Had a Overclocking Dream Lately?

Overclocking Tips and Tricks

Getting Warm?

For thermal testing, Karhu stress test software was used for 30 minutes, after which both DIMM temperatures from the SPD hub sensors are averaged together. Testing was performed with and without a fan at the XMP Profile of 1.40 V. DDR5 SK Hynix memory ICs are susceptible to temperature related system errors. Since more voltage is necessary for higher frequency memory, caution is advised when overclocking without direct airflow across the memory, or at least decent case airflow. Otherwise, unexpected stability issues may occur as the memory warms up.

This became all too apparent when using the ASUS Z790 Maximus APEX during the last review, which mind you, only has 2-DIMM slots! Learning from mistakes previously, thermal testing was performed on the MSI Z790 ACE because of the 4-DIMM slots allowing an air gap in between each memory DIMM. For reference, Due to the close DIMM proximity using the ASUS Z790 APEX, the T-Force Delta RGB DDR5-7200 errored out around 70°C, just like the G.SKILL Trident Z5 RGB kit before it.

Intel Results

DDR5-7800 Submission Link


DDR5-8000 Submission Link


DDR5-8200 Submission Link

Behind the scenes, we are lucky enough to have acquired a number of SK Hynix A-Die based memory kits. So far they all have been able to reach a 7600 MT/s overclock with various primary timings and voltages. To achieve these results, a direct fan on the memory was needed to pass memory stress tests. It is essential for performance grade memory to have proper cooling. There is certainly a direct correlation to system stability and DRAM temperature that can be made. Lower is always better and the target goal here is 45°C.

Related to system stability and DRAM temperature, those who are tweaking the memory sub-timings may run into problems when maxing out the tREFi value. This is not a good idea without proper cooling such as a waterblock at these higher memory frequencies. Longer periods between cell refresh cycles can lead to data corruption, if not those pesky system errors that are hard to track down. All overclocking here was performed using a spare OS, that if corrupted would not hinder the review process.

With good success previously in the G.SKILL Trident Z5 DDR5-7200 review, a few things came out of that overclocking session. Firstly, 8400 MT/s works to some extent, and 8000 MT/s is quite easy with this ASUS Z790 APEX motherboard. The G.SKILL OC attempt was done using two different 8000 MT/s profiles. One was copied from a G.SKILL 8000 MT/s kit and the other was a Team Group profile. Once again, it was easy to reach 8000 MT/s. In fact, this time it was a direct copy of the Team Group T-Force Delta RGB DDR5-8000 SPD data with a few tweaks to tREFi and tRRD_sg values. This SK Hynix A-Die based kit from Team Group was quite good, only needing 1.45 V for the initial 8000 MT/s goal. 8200 MT/s used 1.50 V. A quick run at 8400 MT/s with loose timings didn't help. Most things in Windows worked, but wasn't passing y-cruncher 2.5B. At the end of the day, DDR5-8000 was a happy medium since it did not require excessive DRAM voltage and could be used for daily operations.

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





Overclocking at this higher speeds is more about showing you can do it. For gaming, this RTX 3080 Ti is the true limiting factor here. That being said, using the ASUS Z790 APEX, DDR5-8000 was a breeze this time around. Settings were copied directly from the Team Group T-Force Delta 8000 kit, and things happened to work out well. As always it's important to point out that no overclock is ever guaranteed, which makes memory overclocking a new adventure every single time. This also means that the same settings may require more voltage for another kit, or may not be possible depending on the motherboard, CPU and how well the memory performs beyond the rated XMP profile.