Need help for RAM overclocking (questions about voltage)

[Jeepy31]

My system : 

  Asus Prime Z790-P

   Intel Core i5-14600KF

   Ram kit Corsair DDR5 2x16     CMK32GX5M2E6000C36  XMP 6000 Mhz  36-44-44-96, 1.4V

When enabling XMP without other change, my system is not stable.

After many tests, my system is perfectly stable with the following settings :

- 6000 Mhz  36-44-44-96, 1.2V  (corresponds to XMP but with 1.2V instead of 1.4V)

- 6000 Mhz  36-44-44-96, 1.3V (corresponds to XMP but with 1.3V instead of 1.4V)

(not stable with same settings at 1.1V).

1) Do you have some explanations why I cannot increase the voltage to 1.4V ?  Does it mean that my motherboard is defective ?

2) Is it better to put voltage to 1.2V (closer to reference 1.1V) or 1.3V (closer to XMP profile) ? For what reasons ?

It is probably not significant, but in all my tests, I seem to get  a slightly better performances (bitrate + latency) with 1.2V rather than with 1.3V.

3) Does it make sense ? Is it possible or just related to 'random' aspect of benchmarks ?

Edit : this morning, I get opposite results (1.3V better than 1.2V). Cold PC ? Not significant results ?

Thanks in advance

[Silent_Scone]

Hi @Jeepy31 

It seems like the stability is conditional under all the configurations you've tried. The problems you're having are likely due to the method you're using to diagnose them.

1. CMK32GX5M2E6000C36 does not appear to be on the motherboard QVL
2. Use Karhu RAM (2k to 6k% coverage) Test or HCI Memtest to test stability (500-1000% coverage)

[Jeepy31]

Thanks for your reply. However, your answer doesn't really seem to correspond to my 3 questions.

My system is stable under the conditions I indicated. I just have some precise questions related to voltage.

Please could you re-read my message  and answer my questions (or at least some of them) independently.

Thanks

[Silent_Scone]

See here...Edit : this morning, I get opposite results (1.3V better than 1.2V). Cold PC ? Not significant results ?

You need to test stability correctly, so my answer directly corresponds with this comment. You're likely not seeing opposite results, but the system stability was conditional in the first place.

For anyone to assist you vicariously, you need to use the proper tools to stress the system, otherwise, you have the issue of waiting periods of time using the system before instability occurs again.

Secondly, if the kit is validated and binned at a specific voltage by the memory vendor, then this is the voltage the kit should run at. Lowering the voltage (VDD/VDDQ) can induce instability as the memory vendor has not accounted for the user reducing the voltage for the designated frequency and timings. If performance appears marginally improved at a lower voltage this might be temperature induced or the impact of signalling related artifacts such as crosstalk and noise, due to being on an entry level board - however when measuring DRAM performance background activity on the system also plays a part.

[Jeepy31]

I understand, I shouldn't have added the "Edit" since it is confusing.

My system is stable under both 1.2V and 1.3V. But in my first tests, bitrate seemed a little better with 1.2V and during my last test, bitrate seemed a little better with 1.3V.  I just think it is not significant and I got the same results with 1.2V and 1.3V.

I would have preferred using 1.4V, as validated by the kit provider, but it is not stable at 1.4V. With 1.2V or 1.3V, I 've had no issue for a week, and memtest is ok (I will check with the tools you suggest too).

So, given that my system is stable at 1.2V and 1.3V, not stable at 1.4V and that the RAM kit is validated at 1.4 :
- Do you have some explanations why I cannot increase the voltage to 1.4V ?  Does it mean that my motherboard is defective ?

- Is it better to put voltage to 1.2V (closer to reference 1.1V) or 1.3V (closer to XMP profile) ? For what reasons ?

Thanks

[Silent_Scone]

Again, you need to run the aforementioned stress tools to see what's stable and what isn't. Depending on the voltage guardband the memory vendor has instilled whilst binning, there is no issue running with a lower VDD / VDDQ assuming the system is in fact stable, which loops back to my opening statement. VDDQ/VDD if signal margins are quite tight can sometimes be CPU dependant, so use what's stable.

[Jeepy31]

Thanks.

I will test with the aforementionned stress tools, with 1.3V since it is closer to the memory vendor reference voltage.

However, I would really appreciate an answer to my precise questions.
First question in particular is in fact : do I have to return my motherboard ?
Second question  is : if both 1.2 and 1.3 are suitable, which one is best

[Silent_Scone]

Unless the board is unstable at Optimised Defaults then you have no reason to return anything. XMP is overclocking, and as such holds no assurances due to variances between parts. Some more information can be found here

[Jeepy31]

Thanks.

I suppose that "Optimised default" correspond to 4800 Mz.

To be noticed  that CMK32GX5M2E6000C36  is not in  motherboard QVL but CMT32GX5M2E6000C36 is, with the exact same characteristics is. So I am surprised it doesn't work anyway.

Does Asus certify new ram kits with time (through BIOS updates) ?

I bought an expansion kit composed of motherboard, processor and ram kit. It would have been a good idea that my seller included a "QVL" ram kit in this expansion kit.

[Silent_Scone]

If more memory kits are validated they are added to the QVL list.

XMP I is the profile that is validated by both the memory and motherboard vendors.

CPU IMC variance plays a large part in the equation as well as the motherboard DRAM topology and PCB layers to name just a few. Again, overclocking holds no assurances and manual tuning can often be required.