Tweaking for Speed

With testing out of the way, I endeavored to see if these sticks have any headroom. For Intel, I kept the same procedure I have been using: I used the XMP profile and increased the frequency until the system lost stability. After finding that limit, I manually tweaked for the maximum frequency and lowest possible timings. Voltage modification from stock is allowed. After all, this is overclocking!

The 11th Gen Intel Core processor paved the way for things to come. The introduction of the memory controller Gear Ratio allowed the system memory to run in synchronous 1:1 mode with the CPU memory controller, or 2:1 ratio. With the release of Intel's 12th Gen Alder Lake based processors came DDR5 support and additional 4:1 ratio.

1:1 ratio generally falls between between 3600 and 4000 MT/s for Alder Lake CPUs. This is completely dependent on the CPU memory controller and supporting voltages. In rare instances, higher-end motherboards can increase this slightly and offer better overall compatibility. My Intel Core i9-12900K maxes out at DDR4 4133 MT/s, which is rare if going by the sheer number of forum posts about many struggling to reach 3800 MT/s. It is safe to say that anything greater than 3600 MT/s will require a bit of hands-on tuning.

With this information, Intel 12th Gen Alder Lake paired with DDR5 will gain the most from the highest-possible frequency without giving up the benefit of the increased bandwidth. Because DDR5 has a higher operational frequency and dual 32-bit data bus, synchronously operating it in 1:1 is unlikely. That only leaves 2:1 and above as a viable option for any DDR5-based setup. The motherboard should automatically switch to the 2:1 ratio. If all else fails, you can manually set this in the BIOS.

Those looking to overclock will generally find a hard barrier around 6600 MT/s on the Intel Z690 as many 4-slot motherboards do not support higher speeds. Only a handful of motherboards are designed to support 6666 MT/s and up. These are the ASUS Z690 Apex, Gigabyte Z690 Tachyon, MSI Z690 Unify-X, ASRock Z690 AQUA OC, and EVGA Z690 Dark.

Caution is advised with DRAM voltage over the rated XMP profile. Direct airflow or a waterblock may be necessary for long-term stability. This extends to the CPU as well. Raising the integrated memory controller voltage (vDD2), System Agent (SA), and VDDQ_TX may cause irreparable damage. Please proceed with care and do research before attempting this. Do not copy and paste values without understanding the impact first, especially if taken from screenshots posted on Discord or Reddit.


After numerous requests to thermally test DDR5, I am still not ready to fully commit to a finalized testing method. For the time being Karhu stress test software is used for 30 minutes, after which both DIMM temperatures from the SPD hub sensor combined are averaged. With the limited time I had available for this review, testing was preformed with and without a fan at the XMP of 1.3 V and a overclocking voltage of 1.5. After failing the stress test at 6800 MT/s 1.5 V, the v-color memory refused to boot again until it cooled off. I caution against overclocking without a fan directly placed on the memory.

Intel Results

I may have gone overboard with the overclocking. Now that I have enough Hynix M-Die kits, accidentally killing one with over-voltage isn't detrimental anymore. v-color's memory kit unintentionally became my guinea pig!

In any case, knowing I would eventually have to switch motherboards if wanted to reach above 6600 MT/s, I swapped the ASUS Z690 Hero used for the benchmarking portion out for the overclocking-orientated Gigabyte Z690 Tachyon right at the start of this overclocking session. First was to go up using the XMP profile without changing the values, which ended in failure. Because the XMP voltage is 1.3, I was unable to stabilize anything else further. A small reminder for those new to memory overclocking: Just because it boots into Windows does not mean it is stable.

Next up was to try a number of different parameters based on past overclocking. After extensive testing and dozens of hours later, we have some solid results. Since the stock XMP DRAM voltage of 1.3 was out of the question for anything useful in terms of further overclocking, the first major attempt was to raise it to 1.4 V and see what can be done. 1.4 V allowed for up to 6600 MT/s. Tightening the timings further was possible at 6800 MT/s, but required 1.55 V. This is above the "absolute limit" of 1.5 V listed in the Hynix data sheet. I of course did try for a higher voltage, but the PMIC was inconsistent on the supplied voltage and would often not boot at 1.5 V, let alone anything higher.

Officially, 1.475 V is the limit for this sample without giving me problems—at 1.475 V, the voltage didn't dip occasionally and cause lock-ups. With this voltage limitation out of my hands, setting 7000 MT/s booted into Windows, but was unstable in testing. This seemed to be due to a combination of a lack of DRAM and IMC voltage. As I wasn't willing to increase it further, the compromise was 6933 MT/s CL34.

For newcomers, getting DDR5 overclocking stable is a bit complicated. To reach 6933 MT/s, major adjustments to the System Agent (SA), vDD2 (memory controller), and TX VDDQ were needed. I will not share my final voltage settings in fear of someone just copy and pasting the values. However, those who enjoy the process of tinkering will be rewarded.

Note: All memory overclocks passed Karhu stress test 2000% or more.