cancel
Showing results for 
Search instead for 
Did you mean: 

CMR32GX4M4C3466C16 XMP errors on X HERO

Deepcuts
Level 10
Hello,

Just posting this in case others bought the same KIT and had better luck then me.
I had to RMA my CMK32GX4M4B3200C16 because one stick out of four was throwing errors.
Because the retailer did not have the same KIT anymore, they replaced it with CMR32GX4M4C3466C16
I agreed after checking http://dlcdnet.asus.com/pub/ASUS/mb/LGA1151/ROG_MAXIMUS_X_FORMULA/MEMORY_QVL_0108.pdf and saw the KIT as qualified.

CMR32GX4M4C3466C16 is a DDR4 Corsair @3466 Mhz 16-18-18-36 (4x8GB)
At JEDEC spec, the KIT runs just fine.
At XMP tho, the system does not even boot most of the time. When it does, Windows will just freeze without any BSOD and the system will restart.
I also tried manually setting up the speed and timings without XMP. Tried Auto, Mode1 and Mode 2. Loaded BIOS defaults and only enabled XMP. Same story.
I have managed to get the system as stable as I can by setting DRAM Voltage to 1.400 instead of 1.350, but even tho I could run AIDA64 Cache & Memory and RealBench successfully for about 1 hour, Windows starts very slow and a MemTest86 run shows errors after only 4 minutes.
At the moment, I am running the KIT @3200MHz with 15-15-15-35, 1.350V and so far, it seems stable. MemTest86 does not show any errors.
Not holding my breath, but hoping for a future BIOS update that will actually support this memory KIT as stated on QVL link above.

So I guess the question is: does anyone else have this memory KIT and can run it at XMP spec? Or at least a manual setup at XMP spec?
Maybe I have such bad luck to hit two bad kits in a row?


3200Mhz with 15-15-15-35, 1.350V
70647

72329

Thank you.
272 Views
40 REPLIES 40

Can you try lowering your max cache ratio from bios to 42? It is set to "auto" by default and this makes it run a bit faster (NB frequency on cpu-z memory tab). Also I would recommend running VCCSA=1,15V and VCCIO=1,10V for ram speeds above 3200MHz, something lower like 1,10V and 1,05V respectively might work, but having both at 1,00V with those memory speeds is not really recommeded, since VCCSA defaults to 1,05V and VCCIO to 0,95V at the SPD speeds (2133MHz).

You could also downgrade to bios 0802 since I have seen some people mentioning that the bios 1003 has made their memory overclocks worse.

Renzo wrote:
Can you try lowering your max cache ratio from bios to 42? It is set to "auto" by default and this makes it run a bit faster (NB frequency on cpu-z memory tab). Also I would recommend running VCCSA=1,15V and VCCIO=1,10V for ram speeds above 3200MHz, something lower like 1,10V and 1,05V respectively might work, but having both at 1,00V with those memory speeds is not really recommeded, since VCCSA defaults to 1,05V and VCCIO to 0,95V at the SPD speeds (2133MHz).

You could also downgrade to bios 0802 since I have seen some people mentioning that the bios 1003 has made their memory overclocks worse.

I have the CMR16GX4M2C34 2x8 GB 3466 MHz CL 16 kit.

I had a lot of errors before on my "old" i7 8700K processor but I had another one here too so I tried that one in case it just was the memory controller. On ASUS Z370-F Strix with 0607.

The first screenshot below is likely memtest86 sequential test #6 block move for one hour on the ASUS Z370-F Strix BIOS 0607 with another i7 8700K. 284 errors.

At the same time I bought this motherboard I had also ordered an MSI Z370 Gaming M5 because this was on black friday and both was on sale. Though the ASUS board have better sound and the one I really wanted was the Z370-E Strix. But it's time to return items and I figured I'd put it together anyway. The second "to try if it's better" i7 8700K I've now returned so this is my same old processor which also never have worked on the ASUS board with XMP on.
The second image show this processor simply with XMP on and I think the FULL test and with parallell processor core usage.
The second image show the same processor still on the Z370 Gaming M5 (BIOS 1.20) simply with XMP on and parallell processor core usage and only test 6 block move for 1 hour 30 minutes and 52 passes with no errors.

My RAM is on the QVL for the ASUS board but it's not on the QVL for the MSI board. But clearly it work better there. Also as I've tried TWO i7 8700K I doubt it's all that much of "bad luck."

On the MSI board however if I choose sequential core usage and still do test 6 then the program simply seem to hang after 4 seconds. It stopped updating after that at-least. Tried it twice and same result both times.

HWmonitor list +0.000V for the offset value for SA don't remember what the other one was but IA, GT, LLC/Ring is all +0.000. VID min 0.670 and max 1.345.

CPU-Z says: 1741.9 MHz, FSB:DRAM 1:26, CL, tRCD, tRP and tRAS: 16-18-18-36, tRFC / Row Refresh cycle Time: 607 cycles, Command Rate / CR: 2T

No idea whatever that would be of any help on the ASUS board. It's really about my last change to return any of these so I really should make a decision. I've never used the MSI board before and at-least the Pro Carbon AC (Pro) has shown worse audio test results than the Strix one in some benchmark likely / possibly because of better input on the ASUS board since they claim they have improved that and it was likely tested using a loop between the output and input. Also the MSI board come with Killer NIC which use more CPU and did cost 200 SEK more though it's normally cheaper (came with Assassin´s Creed Origins) otherwise I guess they are pretty similar. ASUS RGB solution is much more shiny also everything whatever it's heat-sinks, capacitors for the VRM or for the audio look larger on the ASUS board. No idea if that says "quality" but assuming the ASUS board worked as well I would likely had preferred that. Now it's harder. Also my case is old and I have no cut-out so switching from one to the other force me to rebuild the whole setup.

Because of what seem to be that complete lock-up maybe I can't be sure it work on the MSI board either. Unless I get it going on the ASUS board .. I will have a hard time to decide what to do =P
Edit: I'm still upset simply changing the frequency to 3200 MHz put the memory controller and RAM at 1.5 volt!!
Plenty of people seem to have problem with the Corsair kits on the ASUS boards?

I've copied all the values which the MSI BIOS showed in case that can help run it on the ASUS board too.
CPU IO seem to be 1.25 volt and CPU SA 1.31 at-least.

(Also is there a tiny red dot at the middle row left most half of the CPU socket furthest to the right? My eye-sight and brightness isn't enough to see. Maybe it's a piece of dust? How convenient! https://pasteboard.co/H7mqKzu.png (Never mind, if there ever was a thing there it's gone now. Can't see anything with my eyes at-least, there seem to be a ridge at the plastic to the right of it, as can be seen on the photo, maybe that was what I saw or there was a dust spec which has moved since.)

Raja
Level 13
I'd try tuning the sa and io voltages before you do that.

Deepcuts
Level 10
Thank you Raja.
Will try tuning SA and IO some more.

Deepcuts
Level 10
Tried increasing SA and IO in 0.10 steps up to 1.360V each and DDR Voltage at 1.450V. No luck with XMP or manual 3466Mhz.
So I decided to leave the speed at 3200Mhz and lower the timings.
So far, 3200Mhz 13-13-13-30 at 1.400V. Rock solid. I am but a beginner in overclocking my I would say those are some nice timings for 3200Mhz.
SA and IO at 1.050V. Works with 1.000V also but 0.050V is extra on the house, because I am feeling generous.
CPU Standby at 1.000V
RAM temperatures never rises above 44 Celsius under load.
70787

Corsair replied to my ticket saying that in their opinion, the BIOS on X HERO is to blame.
So I am waiting for a miracle BIOS update.

Raja
Level 13
You still haven't tried voltages between 1.10v and 1.30V For 3466? Has been suggested twice that you try tuning the voltages in both directions. Go in small steps. BIOSES cannot be tuned for every type of CPU; some are more sensitive to sa and io voltages than others.

Deepcuts
Level 10
"Tried increasing SA and IO in 0.10 steps up to 1.360V each" is missing a 0.
I tried upping the voltage in 0.010 steps up to 1.360V, from 1.000V for both SA and IO, just like you said.
Every time (if the system booted) ran MemTest86 until errors were found. Then up the voltage again.
That would be almost a full day just doing that 😞
All I am seeing are Volts and decimals now.

later edit: I need a couple of days to forget the "horror" and will try from 1.360 backwards.


Deepcuts wrote:
"Tried increasing SA and IO in 0.10 steps up to 1.360V each" is missing a 0.
I tried upping the voltage in 0.010 steps up to 1.360V, from 1.000V for both SA and IO, just like you said.
Every time (if the system booted) ran MemTest86 until errors were found. Then up the voltage again.
That would be almost a full day just doing that 😞
All I am seeing are Volts and decimals now.

later edit: I need a couple of days to forget the "horror" and will try from 1.360 backwards.




Going downwards from 1.36V isn't going to be any different from coming upwards towards it.

Menthol
Level 14
3200 13-13-13 is probably better than 3466 at looser timings, I would stick with that

Deepcuts
Level 10
Replaced the KIT with an identical one. Same story.
So, more testing with interesting results.

Tested each stick of RAM, one at a time in A1.
Loaded XMP at 3.466Mhz with 1.350V DDR voltage. Rest Auto.
Each stick in A1 was rock solid. No errors in MemTest86.

Then, I tested two sticks of RAM. Same settings
A1 and B1 rock solid. No errors in MemTest86.
A2 and B2 rock solid. No errors in MemTest86.

Took out the first two sticks and put in the other two.
A1 and B1 rock solid. No errors in MemTest86.
A2 and B2 rock solid. No errors in MemTest86.

Just for fun, with two sticks of RAM in A1 and B1, I setup SA, IO and CPU Standby at 1.000V
Ran AIDA64. Rock solid.
Ran MemTest86. No errors whatsoever.

Setup SA, IO and CPU Stanby back to Auto and inserted all four stick. System did NOT post at all. It kept cutting power and powering up again until it posted in safe mode.
Setup SA, IO and CPU Stanby to 1.000V and same story. No go.

So my conclusion: all four sticks of RAM are OK.
For some reason, Maximus X HERO does not like foursome with my sticks 🙂

/rant
I would like to postulate that the QVL list is a bit BS. Either the dep. who does the testing only tests one or two sticks out of every KIT, or they just pick random/popular KITs and paste in QVL.
/endrant

Opened a ticket with Asus and ordered a Voodoo doll and pins for the people responsible for QVL.