After thinking i got [this](https://www.reddit.com/r/overclocking/comments/18bpdtf/2x24_gb_hynix_mdie_oc_results/), it turned out unstable:
>!OCCT 4h nothing, Ycruncher 4h stable, Memtest86 3h stable, TestMem5 extreme + absolute 3 cycles stable (each about 2h), last run:!< TestMem5 with 1usmus over night: 97 Errors
I looked at HwInfo64, saw my sticks were consistently at 60-65°C so fix the problem with [this heatsink mod](https://www.reddit.com/r/overclocking/comments/18juo01/ddr5_heatsink_mod/). Worked wonders, 20°C temp drop reported, probably even more for the chips themself.
Redid all the tests: **1 Error** after a total of 16h stress testing.
​
I know i could probably fix this with fiddling around with the voltages even more, but i don’t want a system that is right on the edge of stability.
At least i know where the edge is >!\- 7200: no matter what i do, even with extremely loose timings and different voltages from high to low – errors after some minutes!<
>!7000 probably possible to get stable with finding the exact sweet spot of those voltages.!<
​
So i settled on this:
https://preview.redd.it/927bf45fk07c1.png?width=527&format=png&auto=webp&s=b16daba24858a9b452aa625cee1973c26801b51e
Tight but still conservative, a lot of timings can go lower (nearly every single on of those, except tCL, tRCD, tRRD\_L and tFAW) – tho i could probably dump a lot of dram vdd into it to get cl30, not worth it imo
CPU VCCSA: 1,16 v
VDD2 (IMC): 1,38 v
CPU VDDQ: 1,38 v
DRAM VDD: 1,42 v
DRAM VDDQ: 1,38 v
(voltages got reduced as far as possible to see a quick limit, then set with +0,03v over the limit – then stress tested for longer)
​
I know – nothing special, but maybe it helps someone here that gets frustrated trying to reach something above 7000mHz (after reading all those 8000+ results here).
It’s just the reality of current gen 4-dimm motherboards + cpus and [not worth it anyways](https://www.youtube.com/watch?v=RTmbYak_8gE&t=289s)
6 Comments
Usually my process is to find the highest overclock that I can’t find any errors in, then I just dial it back 1-2%. Yeah you might lose a bit of e peen going -100Mhz on clock but you’ll have an infinitely more stable system for realistically a 0-1% performance loss.
13700K user here, my memory controller couldn’t handle 6800C34 without 1.45V……didn’t matter the secondaries/tertiaries, I hope Intel does a better job on 15th gen (14th gen’s memory control has zero improvement)
I’m on a 6layers PCB 4dimmer board (the aorus elite ax) and I find your thread perfectly in line with my experience. I’m now on 7000 stable, tried for so long to stabilize VST at 7200 but the most I could do wasn’t cold-boot stable.
I personally ignore all the people trying to sell high memory speeds on these boards
I finally got my rig stable and something I’ve noticed about ddr5 is that it’s highly highly influenced by vcore, more so than earlier generations seemed to be. More specifically, for my build I have to make sure that my vcore is always less than my IMC/IVX_TR or I’ll wind up with an access violations. Everything will pass and I can even run stress tests, but as soon as I open a website or run any other program it would crash.
– Verify that the 12/4/16 tRRD and tFAW profile improves your PYPrime time. When I tried 12/4/16, PYPrime was almost consistently 0.4s higher than when using 12/8/32.
– tCL at 30 is probably possible at or slightly above 1.45v VDD.
– tRDRD_dg and tWRWR_dg can easily go to 7
6800 was the highest stable speed I could reach too. Glad to see others post this.
i7-14700k
Gigabyte Z790 AORUS ELITE AX ATX