Ryzen reboot at idle - Solved? (RMA)

[mtavel]

I know there has been a lot of chatter about 3xxx and 5xxx (and maybe earlier) Ryzen systems rebooting at idle.  Just browsing the web, listening to music, or just letting the PC sit - leads to the PC rebooting.  This exhibits itself as:

• PC is at idle and the screen goes black, but the system still has power (not the PSU clamping)
• PC reboots and no real errors in Event Viewer beyond (the last shutdown was unexpected)

It looks like most people seeing this are running PBO, even with stock PBO settings (just enabling PBO and leaving everything to Auto).  The shutdown/reboots are not related to thermals (seen it happen @42c and lower die temps) and not due to memory instability.  I can run Memtest for hours with no issues - and in fact - it never reboots under any substantial load like a Memtest, Cinebench, 3dMark or PCMark, etc. Only at idle.

 

What I've found:

It looks like the low power state is lowering the core voltages below stable minimums.  It's normal for cores to go to 'sleep' and for voltages to drop to a fraction of a volt - but seems like it's being too aggressive.  

 

If I configure a + Offset for the CPU voltage (where it adds the specified offset voltage to the standard voltage), it seems to eliminate the issue.  I'm still doing more testing to validate - but I was going insane with this issue previously, with reboots happening several times a day.  I have +.03750 Offset now, that may be more than needed, I'm still trying different values.  Also, this may raise your idle temps a bit.  Mine went from 43-44 to 45-46. 

 

Has anyone who has experienced this issue found a workaround that resolve the issue for them?

 

Thanks

[Guest]

Hi and welcome to the forums. 

 

I'm your friendly personal astrophysicist (actually PC enthusiast and hobbiest) and I'll try my best to help explain about restarts, and why they happen. 

 

So mostly from a lot of reading and discussion, helping people try and remedy the idle restarts, a lot of the time the OS will throw a Kernel 41 error. Which is basically saying the power went out unexpectedly and generally does not help the user with this type of symptom.

 

I take notice this seems to happen mostly when XMP is enabled in people's systems. At stock, almost never an issue while running 2133mhz stock. Well, the memory will throttle voltage when in an idle state. This can cause the system to simply restart. No real accompanying errors, just a sudden restart.  Bios posts normal, almost never in safe mode when this occurs also.

 

The only thing I have been able to pin my finger on to aid the situation is to eliminate throttling. When XMP is enabled, the bios does not disable power down mode in the sub timings menu. This setting stops the memory from going into low power states. 

 

Also, processors are also known for low volt issues, but almost always ends in a frozen screen and system not responsive. This is how I can distinguish between a Cpu low volt hang and memory (or other) low volt symptom. 

 

Hopefully this provides some guidance for those that have this issue. 

 

Have specific questions based on this theory, feel free to ask.

 

~ŞɧʁɨmǷβʁɨmЄ

[mtavel]

Thanks for the reply!  I'll definitely look into this.  I haven't gone down the path of memory settings and really started focusing on CPU power when I noticed my reboots stopped after adding a CPU voltage offset. I do typically have XMP enabled but still had the reboot with XMP disabled.  I know offset isn't an ideal solution for several reasons, but it's better than frustrating and frequent reboots.

 

I'll look into this and post a follow-up.  Thanks again!

[Plermpel]

I ran a 1600X a 3600x and currently a 3700X and a 5950X since today and never have experienced anything like that.

[Guest]

1 hour ago, mtavel said:

Thanks for the reply!  I'll definitely look into this.  I haven't gone down the path of memory settings and really started focusing on CPU power when I noticed my reboots stopped after adding a CPU voltage offset. I do typically have XMP enabled but still had the reboot with XMP disabled.  I know offset isn't an ideal solution for several reasons, but it's better than frustrating and frequent reboots.

 

I'll look into this and post a follow-up.  Thanks again!

Yes, I've read also some cpu v-core increase via positive offset or increased LLC can also help with restart issues also. 

There seems to be a large grey are where people have the problem, but yet in google search never a definitive answer.

Besides that we know it's a power issue and the motherboard doesn't like it. 

 

I have experienced this with 220ge and certain memory that would be installed. 

One set is Corsair Vengeance LED 3000 sticks. This set installed after my 3600mhz B-die took several attempts to post.

Then it was a stability nightmare, mostly none of Cpu alterations seemed to make a great deal of difference. 

Realizing that setting XMP doesn't always work, I found timings that the board while on auto, sets incorrectly.

Checking the SPD profiles in bios and manually inputting them seems to help quite a bit even with a memory voltage bump also.

But Gear Down Mode, in accompany with manually setting up correctly the timings bios SPD states, seemed to help a great deal.

[mtavel]

I reset my BIOS to defaults and kept it (fairly) basic while implementing your recommendation to disable power down mode in the memory timings. I also enabled PBO (full auto) and enabled XMP (manually checking and setting the appropriate timings for my G. Skill Z Neo 3600 as I also found some of the auto timings being quite out of whack).

 

IT WORKED!  No more random and frequent shutdowns at low utilization/idle!  I have a usable PC again!  Thanks so much for your suggestion.  I was DEFINITELY dissatisfied with the CPU voltage offset workaround.  The memory power down mode change had no discernable negative side effects (other than maybe a small increase in power draw which I won't lose sleep over) and was an immediate fix.

 

Since I didn't include this info initially, my build is:

• 5950x
• Asus Crosshair VIII Dark Hero
• 850w Seasonic Prime GX-850
• 64gb G. Skill 3600 CAS16
• EVGA 3080 XC3 Ultra

Thanks again for taking the time to respond with some great guidance!

[Guest]

8 minutes ago, mtavel said:

I reset my BIOS to defaults and kept it (fairly) basic while implementing your recommendation to disable power down mode in the memory timings. I also enabled PBO (full auto) and enabled XMP (manually checking and setting the appropriate timings for my G. Skill Z Neo 3600 as I also found some of the auto timings being quite out of whack).

 

IT WORKED!  No more random and frequent shutdowns at low utilization/idle!  I have a usable PC again!  Thanks so much for your suggestion.  I was DEFINITELY dissatisfied with the CPU voltage offset workaround.  The memory power down mode change had no discernable negative side effects (other than maybe a small increase in power draw which I won't lose sleep over) and was an immediate fix.

 

Since I didn't include this info initially, my build is:

• 5950x
• Asus Crosshair VIII Dark Hero
• 850w Seasonic Prime GX-850
• 64gb G. Skill 3600 CAS16
• EVGA 3080 XC3 Ultra

Thanks again for taking the time to respond with some great guidance!

Most welcome!

[mtavel]

To provide an update, while the change to prevent memory power down (in the memory timing settings menu) went a long way to reduce the number of low idle shutdowns, it did not completely eliminate it.  

 

It's been quite frustrating because I can run memory tests (memtest86) for 7+ hours with no errors, and I can run CPU load tests as long as I care to and not get any errors or thermal throttling, but during low utilization/idle time, the screens go black and the system reboots with no errors in event viewer.  Just complaining that the previous shutdown was unexpected (yeah Windows, unexpected for you and me both).

 

So, any additional tips are appreciated as always, and I'll keep this updated if/when I learn more.  Thanks everyone for making a great community!

[Guest]

4 hours ago, mtavel said:

To provide an update, while the change to prevent memory power down (in the memory timing settings menu) went a long way to reduce the number of low idle shutdowns, it did not completely eliminate it.  

 

It's been quite frustrating because I can run memory tests (memtest86) for 7+ hours with no errors, and I can run CPU load tests as long as I care to and not get any errors or thermal throttling, but during low utilization/idle time, the screens go black and the system reboots with no errors in event viewer.  Just complaining that the previous shutdown was unexpected (yeah Windows, unexpected for you and me both).

 

So, any additional tips are appreciated as always, and I'll keep this updated if/when I learn more.  Thanks everyone for making a great community!

Well headed for the right track...

 

Look in the bios at the SPD for the memory.

Check the timings that bios is setting. 

Change any incorrect timings manually.

Then do some idle testing.

[mtavel]

Feb 2 Update

 

Based on previous troubleshooting, my idle reboots appeared to be related to CPU instability at low power (since I can run CPU and Memory stress tests for hours and hours with no issues, but an hour or two of idle time leads to a reboot).

 

My new hobby is watching Blue Screens and Event Viewer logs   I have seen several Blue Screens (bugchecks) that displayed WHEA faults on the error screen, but the dmp file was never written out and no record of it existed in Event Viewer.  I started suspecting that some proportion of the unlogged reboots might not just be abrupt shutdowns, but may be WHEA errors that weren't getting logged because of the nature of the fault.  

 

I started looking at my WHEA errors and recording which APIC (which thread on which core) the error was attributed to.  My WHEA errors consistently looked like this one:

 

Quote

 

A fatal hardware error has occurred.

Reported by component: Processor Core

Error Source: Machine Check Exception

Error Type: Cache Hierarchy Error

Processor APIC ID: 27

 

 

Over the 16 day period the event log covered, I had 42 WHEA errors logged with the following APIC distribution (simplified to cores):

 

 

It's pretty clear that core 14 on CCX 1 is a main offender.  I then went into Curve Optimizer and applied a positive curve offset to the cores with attributed errors (more offset for more error instances).  This did help. but I was still seeing (less frequent) reboots.  

 

The most recent troubleshooting step I tried was to disable "Global C-State Control" which prevents the cores from going to sleep.  This has several real drawbacks including:

• Increasing power draw and heat at idle (I'm now at 50 degree idle temps at the desktop, up about 7-8 degrees C when I had sleep enabled).
• Decreased single core boost since that's apparently partially dependent on how much additional power budget is available to the active cores from other cores being asleep.
• Less thermal headroom for bursty workloads.  If the CPU is always 8 degrees hotter than usual, that's 8 degrees less headroom before hitting tjmax.

Disabling C-States appears to have eliminated the reboots completely.  I'm hesitant to say it's a fix, since I've experienced settings that improved stability for a bit before the reboot behavior returned, so I'm keeping an eye on it.  But it may be worth a try if you're seeing the issue.  

 

It seems like the reboot issue is a real result of the silicon lottery and the quality of the cores in your CPU.  As you can imagine, a 5950x has a lot more probability of having a core that performs poorly at low voltage compared to a 6 core part.  I can understand people that have replaced the CPU and resolved the issue (finally won the silicon lottery) and those that had the issue appear even after replacement (lost the lottery twice).

 

I will be looking into RMA'ing this CPU, but not looking forward to the downtime considering the stock issues.

[mtavel]

Feb 3 Update

 

I started the RMA process for the 5950x.  In preparation for an unknown period of time without a CPU, I bought a 5600x at Micro Center.  I replaced my 5950x with the 5600x and reset to BIOS defaults again (which enabled Global C-State Control) and I also enabled PBO and XMP.  

 

With all the same hardware (except for the new CPU), the idle shutdown/reboot issues (and WHEA errors) have not returned.  I'll keep an eye on it, but so far it looks like the same system that was experiencing unexpected idle restarts as often as several times an hour is now running reliably with the new CPU - with no tweaked settings needed (Memory power down mode enabled as default, C-States enabled as default, no positive CPU voltage offset or PBO curve needed, etc).  It just works (so far).

 

I'll keep you updated about ongoing stability and RMA turn around time, etc.

 

Thanks again everyone for your help with this.  It's been insanely frustrating but I'm happier than ever to have such a great community out there to help.

 

[Guest]

18 minutes ago, mtavel said:

Feb 3 Update

 

I started the RMA process for the 5950x.  In preparation for an unknown period of time without a CPU, I bought a 5600x at Micro Center.  I replaced my 5950x with the 5600x and reset to BIOS defaults again (which enabled Global C-State Control) and I also enabled PBO and XMP.  

 

With all the same hardware (except for the new CPU), the idle shutdown/reboot issues (and WHEA errors) have not returned.  I'll keep an eye on it, but so far it looks like the same system that was experiencing unexpected idle restarts as often as several times an hour is now running reliably with the new CPU - with no tweaked settings needed (Memory power down mode enabled as default, C-States enabled as default, no positive CPU voltage offset or PBO curve needed, etc).  It just works (so far).

 

I'll keep you updated about ongoing stability and RMA turn around time, etc.

 

Thanks again everyone for your help with this.  It's been insanely frustrating but I'm happier than ever to have such a great community out there to help.

Well it's a step in the right direction. Hope the replacement is good to go without errors. One thing is certain, the 5950X is no slouch on a power delivery system at load. It's not the voltage, it's the current (amps) it can draw. Probably pegs the EDC power limit no doubt when at load.

[mtavel]

2 minutes ago, ShrimpBrime said:

Well it's a step in the right direction. Hope the replacement is good to go without errors. One thing is certain, the 5950X is no slouch on a power delivery system at load. It's not the voltage, it's the current (amps) it can draw. Probably pegs the EDC power limit no doubt when at load.

The 5950x is definitely a completely different animal in regards to all core power draw during heavy load compared to the 5600x.  No question you can hit EDC limit and get the package temp up to 90 C even with decent water cooling.  But that's what was so frustrating about this issue, the only time I could rely on my PC being stable was under full load.  Once I was out of Prime95 or Cinebench or whatever and the fans started spinning down and the cores started going to sleep - that's where the danger lurked and the reboot gremlins came out!  "Hey user, it looks like you're not hammering all your cores at the moment, I guess you won't mind shutting down and starting from scratch!" 

[Guest]

4 minutes ago, mtavel said:

The 5950x is definitely a completely different animal in regards to all core power draw during heavy load compared to the 5600x.  No question you can hit EDC limit and get the package temp up to 90 C even with decent water cooling.  But that's what was so frustrating about this issue, the only time I could rely on my PC being stable was under full load.  Once I was out of Prime95 or Cinebench or whatever and the fans started spinning down and the cores started going to sleep - that's where the danger lurked and the reboot gremlins came out!  "Hey user, it looks like you're not hammering all your cores at the moment, I guess you won't mind shutting down and starting from scratch!" 

Lol. Have been there many times. But haven't bought into 3000 or 5000 series chips. The one thing I noticed when "tweaking" Zen and Zen+, the reference clock dips on too much under-volt to compensate. This reduces consumption slightly, but at risk of transistors stopping. 

The SOC also another chip that likes to leak and really doesn't appreciate much under-volt might make a system unstable perhaps from a momentary droop. 

 

WHEA errors also a low volt deal on the Cpu core. I often wondered if anyone has tested the hang and whea errors with SMT disabled for any length of time. No surprise would it be a bump in v-core cures the issue with all threads running. 

I suppose there's quite a bit to it all. In conjunction with board and memory combinations, anything can happen. 

 

 

[mtavel]

RMA Process & Experience so far:

 

Sunday Jan 31 @ 5pm: Submitted initial RMA request form

Sunday Jan 31 @ 8pm: Received automated confirmation with a service request ticket number

Wednesday Feb 3 @ 5:45am: Received a response and service history request from AMD support requesting:

• Pic of the CPU installed in the MOBO (cooler removed with model and S/N visible)
• Original vendor invoice (pdf)
• Make/Model/Bios Version of the MOBO
• Details of the issue & troubleshooting steps performed

Wednesday Feb 3 @ 3:35pm: Replied with requested information.  Received automated confirmation of new service request details @ 3:37pm

Wednesday Feb 3 @ 9:57pm: Received "RMA Approved" message with details.  Received a FedEx Ground shipping label in a separate email.

Thursday Feb 4 @ 8am: Shipped the 5950x to AMD using the provided shipping label (handed to employee at a "FedEx Office" shipping location)

• CPU shipped back in its original plastic clamshell to protect the pins, and also in a rigid cardboard box with enough bubble wrap and foam to withstand a SpaceX Starship SN9 landing attempt.

Monday Feb 8 @ 1pm: 5950x was received at the service center in Miami, FL.

Tuesday Feb 9 @ 6:37am: Received confirmation that the my processor:

• Quote

  • has successfully passed the inspection and your replacement product is now approved

Tuesday Feb 9 @ 6:47am: Received another update indicating:

• Quote

  • Your replacement processor is ready to ship.
    If you do not receive your replacement processor within next 5 business days or have any other queries with respects to this RMA, then please submit an online service request...

• No shipment tracking number provided.  Only RMA number and serial number provided for tracking the request.

Tuesday Feb 9 @ 7:20pm: Shipping label created by service center (2-day Fedex). Package not yet received by Fedex.

 

Thursday Feb 11 @ 5:04pm: FedEx received the package from the service center.  Expected delivery on Monday Feb 15th.

 

Monday Feb 15 @ 9:07am: FedEx delivered the package (signature required). New CPU in full retail box packaging.

 

The new CPU batch number is 2104PGS (produced the 4th week [Jan 25-31] of 2021 in Penang).  I received the CPU in the 7th week, so not a lot of time in transit.  My previous CPU was 2043PGS (mfg. Oct 19-25th 2020 also in Penang).

 

I'll swap in the new CPU a bit later today and post a reply to this thread with an update after I have some time to test.

 

I'll keep this updated as I go through the process.

[suizman]

Hi everyone, this is my first post in this forum. I unable to post to ASUS or G. Skill forum because of the activation process is taking eons, anyways I'm not here to rant about that xD

 

Setup

 

NZXT H710i
CPU AMD Ryzen 5 5600XT in exchange of Ryzen 5 3600 while waiting for 5950x stock...
MB Asus ROG CROSSHAIR VIII DARK HERO
RAM G.Skill Trident Z Neo DDR4 3600 PC4-28800 64GB 4x16GB CL16 (F4-3600C16Q-64GTZNC)

GPU PowerColor Red DEvil RX5700XT 8GB
SSD Samsung 980 Pro M.2 1000 GB
Cooler AIO NZXT Kraken X73 360 mm
PSU Corsair HX1200W Platinum

 

First at all, @mtavel I have the vary same symptoms, my PC can handle stress tests forever, CPU, RAM, GPU under 100% no problem with or without XMP profiles, stock OC, PB, PBO whatever. But!!! idle oh man.... idle with XMP default settings is a real nightmare as you've proved on your on flesh...

 

While debugging I've also seen the WHEA errors, 4 in total. I thought my provisional Ryzen 5 3600 was faulty and cannot handle 3600Mhz RAM so I exchange it in the store for the Ryzen 5600XT, but unfortunately the issues persist... I also have the Dark Hero motherboard with the latest BIOS 3204 , but this error also happens with the stock BIOS which in my case was 2601.

 

So, for this last week I was scraping the whole internet reading about Ryzen issues related to XMP profiles and low voltages in power saving modes... I found this thread from the same guy in Asus and G. Skill forum. He fixed this issue increasing DRAM voltage from default 1.35 to 1.36. What I don't know is what he exactly means by "system instability", idle, random crashes?? I cannot ask him because of the account activation issue....

 

As I'm writing my PC is "idleing" with default BIOS settings and XMP activated + DRAM voltage set to 1.36. I'll check if this helps in a couple hours after sleep. Normally the PC freezes in about ~2h idle.

 

I have few questions for you @mtavel. What is the exact model of your RAM? Your PC reboots, freezes??? Did it throw any Q-Code? Mine completely freezes, USB stop working, monitor goes black with "No displayPort signal" error and I have to push the switch on the PSU and boot the PC again... Also the Q-Code on the motherboard is 00 and the DRAM led is on with yellow color. For me is hard to believe that the RAM modules are defective. If it was the case I probably would have POST issues, stress tests like MemTest86 for 8h should crash...

 

One last thing. Do you have Asus AI Suite 3 installed or any Asus software?? I have it all running and what I've noticed in the event viewer after every crash there is an DipAwayMode.exe regarding I have disable it from the AI Suite software... Can you check your Event Viewer for  "UserPowerMode" in INFO level calling -> C:\Program Files (x86)\AI Suite III\DIP4\DipAwayMode\DipAwayMode.exe ???

 

If tonight test at DRAM v1.36 works fine tomorrow I'll test again with v1.35 and AI Suite uninstalled and DipAwayMode removed from Windows Task Scheduler. No matter if you disable it in the interface this task is still running on the system if you're logged in...

 

I'll keep posting new info. (Hope your new 5950x arrives fast...)

 

GN

[mtavel]

I have G.Skill F4-3600C16-16GTZNC

 

I saw some of the same hard hangs you saw, with the 00 Q-Code and I had to shut down the PC at the power supply to recover.  I got them when I was adding positive Curve Optimizer offset on the worst core trying to stabilize it, but I started getting these hard hangs sometimes.  That's when I gave up trying to 'fix' my 5950x that was obviously a binning mistake.  The only thing that EVER worked around my hangs/reboots was disabling C-states.

 

I did see a few event viewer entries related to UserModePowerService - none related to any AI Suite 3 or any other ASUS apps (which I do have installed).  They seem to be caused by some other unrelated (infrequent) app crashes - not related to my reboot issue. 

 

 

[Guest]

4 hours ago, mtavel said:

I have G.Skill F4-3600C16-16GTZNC

 

Oh this is starting to make a lot of sense now.....

 

The kit you should have boughten is the Same part number, but leave the C off the end. I'd be willing to bet, this set of memory is your problem. 

 

why did I miss this before...? Gah, I'm sorry.

 

Here's the kit you want for your system.

https://www.newegg.com/g-skill-16gb-288-pin-ddr4-sdram/p/N82E16820232858

[Guest]

On 2/6/2021 at 5:04 PM, suizman said:

RAM G.Skill Trident Z Neo DDR4 3600 PC4-28800 64GB 4x16GB CL16 (F4-3600C16Q-64GTZNC)

 

For your system, this set of Ram. Just like mtavel, drop the C off the part number.

 

https://www.newegg.com/g-skill-64gb-288-pin-ddr4-sdram/p/N82E16820232864

[mtavel]

13 hours ago, ShrimpBrime said:

leave the C off the end. I'd be willing to bet, this set of memory is your problem. 

What's the issue with the memory kit? 

 

I have been able to run this memory kit at D.O.C.P. 3600 with no system issues at all - just by disabling Global C-States on the CPU (I was having the issue with idle state reboots with my 5950x). 

 

I swapped out my 5950x for a 5600x and am running the same memory with the new CPU with no issues and all default values plus D.O.C.P. and PBO.  System is running like a dream now with the different CPU - not a single idle reboot or any other issue.  I'm getting my 5950x replaced now.

[mtavel]

Looks like the "C" at the end indicates a different timing, it is 16-19-19-39 instead of the 16-16-16-36 in the kit you linked to - It shouldn't be an issue as long as the timings are configured properly in BIOS.

 

I found this describing G. Skill memory model numbers:

Memory type:
F1 = DDR
F2 = DDR2
F3 = DDR3
F4 = DDR4
FA = For Mac

Frequency:
8500 = 1066MHz
10600/10666 = 1333MHz
12800 = 1600MHz
14900 = 1866MHz
17000 = 2133MHz
19200 = 2400MHz
Other 4-digit numbers represent the operating frequency with no changes.

CL Timing:
CL15 or C15 = CL timing (tCL) is 15 (i.e. CL 15-x-x-x)

Modules per kit:
S = Single module kit (x1)
D = Dual module kit (x2)
Q = Quad module kit (x4)
Q2 = Eight-module kit (x8)
T = Triple module kit (x3)
T2 = Six-module kit (x6)
H2 = Twelve-module kit (x12)
* If the model number does not have this letter to identify the number of modules for this kit, then this model number is for this specific module only, and is not the complete model number for your memory kit. For the FULL model number for your memory kit, please refer to the label on the memory module or on the package.

Total capacity:
2 / 4 / 6 / 8 / 12 / 16 / 24 / 32 / 64 / 128 / 192GB

Series:
TZ = Trident Z
TZRX = Trident Z RGB (for AMD)
TZR = Trident Z RGB
TZDC = Trident Z RGB DC (Double Capacity)
TRG / TRS = Tridnet Z Royal (Gold / Silver)
TRGU / TRSU = Trident Z Royal Collector Edition (Gold / Silver)
TZN = Trident Z Neo
SXF / SXK / SXW = Sniper X (Classic Camo-Green / Digital Camo-Black / Urban Camo-White)
FT = FORTIS (for AMD)
VR / VB / VK / VS / VG = Ripjaws V (Red / Blue / Black / Silver / Gunmetal Gray)
RR / RB / RK = Ripjaws 4 (Red / Blue / Black)
TX = TridentX
XL / XM / XH = RipjawsX
ZL / ZM / ZH = RipjawsZ
RL / RM / RH = Ripjaws
TD = Trident
PI = PI
AB / AO / AR = ARES
ECO = ECO
SR = Sniper
IS / ISL = Aegis
NQ / PK / HK = Performance
NT / NS = Value RAM
GRS / GRSL = Ripjaws SO-DIMM for laptop
SQ / SK = SO-DIMM for laptop

Additional Trident Z Color:
TZ = Classic Trident Z, Dual Black&Silver body w/ Red top-bar
TZSW = Silver body w/ White top-bar
TZSK = Silver body w/ Black top-bar
TZKW = Black body w/ White top-bar
TZKO = Black body w/ Orange top-bar
TZKY = Black body w/ Yellow top-bar

*If there is a “D” after the series name, then memory cooling fan is included, such as “8GTXD”
*You can find the capacity of each module by dividing the Total Capacity by Modules per kit. For example, F3-12800CL9Q-16GBXL = 16GB/4(Q) = 4GB per module, while F3-12800CL9D-16GBXL = 16GB/2(D) = 8GB per module.
*If there is a "A", "B" or "C" as the third letter of the series, such as "TZB" or "RKB", then it typically designates a different CL timing value when compared to the original model without the "B" in the third letter.

[Guest]

Ryzen chips really do well with better Ram quality kits IE 16-16-16-36 is a sweet performance spot for Ryzen across all 4 revisions 2 of which I've tested personally and the newer 2 Ryzen releases by close friends and family. 

 

That said, I've read a few threads running your exact kits of memory, all of which had different symptoms accompanying the issues.

 

Take that as you want. I'm not telling you to replace the memory, just showing which would have been the better purchase, I do feel the memory contributes to your issue.

[mtavel]

Thanks, and I appreciate your pointing out the timing difference.  When troubleshooting issues this fundamental (intermittent shutdown/reboot), all core components are not only suspect, but definitely could contribute and taking everyone's experience into account has been SUPER helpful.  

 

I very heavily suspected memory too, but started leaning away from it in my case when replacing the CPU and keeping everything else the same eliminated the issue.  It could be that the tolerances of this replacement CPU are compensating for some issue in the memory too... that's the nature of these kinds of complex systems with variable real-world performance between samples.

 

Thanks again

[suizman]

Hi, in Ryzen architecture memory plays a crucial role. It could be related with the IMC(memory controller) of 5950x. As @ShrimpBrime said there are a lot o people experiencing issues with this memory kit. But in my opinion the main issue resides inside XMP it self is Intel feature and the default timings for the profile are usually based on Intel CPU. So motherboard manufacturers should take this in account and configure better the safe defaults when you enable XMP(DOCP) profiles.

I don't thing there are that much faulty RAM modules and btw F4-3600C16-64GTZNC is not the only affected also F4-3600C16-32GTZN which is listed in ASUS QVL. In the 99% of the cases are bad BIOS settings. When I bumped the voltage to 1.36 the issue disappeared when I turn it back to 1.35 and manually set "Power Down mode"(This feature is AMD Ryzen only. Not Threadripper nor Intel ) to disable it just works!! No more random crashes on idle(never had them under heavy load even for 24/7 workloads).

This lead me to the next question WTF is doing the "Power Down mode" when its on Auto but default at v1.35?? Is it Enable, disabled?? No one knows because there is no F description nor manual from Asus for this 500€ motherboard... Maybe when you set it anything higher than 1.35 like 1.36 Auto set the Power Down to disabled and anything equal or lower than 1.35. No one knows the behavior of that feature.... It could be just that stupid and is causing a lot of unnecessary troubles or misleading people form some real issues with the CPU, etc..

 

And btw I've been monitoring the voltage of the DRAM and it randomly drops from 1.352 to 1.344 or to 1.352 from 1.360 exactly 0,008 in both cases.

 

Just check how other suffer the same issues. 4 sticks double rank are pushing to the limit the IMC

https://rog.asus.com/forum/showthread.php?122733-Gskill-ddr464gb-4-sticks-3600mhz-and-rog-crosshair-VIII-formula-problems

https://www.gskill.us/forum/forum/product-discussion/ddr4/166256-f4-3600c16q-64gtznc-64gb-3600mhz-problem-witn-ryzen-9-5950x

 

 

[mtavel]

2 hours ago, suizman said:

It could be related with the IMC(memory controller) of 5950x

Don't all Zen 3 CPU's share the same IMC?  After switching from my 5950x to a 5600x (to get me through the RMA wait) - with the same settings configured (BIOS defaults with D.O.C.P. 3600 and PBO enabled) - I have not had an idle reboot in 8 days of nearly 24 hour uptime per day (I did hibernate it a few nights). With my 5950x, I would have had maybe 4 reboots a day. 

 

One other note, my shutdown/reboots were largely associated with WHEA errors in Event Viewer identifying the CPU (specifically the APIC associated with Core 14) as the point of failure.

 

I'll definitely be watching for any reboots when the new 5950x arrives (they will be easy to spot since I have ZERO now). 

 

Whether the low power reboots are related to CCX core instability, IMC instability, Memory module instability, etc. leads to the difficulty getting root cause analysis and explains why settings that work for one person have little or no impact for someone else.