At wits end. 4 dimms WONT run properly on maximus x formula (macrium reflect errors)

[klepp0906]

Oh god, so without making this as long as a book. My ram has been rock stable for a loooooong time. However ive foolishly spent too many years without backups and began using macrium reflect recently. Who woulda known its a hardcore stability test.

The problem is - my images all fail verification, but only full images, and only to internal drives.

two things allow me to pass verification. 1) write to a usb drive. 2) pull 2 sticks of my memory. 4x8 CMR16GX4M2C3600C18

The ram passes prime. Ive ran hci memtest pro overnight with no errors. I have no issues elsewhere.

After spending not days, but weeks on this, and blasting my disks with many terabytes of written and re written images to test, im spent.

I ended up checking the QVL (and while before i never put much stock in them, now i begin to wonder) and while my memory is ON the list, its only on it for 2 sticks - not 4.

Thing is, ive tried it on auto which bumps the VCCSA/IO up to 1.4+, ive tried knocking it down from 3600 to 3200 at the same timings. Ive tried with my cpu at stock.

Literally the ONLY two things that can get me to pass is pulling two sticks (yes ive tried pair by pair, and in both sets of slots) or using an external drive.

What on gods green earth is going on?

Anybody using macrium and 4 dimms able to verify problem free? I really dont know where to go with this. I certainly wish I could pull two sticks, but I cant. I'm a heavy multitasker and vm user and id hit that wall more often than not.

[klepp0906]

Enderwiggin03 wrote:
Something else, are these SATA drives you're having issues with? Might want to swap SATA cables/ports for trouble shooting proposes.

they are yes. and I did try that. Was about to try again when I ultimately (fingers crossed) stumbled on root of the issue. Common sense told me thats what it was, but i was operating under constraints/in a box I shouldnt have been.

I just couldnt fathom how or why i could be limited to below 3600. I mean they wont do it for anything. (except when I put the voltages to auto and asus wanted to give them over 1.4v each)

I guess the IMC on this chip is just THAT bad. Still doesnt explain why asus employees are on record on the forums saying not to go over 1.3ish, and the bios itself lists those voltages as RED as opposed to pink or yellow, yet they use them by default. if taken at face value, preeetty reckless.

Anyhow, more ~30m a pop verifies, until it wont anymore. Then im gonna bump the voltage back up a peg. Run another full image create/verify. and call it.

So hopefully by the end of the night I can report back with the final tally 😛 For science

[Enderwiggin03]

1.4v is entirely too much, I'm running at 4000mhz for my kit, and only running 1.264v VCCIO and 1.280v VCCSA. I've been reading above 1.3v+ can degrade IMC. Most motherboard auto settings are way too high.

[klepp0906]

well... its a bit early to say we're there. and im feeling just as much disappointment as relief.

BUT - it seems for whatever reason (unless anyone can think of something else to try) these just wont do 3600 with a reasonable amount of IO/SA voltage. Kick them down just one peg to 3466 (actually thats two pegs, 3500 is allowed as well but it just looks funny to me lol), and I'm running and verified twice now. Basically the voltages (io/sa) that are enough to pass everything BUT images (including 24hr prime ffs) at 3600, arent enough to image even at 3466.

I moved to 3466 and set IO/SA at auto. This gave me 1.32xx IO and 1.26XX SA. Ran/passed. So i moved the SA back down to what I was using for 3600 (1.23XX) and ran again. Good. Now im walking the IO back. currently doing a 1.25XX run. Seems like SA was always okay and for some reason the chip needs alot of IO voltage.

I dont know the difference between them but I think i ultimately found my problem.

Should have been found sooner but it was complicated by the fact that A) i was stuck on them being able to do XMP especially considering theyre b die. B) The general e-consensus that 1.3v was too much for 24/7 and would degrade IMC. C) 3600 even with tightened timings would pass prime/hci/ibt/aida etc at the previous settings.

Seriously, if you think your memory is stable - image your discs. Verify. Still amazed thats whats catching errors.

Relieved I know for 100% im now stable through and through. Still sad and a bit perplexed as to how I can have such potentially decent memory, that I cant even get XMP out of.

At least the chip itself is a decent clocker. 5 at 1.31v. I'm on custom water so im sure 5.1 is relatively easily obtainable but this build has to last while, so i'm taking it easy. Not really chasing the numbers like I used to 😛

BTW, arent you supposed to keep I/O voltage BELOW the VCCSA? Cause its sure not working out that way in this case. I sure hope i dont need to arbitrarily bump my SA voltage due to some technicality beyond my feeble brain 😛

[Enderwiggin03]

klepp0906 wrote:
well... its a bit early to say we're there. and im feeling just as much disappointment as relief.

BUT - it seems for whatever reason (unless anyone can think of something else to try) these just wont do 3600 with a reasonable amount of IO/SA voltage. Kick them down just one peg to 3466 (actually thats two pegs, 3500 is allowed as well but it just looks funny to me lol), and I'm running and verified twice now. Basically the voltages (io/sa) that are enough to pass everything BUT images (including 24hr prime ffs) at 3600, arent enough to image even at 3466.

I moved to 3466 and set IO/SA at auto. This gave me 1.32xx IO and 1.26XX SA. Ran/passed. So i moved the SA back down to what I was using for 3600 (1.23XX) and ran again. Good. Now im walking the IO back. currently doing a 1.25XX run. Seems like SA was always okay and for some reason the chip needs alot of IO voltage.

I dont know the difference between them but I think i ultimately found my problem.

Should have been found sooner but it was complicated by the fact that A) i was stuck on them being able to do XMP especially considering theyre b die. B) The general e-consensus that 1.3v was too much for 24/7 and would degrade IMC. C) 3600 even with tightened timings would pass prime/hci/ibt/aida etc at the previous settings.

Seriously, if you think your memory is stable - image your discs. Verify. Still amazed thats whats catching errors.

Relieved I know for 100% im now stable through and through. Still sad and a bit perplexed as to how I can have such potentially decent memory, that I cant even get XMP out of.

At least the chip itself is a decent clocker. 5 at 1.31v. I'm on custom water so im sure 5.1 is relatively easily obtainable but this build has to last while, so i'm taking it easy. Not really chasing the numbers like I used to 😛

BTW, arent you supposed to keep I/O voltage BELOW the VCCSA? Cause its sure not working out that way in this case. I sure hope i dont need to arbitrarily bump my SA voltage due to some technicality beyond my feeble brain 😛

Sometimes too high VCCIO/VCCSA is just as bad as too low of voltage.

[klepp0906]

Enderwiggin03 wrote:
Sometimes too high VCCIO/VCCSA is just as bad as too low of voltage.

ha - hadnt refreshed the page. and you said what appears to be the case potentially

[mdzcpa]

Seems like you are already sorting this out.

I run Macrium with 4 sticks (32 GB) at 4000mhz. SA/IO at 1.2v with 9900K on Hero XI.

As you have likely surmised by now.... never ever match 2 kits together. Always get the 4 stick kit. They are binned that way. The QVL is indeed very helpful, so do not disregard it.

The slightest memory instability can be the most annoying thing. Sometimes it doesn't easily reproduce. It's not consistent. Because you have 2 similar kits the instability is very slight. And your IMC seems to be on the week side. What makes your system "usable" for less demanding tasks is the memory training built into these later CPUs the last few years.

If you want to run rated speeds, you will have better luck with a 4 stick kit.

[cniedzi]

klepp0906 wrote:
well... its a bit early to say we're there. and im feeling just as much disappointment as relief.

BUT - it seems for whatever reason (unless anyone can think of something else to try) these just wont do 3600 with a reasonable amount of IO/SA voltage. Kick them down just one peg to 3466 (actually thats two pegs, 3500 is allowed as well but it just looks funny to me lol), and I'm running and verified twice now. Basically the voltages (io/sa) that are enough to pass everything BUT images (including 24hr prime ffs) at 3600, arent enough to image even at 3466.

I moved to 3466 and set IO/SA at auto. This gave me 1.32xx IO and 1.26XX SA. Ran/passed. So i moved the SA back down to what I was using for 3600 (1.23XX) and ran again. Good. Now im walking the IO back. currently doing a 1.25XX run. Seems like SA was always okay and for some reason the chip needs alot of IO voltage.

I dont know the difference between them but I think i ultimately found my problem.

Should have been found sooner but it was complicated by the fact that A) i was stuck on them being able to do XMP especially considering theyre b die. B) The general e-consensus that 1.3v was too much for 24/7 and would degrade IMC. C) 3600 even with tightened timings would pass prime/hci/ibt/aida etc at the previous settings.

Seriously, if you think your memory is stable - image your discs. Verify. Still amazed thats whats catching errors.

Relieved I know for 100% im now stable through and through. Still sad and a bit perplexed as to how I can have such potentially decent memory, that I cant even get XMP out of.

At least the chip itself is a decent clocker. 5 at 1.31v. I'm on custom water so im sure 5.1 is relatively easily obtainable but this build has to last while, so i'm taking it easy. Not really chasing the numbers like I used to 😛

BTW, arent you supposed to keep I/O voltage BELOW the VCCSA? Cause its sure not working out that way in this case. I sure hope i dont need to arbitrarily bump my SA voltage due to some technicality beyond my feeble brain 😛

I have similar problem but with WinRar CRC errors (https://rog.asus.com/forum/showthread.php?110378-Looking-for-someone-with-i9-9900k-Maximus-XI-HERO-T...).
I've resolved it by using memory Maximus Tweaker Mode 2 (instead of Auto or Mode 1). Didn't try to play with IO/SA voltages too much...

What Mode do you use?

[klepp0906]

i hate ram. its so broken.

now im passing verification @3466 at 1.10 SA and 1.05 IO. something fishy going on here.

either 3600 is the cold stone hard cutoff for my IMC, or perhaps my chip likes way less voltage?

I aaaaaalmost want to try 3600 again, but im scared. It was daaays of tests and back and forths to get 3466 running, and with that memory retraining stuff - changing anything at this point that triggers that scares me to my core 😛

[kens30]

klepp0906 wrote:
i hate ram. its so broken.

now im passing verification @3466 at 1.10 SA and 1.05 IO. something fishy going on here.

either 3600 is the cold stone hard cutoff for my IMC, or perhaps my chip likes way less voltage?

I aaaaaalmost want to try 3600 again, but im scared. It was daaays of tests and back and forths to get 3466 running, and with that memory retraining stuff - changing anything at this point that triggers that scares me to my core 😛

Hi Klepp0906,

I am glad you are finally sorting your issue with your ram out.
I suggest you just leave your ram at 3466mhz for now that you have gotten it stable, you are not losing much on performance compared to 3600mhz.
Make all your backups and fresh install Windows as your unstable memory before has for sure corrupted your Windows install even though it seems to be working ok.

Just to let you know my ram 3200mhz (samsung b die) can run at 3466mhz 1.35v cl 14 -14- 14- 34 cr 2 at the default VCCIO/VCCSA voltages 0.950v for IO and 1.050v for SA.
Leaving my RAM on auto gives way to much voltage around 1.35v io,1.25v sa .
I got lucky with my IMC needing only default voltages.

You might want to read this...

https://rog.asus.com/forum/showthread.php?57038-Don%92t-combine-memory-kits!-The-meat-and-potatoes-o...

[Phoenix2000]

Hi klepp0906

At first please download Thyphoon burner and look that the IC´s on the kit´s the Same, by Corsair you see at also at the Version number, like 4.31 for Samsung B-Die.At first i had 2 different Kit´s FLareX and Trident-Z RGB, this is A1 and A2 PCB Samsung B-Die.

The only Way to get this really stable was to put Kit 1 in Slot A1/B1 and Kit 2 in A2/B2.The same do if you have the same Kit´s.
If you dont do this this the Board has problem´s to set the RTL/IOL.You can have by Corsair same Kit´s with different ic´s.

The next Thing is if you have Mode1 and Mode2 use Mode1 by Dram Setting.This Board dont like different Kit´s to get this stable it´s not easy.
The maximum with A2 PCB B-Die Gskill stable i can reach is so 4500 in slow and fast 4350Mhz with Custom sub´s 4x8GB on Wifi XI.
With mixed B-Die A1-PCB/A2-PCB in the right Slot´s 4133Mhz stable.

If you really cant get stable, put the CPU and Memory out, clean the Gold contact an be really sure that all is fine.That´s the most important Thing, alsoo the soket pin musst be perfect, take ther also a look that no pin be "down".