Maximus 12 Formula with 11900k unable to POST RAM at 3866 gear 1 or 4400/4533 gear 2

[morph.]

Some assistance would be greatly appreciated... So I have an M12F bios 2201 the story is the same for all the previous bios's released of late paired with a 11900k SP85.

I've been unable to Post at 3866 Gear 1, 4400 gear 2 or 4533 gear 2.

My RAM is G.Skill Trident Z Royals F4-3600C14-16GTRS (dual rank bdie)

With the same kit but back to my old sp79 10850k CPU I can run 4400c17 stable without any issues.

I've tried some aggressive voltages by increasing dram voltage to 1.55v, SA voltage to 1.45v, mem io to 1.35v but still no POST even with timings on auto.

I also seem to hit a stability problem when my cache/un-core ratio goes above 4.5ghz which seems fairly low, I believe my vcore is sufficient.

Here is what my 3733 g1 is looking like at the moment couple more values I can possibly decrease with TRTP/Twrrd_dg

[thomas_yiu@ROG]

Hello Morph,


I am sorry to hear you having issue with it. The memory is not in the Asus Memory QVL list. but in the Gskill memory compatible list.
Please troubleshoot by turning off the system (unplug power) and than reseating the memory, cleaning the memory, make sure the timing is
16-19-19-39 as spec, I suggest also to run Memtest86 to see if memory is working properly or not. TY

[Braegnok]

Hello Morph,

Most likely your memory is not the issue, as it's stable and working fine at 3733.

Which microcode does BIOS 2201 have, are you running 0x34, 0x3c, 0x39,..

I'm thinking perhaps BIOS/Microcode update could be the answer for stability when overclocking memory, running RL chip in your M12F.

[morph.]

Hello Morph,


I am sorry to hear you having issue with it. The memory is not in the Asus Memory QVL list. but in the Gskill memory compatible list.
Please troubleshoot by turning off the system (unplug power) and than reseating the memory, cleaning the memory, make sure the timing is
16-19-19-39 as spec, I suggest also to run Memtest86 to see if memory is working properly or not. TY

Thanks for your response however I believe my screenshot has shown this working perfectly fine with a mem test as well for 3733, I can also show you it working perfectly fine with 4400 on the same motherboard/ram if you want just a different CPU 10850k previously.

While I understand the CPU is different and IMC's can be different I believe this is more down to the bios support for the 11900k and mine is of SP85.

Hoping something Shamino/Safedisk can turn their attention to and support.

Here is another screenshot of even further tightened secondary and tertiaries running stable at 3733 gear 1 with a successful pass of memtest.

[morph.]

Hello Morph,

Most likely your memory is not the issue, as it's stable and working fine at 3733.

Which microcode does BIOS 2201 have, are you running 0x34, 0x3c, 0x39,..

I'm thinking perhaps BIOS/Microcode update could be the answer for stability when overclocking memory, running RL chip in your M12F.

I've tried with all the new various iterations of bios's from 2101, 2102, 2103 and now v2201 (as provided by safedisk).

My microcode is currently 3Ch

I am aware that running 3866 at gear 1 mode is very hard to do so successfully as of now but seems to be just attainable with either SR or apex boards for dual rank.

However, I'm fairly stumped why I can't get Gear 2 4400/4533 to post/train even with loose/auto timings and bolstered voltages. I know with the 10850k same everything else 4400 ran solid.

[Silent Scone@ROG]

What Qcode?
SA Voltage at 1.45 is far too high. Is there any reason you’re pumping this rail so far?

These are signal rails and are sensitive to too much voltage, it will hurt stability. I would not recommend using that much voltage long term. How much is needed will depend on the CPU. You’ll need to test this methodically.

Look at relaxing tertiary timings. I would also recommend leaving both the CPU and uncore at stock for the time being. Pushing the uncore will have an adverse effect on maximum obtainable frequency to a point

[morph.]

What Qcode?
SA Voltage at 1.45 is far too high. Is there any reason you’re pumping this rail so far?

These are signal rails and are sensitive to too much voltage, it will hurt stability. I would not recommend using that much voltage long term. How much is needed will depend on the CPU. You’ll need to test this methodically.

Look at relaxing tertiary timings. I would also recommend leaving both the CPU and uncore at stock for the time being. Pushing the uncore will have an adverse effect on maximum obtainable frequency to a point

My VCCSA is normally around 1.30-1.35v for me but in my angst to try to get tighter timings and or higher frequencies to post I cranked it up. I'll start to optimise (dropping voltages down) soon thanks for the reminder is 1.35v too high? (its no longer 1.45v sa).

Must say I haven't thought about dropping cache/uncore as it doesn't feel too high, if anything I was trying to make it higher but ran into stability issues. Previously on my 10850k/8700k i ran cache ratio at 48x. But next time I'll try take a look at dropping that back to default but this would drop my memory latency/bandwidth a little thanks!

Furthermore, when I did try to post at 4400 & 4533 in gear 2 I did leave all my timings to auto.

[Silent Scone@ROG]

My VCCSA 1.30-1.35v for me but in my angst to try to get tighter timings and or a higher frequency I cranked it up. I'll start to optimise (dropping voltages down) soon thanks for the reminder is 1.35v too high? (its no longer 1.45v sa).

Must say I haven't thought about dropping cache/uncore as it doesn't feel too high, if anything I was trying to make it higher but ran into stability issues. Previously on my 10850k/8700k i ran cache ratio at 48x. But next time I'll try take a look at dropping that back to default but this would drop my memory latency/bandwidth a little thanks!

Furthermore, when I did try to post at 4400 & 4533 in gear 2 I did leave all my timings to auto.

For those frequencies you may want to look at dropping them further. Auto rules can be quite tight depending on the scenario.


1.35 is on the high side but should be fine, I would not recommend going far outside this and would consider it to be poor tuning for a daily rig.

Here is an example of excessive VCCSA causing instability. The only change is the increased VCCSA. Memory training is a harder pass than operating system style tests because the requirement falls down to a pass/fail. When exploring highest obtainable frequencies these changes are more important as the margin for error on an electrical level is much smaller.

Admittedly I haven't used this platform much, but these rules are universal depending on the architecture. Moreover, VCCIO shouldn't be far too out of spec from VCCSA, this is also comes down to an alignment issue when overclocking.

I would look to dialling back on these rails and try and dial back on terts and seconds where possible. Some of those rules can still be quite tight for a given ratio even in auto. This is by no means guaranteed to work, but may keep you on the right path. The strength of the memory controller still needs to be considered.

You may wish to aim for your maximum obtainable frequency at CR1, too.

[morph.]

For those frequencies you may want to look at dropping them further. Auto rules can be quite tight depending on the scenario.


1.35 is on the high side but should be fine, I would not recommend going far outside this and would consider it to be poor tuning for a daily rig.

Here is an example of excessive VCCSA causing instability. The only change is the increased VCCSA. Memory training is a harder pass than operating system style tests because the requirement falls down to a pass/fail. When exploring highest obtainable frequencies these changes are more important as the margin for error on an electrical level is much smaller.

Admittedly I haven't used this platform much, but these rules are universal depending on the architecture. Moreover, VCCIO shouldn't be far too out of spec from VCCSA, this is also comes down to an alignment issue when overclocking.

I would look to dialling back on these rails and try and dial back on terts and seconds where possible. Some of those rules can still be quite tight for a given ratio even in auto. This is by no means guaranteed to work, but may keep you on the right path. The strength of the memory controller still needs to be considered.

You may wish to aim for your maximum obtainable frequency at CR1, too.

Thanks Scone are you aware if there are any voltages that increase the likelihood of CR1 being better obrtainable? I

know Memory Trace Centering was a setting that should help.

[Silent Scone@ROG]

Thanks Scone are you aware if there are any voltages that increase the likelihood of CR1 being better obrtainable? I

know Memory Trace Centering was a setting that should help.

Not on this board at those frequencies, no. I've limited experience on that particular platform but Gear 2 will likely be needed for anything beyond 3866 at 1T on that board