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.

Tighter sub-timings and longer cell refresh cycle errors will greatly depend on the memory temperature as well. This is why retail kits generally have looser timings to account for different operational environments. It is recommended to stay below 60°C for 7200 MT/s. As while this memory kit did "pass" the 30 minute mark, the system froze shortly after and hard-locked at 70°C.

Note: 2-DIMM ASUS Z790 APEX was used. Due to close proximity, temperatures are 10°C higher compared to 4-Slot spacing without airflow.

Intel Results

DDR5-7800 Submission Link


DDR5-8000 Submission Link


DDR5-8200 Submission Link

For starters, overclocking with memory that uses SK Hynix A-Die IC was an interesting adventure to say the least. Every A-Die has different limits too. With only five kits on hand, it is hard to give a average limit of what A-Die can achieve for daily use, though they all can do 7600 MT/s at least. That being said, a direct fan on the memory is absolutely required for anything that is remotely memory intensive. There is a direct correlation to system stability and DRAM temperature. Lower is better and 45°C is the target goal. Contrary to what many preach on the forums, maxing out the DDR5 tREFI value is not ideal without proper cooling such as a waterblock at these higher memory frequencies. Longer delays between cell refresh cycle, can lead to stability issues and lost data. Therefore, all overclocking was preformed using a spare OS, that if corrupted would not hinder the review process.

After the first overclocking round with the G.SKILL Trident Z5 RGB 6800 kit previously reviewed, it was time to push the limits some more. Now with a ASUS Z790 APEX motherboard in hand, it was time risk instant death of the memory (which didn't happen). Two things came out of that. One, this memory is completely stable at 8200 MT/s, and two, 8400 MT/s works to some extent.

As discussed already, memory temperatures are a big factor for a successful overclock, which could be resolved by using a waterblock or a direct fan. To achieve 8200 MT/s, the voltage had to be bumped up to 1.55, which is currently up for debate if that this okay for long term daily or not. Most will lend toward no, myself included, but it lasted 6 hours - all the way through the entire overclocking session. That is considered a win, though this does mean 8000 MT/s would be more or less the limit for this particular kit without exceeding 1.5 V. This particular kit did also need 1.5 V for 7800 MT/s, which means it would not have passed G.SKILL retail binning process. Thus, probably why it is binned and sold as a 7200 MT/s kit instead.

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





Overclocking to these speeds is more about saying you can do it, because the benefits start to wane quite quickly beyond 7200 MT/s. For gaming, the RTX 3080 Ti was the limiting factor here. Using the ASUS Z790 APEX, DDR5-7800 overclock setup was simple and just required the correct DRAM voltage and primary / sub-timings timings, which the SPD data used was copied directly from another DDR5-8000 kit. Not really a challenge per say. 8000 MT/s and above is going to be a time sink. No overclock is ever guaranteed, which makes memory overclocking a new adventure every single time.