4x sticks of DDR4 all bad at once?

[Spork3245]

Ok that's interesting.

Pay attention to the USB port you are using and use the same port for all your other USB based testing. Some ports come from the CPU and some from the chipset. Do you know which one you are using? At least the first M.2 comes from the CPU and the others from the chipset. Do you know which M.2 lanes the original boot drive was connected to?

Woke up to memtest stating “failed” for the two 3000mhz Corsair sticks I had installed, I saved the results when prompted. Tried a few suggestions in here with both the 3000mhz and brand new 3600mhz Corsair RAM to no luck. Just started running memtest with the 3600mhz sticks. If the 3600mhz sticks fail, I’m wondering if it’s a motherboard issue (again) and something to do with the memory controller? I haven’t entered Windows yet this morning via the G.Skill RAM, but I’ll post the results of both memtest runs once the 3600mhz test completes and I reinstall the G.Skill RAM to get into Windows.
Here’s a screen of the 3000mhz test result for now:
https://i.postimg.cc/0QP1mzpb/IMG-8400.jpg

I’m using one of my front USB 3.0 ports which is connected to the USB3 header (20pin) on the X570 Dark Hero. Not sure which lanes the boot m.2 is using, but it’s installed into the first m.2 port (this motherboard, despite its original price and “flagship” status at the time of release, only has two m.2 ports because… reasons I suppose lol, but both are full bandwidth gen 4)

IF this was me I would be seeing what the GEIL ram is setting as default timings vs the Corsair stuff and see if anything is being majorly set different. That way you can see if Auto is doing something stupid and setting something either stupid low or stupid high. Yes each kit is going to have differeing timings based upon ICs etc but your looking for something stupid.

16 Mins in a RAM test isnt much at all. You want at least a couple of hours of clean test once you are experiencing issues.

Apologies, it’s actually G.Skill Ripjaw RAM when I actually looked at it to post a pic of the sticker, no idea why I thought it was Geil, but I did buy it closing in on a decade ago lmao.
Per your suggestion I tried running both the 3000 mhz and 3600mhz Corsair RAM at the timings/speed of the GSkill (setting everything manually to the changes that the BIOS makes), but still unable to boot into Windows. I’ll try getting the full info via ZenTimings later today when I reinstall the GSkill RAM and get into Windows.

Some additional thoughts...

In UEFI I think you should configure

• Download & Install ARMOURY CRATE app [Disabled]
• Fast Boot [Disabled]
• PBO Fmax Enhancer [Disabled]
• Precision Boost Overdrive [Disabled]

I don't know what these things do below but I would explore if there is a Normal/Disable option. Generally I agree with Ferrum Master in turning off all the auto overclocking bs and this to me looks like something to try to disable or set to something that is not Extreme.

• DRAM Power Phase Control [Extreme] <<< Normal?
• DRAM R1 Tune [Auto]
• DRAM R2 Tune [Auto]
• DRAM R3 Tune [Auto]
• DRAM R4 Tune [Auto]

... back to your memory testing....

Assuming you get past at least 1 pass of each test in memtest86 I think that's good enough to proceed to a test boot of the linux stick using the same usb port.

• if that works then see if the windows usb will boot up into installation mode.
• if that works then popin the test M.2 into the primary slot and try the windows install again.
  • if that works then proceed with windows install (no need to register) and install chipset drivers and test system.
• if that fails change the test M.2 into a different M.2 slot and try again.
  • if that works then proceed with windows install (no need to register) and install chipset drivers and test system.
• if that fails change the test M.2 into a different M.2 slot and try again if you have a 3rd one (sorry I was too lazy to look).
  • if that works then proceed with windows install (no need to register) and install chipset drivers and test system.
• if that fails use a sata storage device to test the installation.
  • if that works then proceed with windows install (no need to register) and install chipset drivers and test system.

By doing this in my mind we trying to determine if M.2 storage may be a factor contributing to your issues and isolating CPU vs. Motherboard connectivity assuming your M.2 drive isn't a problem.


I think this is a good idea. Especially to borrow ProcODT and RTT settings that were able to successfully boot perhaps even voltages. A ZenTimings screenshot of a successful boot would be nice to see.

I typically do have the first two disabled, I may have missed them among all of the CMOS resets over the past 24-48 hours lmao
I tried the rest of your suggestions and still no luck with either the old or brand new Corsair RAM unfortunately. The FMax enhancer is an interesting one as when I googled to see what it was, it seems as though people tend to experience either no benefit or a performance LOSS with it enabled (ASUS, whatchu doing? Lol).
For DRAM Power Phase Control the options are either “extreme” or “optimal”, from the description in the BIOS it seems as though “extreme” is using default guidelines and “optimal” is using “ASUS’ [custom] optimized settings”. Such odd wording for the settings if that’s the case. But, regardless, no luck with any of the setting changes.

I do not have a spare m.2, unfortunately. I can repurpose my (currently installed) 500gb SATA SSD, however.
BUT, this morning I removed (unplugged) all (5) of my SATA drives (HDD + SSD) and removed the m.2 with Windows on it, leaving only my ~6-month old SK Hynix P41 Platinum m.2 and I’m still unable to boot the Windows Installer from USB with any of the Corsair RAM installed.

The drive and PSU are suspect components. With the drive being the most suspect.

Even with the boot m.2 removed from the system, I still cannot even get into Windows installer via USB unfortunately. The PSU is interesting, but I’d imagine forcing the same settings and voltages of the GSkill RAM would then allow the Corsair RAM to enter into Windows, no? With the GSkill installed I can run a gaming stress tests without issue as well (considering the power draw of both the 5900x and RTX 4090 under load, I’d imagine I’d have system instability if the PSU was going bad - but, maybe not, though?)

 

[A Computer Guy]

Woke up to memtest stating “failed” for the two 3000mhz Corsair sticks I had installed, I saved the results when prompted. Tried a few suggestions in here with both the 3000mhz and brand new 3600mhz Corsair RAM to no luck. Just started running memtest with the 3600mhz sticks. If the 3600mhz sticks fail, I’m wondering if it’s a motherboard issue (again) and something to do with the memory controller? I haven’t entered Windows yet this morning via the G.Skill RAM, but I’ll post the results of both memtest runs once the 3600mhz test completes and I reinstall the G.Skill RAM to get into Windows.
Here’s a screen of the 3000mhz test result for now:
https://i.postimg.cc/0QP1mzpb/IMG-8400.jpg

It will be interesting to see which tests failed and if on both runs with the different memory if the same CPU cores failed.
If the max contiguous errors are low it might simply be a matter of tweaking memory settings to fix it.

I’m using one of my front USB 3.0 ports which is connected to the USB3 header (20pin) on the X570 Dark Hero. Not sure which lanes the boot m.2 is using, but it’s installed into the first m.2 port (this motherboard, despite its original price and “flagship” status at the time of release, only has two m.2 ports because… reasons I suppose lol, but both are full bandwidth gen 4)

I would use a rear I/O usb port (more reliable) and not the one used for flashback.

 

[VuurVOS]

Woke up to memtest stating “failed” for the two 3000mhz Corsair sticks I had installed, I saved the results when prompted. Tried a few suggestions in here with both the 3000mhz and brand new 3600mhz Corsair RAM to no luck. Just started running memtest with the 3600mhz sticks. If the 3600mhz sticks fail, I’m wondering if it’s a motherboard issue (again) and something to do with the memory controller? I haven’t entered Windows yet this morning via the G.Skill RAM, but I’ll post the results of both memtest runs once the 3600mhz test completes and I reinstall the G.Skill RAM to get into Windows.
Here’s a screen of the 3000mhz test result for now:
https://i.postimg.cc/0QP1mzpb/IMG-8400.jpg

Can you set the voltages below. It might give your more stability on 3200MHz.
CPU SOC Voltage: 1.10v
CLDO VDDP Voltage: 0.90v
VDDG CCD Voltage: 0.95v
VDDG IOD Voltage: 1.00v

I am use these voltages for my 5900X with memory overclocked at 3733MHz.

 

[Ferrum Master]

it might simply be a matter of tweaking

It may be... but it is time consuming and often on the blind, you have to make an excel spreadsheet and go through values on by one and then do stability tests, it took me few weeks myself. My board is also very bitchy and I am sure it is an ASUS thing. Actually I have like 2-3 setting combos, that make my system bootable, If I deviate from the golden values, it refuses to boot at all, auto doesn't work at all.

 

[Spork3245]

Realized that memtest takes a bit over an hour per 8gb of RAM. Probably should’ve did the 2x 32gb (64gb) overnight and a pair of the 8gb (16gb) this morning lmao. Thankfully I’m not working today (I use my PC for work).

It will be interesting to see which tests failed and if on both runs with the different memory if the same CPU cores failed.
If the max contiguous errors are low it might simply be a matter of tweaking memory settings to fix it.

I certainly hope it’s something that simple. The way that the 3000mhz went from “totally fine” to “can’t boot Windows” with no settings changed is making me lean towards it could be something deeper. Not to mention the fact that dropping both speed and timings to speeds so slow it’s downright silly on any of the Corsair sticks still result in Windows unable to boot, while the 2400mhz GSkill works even with the DOCP settings of the 3000mhz Corsair sticks (the GSkill does crash if I don’t reset the speed/timings after trying the 3600mhz Corsair, though).

Can you set the voltages below. It might give your more stability on 3200MHz.
CPU SOC Voltage: 1.10v
CLDO VDDP Voltage: 0.90v
VDDG CCD Voltage: 0.95v
VDDG IOD Voltage: 1.00v

I am use these voltages for my 5900X with memory overclocked at 3733MHz.

To confirm, you want me to try this with the 3600mhz Corsair, or both the 3000 and 3600?

Just throwing this out there, do any of you think this could be a CPU issue (ie: the 3.5 year old 5900x going bad)? I’m not sure if a failing CPU would be isolated to just memory issues, though, as it’s stable in other tasks under load for hours at a time.

 

[Ferrum Master]

To confirm, you want me to try this with the 3600mhz Corsair, or both the 3000 and 3600?

Use those for any. I would go even 1.2V for SoC as a starter to exclude CPU IMC instabilities.

 

[Spork3245]

Use those for any. I would go even 1.2V for SoC as a starter to exclude CPU IMC instabilities.

Thanks, I’ll try this when memtest finishes on the 64gb (I might die of old age before then, fair warning! )

 

[A Computer Guy]

Thanks, I’ll try this when memtest finishes on the 64gb (I might die of old age before then, fair warning! )

Are you getting the errors on the first pass? If yes then you can abort subsequent passes and skip the hammer test which takes a long time and skip the bit fade test for now to speed up diagnostics sorry I forgot to mention that before. Usually if you can get through the first pass of all memtest86 tests that's good enough to start boot testing.

I like VuurVOS suggestion about fixing the voltages as suggested for the next memtest86 runs but I would keep your RAM frequency (and other XMP values) to what it's supposed to be per kit. Trying to get a Corsair 3000 kit to do 3200 at the same time we are trying to test the kit is just going to add problems. If for some reason the board wasn't applying a good default voltage VuurVOS suggestion might reveal that if the next test runs pass where they previously failed.

Please make sure between changing ram kits you reset to UEFI defaults to avoid problems with unintended values getting stuck. Unfortuanly that means you need to keep reconfiguring things in UEFI but better safe than sorry and botching your test results by mistake.

I certainly hope it’s something that simple. The way that the 3000mhz went from “totally fine” to “can’t boot Windows” with no settings changed is making me lean towards it could be something deeper.

Yea this bothers me. I believe someone in the thread asked if you tried the problematic memory in a different system but I don't recall the answer?

To confirm, you want me to try this with the 3600mhz Corsair, or both the 3000 and 3600?

Well you got the 3600 kit to replace the 3000 so it does make sense to perhaps just focus on getting the new kit working even if you have to end up down clocking the 3600 kit to 3200.
This would save you a lot of testing time and reduce the wear and tear with swapping ram in and out.
Remember 3200 is in spec and while most AM4 cpu's seemingly can do 3600 but YMMV.
Were either the old or new kits in the QVL for your motherboard by any chance?

 

[Spork3245]

Are you getting the errors on the first pass? If yes then you can abort subsequent passes and skip the hammer test which takes a long time and skip the bit fade test for now to speed up diagnostics sorry I forgot to mention that before. Usually if you can get through the first pass of all memtest86 tests that's good enough to start boot testing.

No errors on the 2x 32gb yet, it’s on pass 3 of 4 right now. Should I end it?

I like VuurVOS suggestion about fixing the voltages as suggested for the next memtest86 runs but I would keep your RAM frequency (and other XMP values) to what it's supposed to be per kit. Trying to get a Corsair 3000 kit to do 3200 at the same time we are trying to test the kit is just going to add problems. If for some reason the board wasn't applying a good default voltage VuurVOS suggestion might reveal that if the next test runs pass where they previously failed.

Please make sure between changing ram kits you reset to UEFI defaults to avoid problems with unintended values getting stuck. Unfortuanly that means you need to keep reconfiguring things in UEFI but better safe than sorry and botching your test results by mistake.

Got it, I’ll try the voltages but leave the mhz + timings alone.
Yup, I’m typically resetting CMOS each time I swap RAM (not counting the instances when I forget to tap the reset button lmao).

Yea this bothers me. I believe someone in the thread asked if you tried the problematic memory in a different system but I don't recall the answer?

I’m going to try it today in an Intel system (I think it’s an 11th gen system). I’ll try both the 3000mhz and the 3600mhz since neither are allowing Windows to boot in my system.

 

[A Computer Guy]

No errors on the 2x 32gb yet, it’s on pass 3 of 4 right now. Should I end it?

If you were typically getting your errors on the first pass then yes.

 

[Spork3245]

MemTest finished on the 2x32gb 3600mhz RAM, it passed oddly enough. Looks like MemTest saved both tests as .html, how can I upload them for you all to see?

Well you got the 3600 kit to replace the 3000 so it does make sense to perhaps just focus on getting the new kit working even if you have to end up down clocking the 3600 kit to 3200.
This would save you a lot of testing time and reduce the wear and tear with swapping ram in and out.
Remember 3200 is in spec and while most AM4 cpu's seemingly can do 3600 but YMMV.
Were either the old or new kits in the QVL for your motherboard by any chance?

If I get the 3000 working I'd actually send that 3600mhz back. I assumed that the 3000 went bad when the GSkill RAM worked, but the 3600mhz isn't working either
Both the 3000mhz and 3600mhz Corsair RAM I have are on the QVL list for the X570 ASUS Crosshair VIII Dark Hero.

I’m going to try it today in an Intel system (I think it’s an 11th gen system). I’ll try both the 3000mhz and the 3600mhz since neither are allowing Windows to boot in my system.

Tried all 4 sticks of the 3000mhz in the Intel system, and it worked without issue. Windows booted immediately and without any difficulty or troubleshooting, even with the XMP profile active. Same with the 3600mhz Corsair Vengeance RAM.

Can you set the voltages below. It might give your more stability on 3200MHz.
CPU SOC Voltage: 1.10v
CLDO VDDP Voltage: 0.90v
VDDG CCD Voltage: 0.95v
VDDG IOD Voltage: 1.00v

I am use these voltages for my 5900X with memory overclocked at 3733MHz.

Tried this with both the 3000mhz (set to 3000mhz instead of 3200mhz) and the 3600mhz (set to 3200mhz manually) and no luck still.

 

[A Computer Guy]

Tried this with both the 3000mhz (set to 3000mhz instead of 3200mhz) and the 3600mhz (set to 3200mhz manually) and no luck still.

When you changed voltages did you rerun the memtest86 or just go straight to trying to boot?

 

[Bobaganoosh]

is it always this complicated to get a 5900X to boot for people? The more I read, the more it sounds like the IMC is going bad on that chip, but I'm not a big AMD expert and everybody else here seems to want to point at everything else. I've certainly read a few threads with Ryzen 5000 IMC's not being good (on individual bad chips, not like "all of them"), so I'd be inclined to suggest you try and replace the CPU.

 

[Spork3245]

When you changed voltages did you rerun the memtest86 or just go straight to trying to boot?

With the voltage changes I tried to boot into Windows. I figured if that worked, I try to reboot to make sure it wasn't a fluke, then run memtest on it. I know that may seem backwards but memtest took about 8 hours for that 64gb combo lol

 

[A Computer Guy]

MemTest finished on the 2x32gb 3600mhz RAM, it passed oddly enough. Looks like MemTest saved both tests as .html, how can I upload them for you all to see?

If it all passed then it doesn't matter. With the passing ram kit I would now proceed to use the Linux boot USB and try it with all USB ports one at a time. Determine if any of them are successful.

 

[Spork3245]

is it always this complicated to get a 5900X to boot for people? The more I read, the more it sounds like the IMC is going bad on that chip, but I'm not a big AMD expert and everybody else here seems to want to point at everything else. I've certainly read a few threads with Ryzen 5000 IMC's not being good (on individual bad chips, not like "all of them"), so I'd be inclined to suggest you try and replace the CPU.

It was working fine for about 3 years without ever giving me issues fwiw. I'm wondering if it's the CPU as well tbh.

 

[A Computer Guy]

It was working fine for about 3 years without ever giving me issues fwiw. I'm wondering if it's the CPU as well tbh.

It is possible something is wrong with the chip or motherboard.

 

[Spork3245]

If it all passed then it doesn't matter. With the passing ram kit I would now proceed to use the Linux boot USB and try it with all USB ports one at a time. Determine if any of them are successful.

Well, the 3000mhz seems to have had two failures. Out of curiosity, let's say Linux boots and works fine - what would be the next step to get me back into my Windows install?

 

[A Computer Guy]

Well, the 3000mhz seems to have had two failures. Out of curiosity, let's say Linux boots and works fine - what would be the next step to get me back into my Windows install?

So we have attempted to validate the new ram and it passed. Going on the assumption that it's good we proceed to try to reinstall the OS with caution that perhaps the USB or M.2 is having an issue.
You would use the working USB port to attempt the windows install with only the 1 M.2 storage device installed into the motherboard.
If the install fails then now you cycle though the storage options and see if any of those work assuming the windows usb is booting correctly (which is why I want you to test that first independently with linux).

 

[VuurVOS]

Tried this with both the 3000mhz (set to 3000mhz instead of 3200mhz) and the 3600mhz (set to 3200mhz manually) and no luck still.

Can you try 2400MHz@CL15-15-15-35 1.35V with only two dimms installed?

 

[Bobaganoosh]

It was working fine for about 3 years without ever giving me issues fwiw. I'm wondering if it's the CPU as well tbh.

Yeah, but it got progressively worse, then you RMA'd the board and it seemed a little better for a while then started failing again. With a new system, the motherboard will try and train the memory and adjust things accordingly for the CPU. If the CPU is a moving target for various reasons, the results will shift with that.

Can you try 2400MHz@CL15-15-15-35 1.35V with only two dimms installed?

I was also just going to ask this.

 

[Spork3245]

So we have attempted to validate the new ram and it passed. Going on the assumption that it's good we proceed to try to reinstall the OS with caution that perhaps the USB or M.2 is having an issue.
You would use the working USB port to attempt the windows install with only the 1 M.2 storage device installed into the motherboard.
If the install fails then now you cycle though the storage options and see if any of those work assuming the windows usb is booting correctly (which is why I want you to test that first independently with linux).

Got it, thanks. I made the flash drive with Mint, but had to run out (my wife’s car won’t start, waiting for a tow). I’ll try it tomorrow.

Can you try 2400MHz@CL15-15-15-35 1.35V with only two dimms installed?

I’ll give this a shot tomorrow with the 3600mhz RAM.

Yeah, but it got progressively worse, then you RMA'd the board and it seemed a little better for a while then started failing again. With a new system, the motherboard will try and train the memory and adjust things accordingly for the CPU. If the CPU is a moving target for various reasons, the results will shift with that.


I was also just going to ask this.

Looks like I can obtain another 5000 series AM4 CPU to try tomorrow. If everything works with the new CPU, we’ll at least know the culprit now!

 

[chrcoluk]

I seems the new ones are Hynix CMRs single rank. Now Put those values in manual use speed for 3200, but timings for 4000 to be safe. But enable GDM. (gear down mode)

View attachment 323585

Nice chart, but I am curious why it bottoms out at 3200, you may think why that low, well running 4 dimms on my AM4 is really tough lol.

On the main subject itself, I think a board or CPU IMC issue is more likely than a dimm issue considering its affecting multiple dimms, bad ram seem rare now days.

The graph doesnt state if those are 2 dimm or 4 dimm timings. I remember the old now abandoned ryzen calculator used to differ quite significantly if you told it that you have 4 dimms.

 

[VuurVOS]

The graph doesnt state if those are 2 dimm or 4 dimm timings. I remember the old now abandoned ryzen calculator used to differ quite significantly if you told it that you have 4 dimms.

Ryzen Calculator isn't a great tool. It dont show the correct RTT Wr, Park, and Nom values when you have 4 single rank DIMM's installed.
To get the "correct" value you must switch to dual rank in the tool.

 

[A Computer Guy]

Looks like I can obtain another 5000 series AM4 CPU to try tomorrow. If everything works with the new CPU, we’ll at least know the culprit now!

Great I think that will be incredibly helpful. When you pop it in use your new memory kit, do a quick 1 pass memtest86 just to make sure nothing goes horribly wrong then try booting the USB and installing windows. I'll be out for the day so good luck!