dig in multiple directions to see what's what. Firstly, we discovered that in the memory frequency range between DDR4-2133 and -2666, the voltage necessary to decrease the CAS latency is linear, except for the very low CAS values (10 and 11) at DDR4-2400 and -2666.

Another surprising characteristic of the Micron chips at hand was the tRCD and tRP behavior. Unlike with Hynix- or Samsung-based kits where the minimal stable value depends primarily on the memory frequency, with Micron it can be lowered along with the CAS latency. This way, it is possible to run tRCD and tRP at one or two values below CAS at low clockspeeds, equal to CAS at medium clockspeeds and one value above CAS at high clockspeeds. As for the tRAS, the lowest available value of 28 was stable throughout the entire testing range.

We tried to keep voltage at safe regions (max of around 1,46V), though we know that DDR4 is capable of running voltages of 1,5V and above.

Below you can see the list of stable configurations that we were able to achieve with our sample.

1066MHz (DDR4-2133)
1200MHz (DDR4-2400)
1333MHz (DDR4-2666)
 CL9
9-8-8-28 @ 1.35V
Not relevant
Not possible
 CL10
10-9-9-28 @ 1.27V
10-9-9-28 @ 1.37V
Not relevant
 CL11
11-10-10-28 @ 1.20V
11-10-10-28 @ 1.29V
11-10-10-28 @ 1.40V
 CL12
12-11-11-28 @ 1.15V
12-10-10-28 @ 1.23V
12-11-11-28 @ 1.31V
 CL13
13-12-12-28 @ 1.10V
13-11-11-28 @ 1.19V
13-12-12-28 @ 1.25V
 CL14
Not relevant
14-12-12-28 @ 1.15V
14-13-13-28 @ 1.20V

1400MHz (DDR4-2800)
1500MHz (DDR4-3000)
1600MHz (DDR4-3200)
 CL10
Not possible
Not possible
Not possible
 CL11
Not possible
Not possible
Not possible
 CL12
12-12-12-28 @ 1.37V
Not relevant
Not possible
 CL13
13-12-12-28 @ 1.31V
13-14-14-28 @ 1.35V
Not possible
 CL14
14-14-14-28 @ 1.26V
14-15-15-28 @ 1.30V
Not relevant
 CL15
15-15-15-28 @1.21V
15-16-16-28 @ 1.26V
15-15-15-28 @ 1.46V

Passing the specs posed no difficulty to our test sample as even 1.10V was enough to clear DDR4-2133 with main timings lowered to 13-12-12.

Pushing the voltage to 1.35V, we were able to drop the CAS latency to 9 and arrived at timing values which are impressive even by DDR3 standards. Our further tests have shown more progress as the memory easily cleared DDR4-2400, -2666, -2800 and -3000, the latter of which seemed like a sweetspot in terms of frequency-voltage ratio.

Driven by the spirit of overclocking, we also achieved stability at DDR4-3200 15-15-15 and DDR4-3333 16-18-18, pushing 1.46V both times around. We are not 100% sure whether these settings are going to be reproducible for daily use, however, they did pass all of our stability tests such as Memtest86 and Intel Burn Test.

We round off the testing page by presenting some benchmark scores to show the performance gain that a normal user can achieve by overclocking the memory.

CMK8GX4M2A2133C13_2400C10XTU CMK8GX4M2A2133C13_3000C14XTU CMK8GX4M2A2133C13_3200C15XTU

The performance gain in synthetic benchmarks like XTU is more than noticeable, however, in tests that only concentrate on raw CPU power, the gain is barely one percent.

Leave a comment