case closed... leave it all stock. haha

 

I am going back to prior BIOS to see if it fixes it.

 

It appears it did. And I believe this is actually a beta BIOS that it worked on now removed from MSI website.

The new tested and supposed to be good BIOS borked DDR5 XMP stability.

Where as the beta BIOS E7D28IMS.A81 actually has XMP DDR5 stable.

 

I had similar issues here is what i found out ONLY Gigabyte is aware of the issue that in some instances XMP may NOT work stable they said REDUCING the frequency by 400mhz should result in a STABLE XMP.

Example if you bought a say 7200mhz DDR5 kit and it doesn't perform stable with XMP go back to BIOS enable "optimized default" then enable XMP then REDUCE frequency by 400mhz (7200mhz - 400mhz = 6800mhz) this can be done with ANY DDR5 frequency speed ie 7600mhz - 400mhz = 7200mhz.

So try this out with your DDR5 kit you have 6600mhz so minus 400mhz = 6200mhz then do a Prime95 and use option "Large FFT's stresses memory controller and RAM" do this for 1 hour.
After that when it is stable you also may try Memtest86 full pass. You are left with XMP cas latency which is to loose for 6200mhz you can lower the cas to make it fast.

I went through similar issues as you had and made a thread about it and it has been fixed one guy in particular to which i am very thankfull for guided me all the way here is the link of my thread hopefully it will be of use to you as it did for me:

Help with 6800mhz XMP on Z790 mb ("SOLVED")

EDIT: I wanna thank all the people who tried to help me and a special thanks goes to chimsunsun9x for guiding me all the way. It is solved albeit at a lower frequency @ 6400mhz for some reason sometimes the actual speed on DDR5 is -400mhz so example pick a 7600mhz if unstable try to lower freq...

www.overclock.net

There may be nothing wrong with your RAM nor ICM its something else. Do read my thread and follow suggestions.

 

I had similar issues here is what i found out ONLY Gigabyte is aware of the issue that in some instances XMP may NOT work stable they said REDUCING the frequency by 400mhz should result in a STABLE XMP.

Example if you bought a say 7200mhz DDR5 kit and it doesn't perform stable with XMP go back to BIOS enable "optimized default" then enable XMP then REDUCE frequency by 400mhz (7200mhz - 400mhz = 6800mhz) this can be done with ANY DDR5 frequency speed ie 7600mhz - 400mhz = 7200mhz.

So try this out with your DDR5 kit you have 6600mhz so minus 400mhz = 6200mhz then do a Prime95 and use option "Large FFT's stresses memory controller and RAM" do this for 1 hour.
After that when it is stable you also may try Memtest86 full pass. You are left with XMP cas latency which is to loose for 6200mhz you can lower the cas to make it fast.

I went through similar issues as you had and made a thread about it and it has been fixed one guy in particular to which i am very thankfull for guided me all the way here is the link of my thread hopefully it will be of use to you as it did for me:

Help with 6800mhz XMP on Z790 mb ("SOLVED")

EDIT: I wanna thank all the people who tried to help me and a special thanks goes to chimsunsun9x for guiding me all the way. It is solved albeit at a lower frequency @ 6400mhz for some reason sometimes the actual speed on DDR5 is -400mhz so example pick a 7600mhz if unstable try to lower freq...

www.overclock.net

There may be nothing wrong with your RAM nor ICM its something else. Do read my thread and follow suggestions.

I had tried downclocking even to 6000MHz on like 4 differnet Asus boards across 4-5 different kits and while it improved stability still was not totally stable as less random WHEA errors in OCCT Large Data Set Variable.

 

Though this:

From Hardware Unboxed has DDR5 6400 on 13900K just completely destroying 3600 CL14 (Which I am sure is Samsung BDie) and I am sure they run it XMP which would default to Gear 1 being 3600.

Not sure in reality how much 4200 with a worse CL16 instead of 14 but 600 extra MT/s and TFRC tweaked to 260 instead of whatever higher XMP value it auto sets narrows or even exceeds the DDR5 6400 for 13900K.

A big part of that answer is going to determine whether I give DDR5 another go or not.

I am not sure if TFRC 260 and tuned 4000 CL15 or 3133 CL15 or 4200 CL16 that GoodOldGamer uses closes that gap or if GoodOldGamer testing methodology is quite different form Steve Walton at HUB. Many if not most of same games are tested. Or some of both?

 

Have you tested all your RAM kits at JDEC default with Memtest86? This will rule out if there is something wrong with your RAM kits very important that you do test at default DDR5 4800mhz which should take around 3 hours for each RAM kit. Memtest86 is pretty good at spotting errors RIGHT away.

As for OCCT i tested memory and CPU without a single issue but it was NOT neccesary for me to do so yet i did anyway i assume the OCCT cpu test has the "Large Data Set Variable" test. The guy who guided me al the way (to which im very very grateful for) said i only needed to test with "Prime95 - Large FFT's stresses memory controller and RAM" after i done testing with ONLY 6400mhz ie Memtest etc etc the "Large FFT's stresses memory controller and RAM" this option is as close to REAL world heavy loads on the IMC and RAM ie long gaming sessions. Since it will torture test the CPU IMC (cpu memory controller) and the RAM it self. Hence he said this test is a good one for my memory controller and RAM to see if my cpu IMC can handle the load.

If it passed the 1 hour mark he said with the new CAS and 6400mhz i would have a stable solid system and so far i have had NO crashes during gaming or anything else it has been smooth sailing (knocking on wood). Previously the system did run stable but the game Arma 3 and Elden Ring did crash altogether 3 times in a time span of 2 months not that much but i was concerned. And did FOUND out the RAM XMP was the culprit (through Windows error report) especially after doing Memtest86 with the stock XMP in less then 1 minute bam...errors. Then default JDEC speed 4800mhz NO errors at all. I found the sweet spot together with his help 6400mhz and the new CAS timings (see my previous post for the CAS timings). I repeated all the stress tests:

Memtest86 full pass (around 2 hours 48 min)
Cinebench no issues
Prime95 "Large FFT's stresses memory controller and RAM" bit over 1 hour no issues.
OCCT - memory 1 hour full pass
OCCT - CPU 1 hour full pass
3DMark for my gpu no issues on various benchmarks.

My system spec since you asked:

CPU - Intel i7-13700K
MB - Asrock Z790 Taichi
RAM - G-skill F5-6800J3445G16GX2-TZ5RK (Hynix A die) 32gb, 2 x 16gb = on QVL
GPU - RTX 4090 (MSI Gaming X Trio)
Drive - Samsung 980 Pro 1TB m2.ssd
PSU - Corsair HX1500i (2022 model = ATX3.0 certified)
Case - Lian Li PC O11 WX

Cooling ALL custom water cooling:

CPU max load - 100% = 55c (capped at 125w ie LP1-125w - LP2-125w) temps on MAX intel spec ie 256w ie LP1-256w - LP2-256w = 71c (minor undervolt used) without UV = 80c
GPU idle 26c or something gaming around 37c - 45c

I researched the impact of DDR5 gaming FPS MHZ i did compared 4800 - 5200 - 5600 - 6000 - 6200 - 6400 - 6800 - 7200
And i can say DDR5 6000mhz is the SWEET spot! above that you end up with diminishing returns.
I never owned a Alder lake system with DDR4 or DDR5. I went straight to Z790 raptor lake. Since my old system 7 years old a i7-6800k was in need of a dire upgrade.

I think your MB and crappy ASUS is to blame screw them i hate to say this since i was a fan of Asus but seeing all the complaints on their XMP on Z690 and Z790 i passed on them.
Man 4 to 5 RAM kits all having issues i don't think magically every KIT was faulty it is then the MB or if the MB is working CORRECTLY which i suspect it is then its ASUS POORLY done XMP and BIOS. Stay clear from them look to Asrock or MSI or EVGA. Don't be overly concerned one can keep doing tests and tests you should only do some manditory tests for ram, cpu, gpu and you should be ok if all pass.

Either RMA the board or keep it in hopes those scum at Asus releases a PROPER BIOS fix for their sloppy XMP or get a new board from another brand.