Missing DIMM Channel C -- The Return

[Flyprdu]

Hello ROG Forum!

I am sad to report that 4GB of my 16GB has gone missing again. I assume that it's from channel C.
I am asking for the forum's advice on how to proceed. I'm running an RVE on 1901 and have 4 sticks of Dominator Platinum 3000 at 1.38v. (0.03v over stock)

How do I proceed to fix this? Or do I have a broken board and need to start the RMA process?

Please advise. Thanks!

[linxeye]

I'd switch CPU voltages to adaptative mode. Can you try that ? Other thing you can try is relax some timings : switch the 15 to 17 for CL.

[Praz]

Hello

Fully clear the UEFI and boot at stock settings. If all the memory is then properly detected you need to correctly tune the system for the chosen memory speed or run at a lower frequency that is within the capabilities of the IMC without the additional manual tuning.

[Flyprdu]

Praz wrote:
Hello

Fully clear the UEFI and boot at stock settings. If all the memory is then properly detected you need to correctly tune the system for the chosen memory speed or run at a lower frequency that is within the capabilities of the IMC without the additional manual tuning.

This is not the first time that this memory has a dropped a stick. The first time, I did everything you instructed and even replaced the memory kit. The memory kit will be fully recognized at stock (2133) memory speeds. But that said, it is completely unacceptable. This board costs too much for it to be wonky and to have DIMMs drop off randomly and no one really knows why. What people are dancing around is that this R5E motherboard has obvious problems with DDR4 3000 MHz sticks. Why? Is it the 125 MHz base clock? What is the official word from ASUS and is there a fix in the works? I reiterate: that Corsair Dominator Platinum DDR4 3000 was on the approved vendor list for this motherboard, but there obviously is a problem.

This needs to be addressed. UEFI resets and another 10 hours of tinkering is not why I dropped $500 US on a motherboard. It should be able to handle the XMP 2.0 Profile that it was advertised to support.

Rant over.

[Nimohtar]

Flyprdu wrote:
Hello ROG Forum!

I am sad to report that 4GB of my 16GB has gone missing again. I assume that it's from channel C.
I am asking for the forum's advice on how to proceed. I'm running an RVE on 1901 and have 4 sticks of Dominator Platinum 3000 at 1.38v. (0.03v over stock)

How do I proceed to fix this? Or do I have a broken board and need to start the RMA process?

Please advise. Thanks!

I actually have a similar problem.
MY RAM kit has 2 XMP profiles, the 2666MHz profile is fine, but I lose channel C most of the time when I use 3000MHz profile. (I also lose channel A very occasionally)
If I manually set everything up just like the XMP profile, but loosen the timing by 1, then channel C seems to work most of the time.
The above happens even when the CPU is not overclocked.

I've noticed channel B and D never drop, so I am using slot B1, B2, D1 and D2 for the time being.
It gives me quad channel and allows me to run 3000MHz XMP without messing with the RAMs.

[campires]

to set manual, not XMP and 1,35v for the memories with 17.15.15.35.2T. I have the same memories and operate correctly.

[cekim]

As Praz says, this is pretty typical symptom of an unstable overclock. Revert to stock, confirm memory is present. If not. Re-seat dimms and repeat. Then once all are present, work back up to the best frequency vs voltage setup your hardware will tolerate

[Nimohtar]

cekim wrote:
As Praz says, this is pretty typical symptom of an unstable overclock. Revert to stock, confirm memory is present. If not. Re-seat dimms and repeat. Then once all are present, work back up to the best frequency vs voltage setup your hardware will tolerate

So, which component is unstable, though?

My RAM is rated for 1.35V 3000MHz 15-16-16 T2.
If I seat them in A1, B1, C1 and D1 then use XMP, channel C is missing most of time.
If I manually set the system to use exactly what XMP uses, channel C is missing most of the time too.
If I manually set the system to 1.35V 3000MHz 15-17-17 T2, channel C is present.

I've had the RAM replaced, still happens to channel C.
I've shuffled the sticks, still happens to channel C.

If I install only A1, B1 and C1, channel C is present, but as soon as I install D1, channel C is gone again.

On the off chance that channel C is present while running 1.35V 3000MHz 15-16-16 T2, the system can run HCI Design MemTest for 1000% without error.
Real bench also ran for over 24 hours without issue.


I ended up installing the RAM in B1, B2, D1 and D2, this setup has been running 1.35V 3000MHz 15-16-16 T2 without any channel missing for over a year now.

So, maybe it's the motherboard?

[Brighttail]

I have had where multiple ram slots will stop reporting, even when not overclocking the RAM and running only on XMP.

I usually just reset to optimal settings > reboot and they come back and then setup my overclock again. It is a silly little "feature" many of these boards have and have had it happen even when I was doing no overclocking but just setting up RAID.

[cekim]

Brighttail wrote:
I have had where multiple ram slots will stop reporting, even when not overclocking the RAM and running only on XMP.

I usually just reset to optimal settings > reboot and they come back and then setup my overclock again. It is a silly little "feature" many of these boards have and have had it happen even when I was doing no overclocking but just setting up RAID.

XMP is an overclock.

When we say "stock", we mean 2133MHz. ANYTHING above that is an overclock and frankly, on X99 systems, XMP is not terribly predictable or reliable. It may or may not actually set everything and its pretty common that the sticker values of the RAM either require tuning to achieve or cannot be achieved on any given processor.

System Agent and Cache voltage (in that order) play a big part in whether your DDR OC ultimately works, but initial voltage (the one on the first extreme tweaking page), plays a determining role in your situation.

You may have to provide additional voltage over the sticker value to get it to detect all your dimms and then "eventual voltage" can be set to the sticker value. I have multiple systems and all of them eventually ended up in this situation to get 64, 128G and higher OC settings working reliably.

For example, I have a 128G 3000CAS14 kit. On one 5960x it required 1.37 initial and 1.35 final to run at that rate.