Think I have a faulty Ryzen 7 PRO 4750G CPU but I cannot return. Trying to decide which CPU and where to buy it for replacement for X570D4U-2L2T

[ScottN]

Long story short. I've been trying to narrow down instability in a server build I have. I've turned off overclocking in the BIOS, performed memtest, Prime95 for 24 hours and they pass just fine. But if I just let the server sit, idle, for about 7 days or so, it randomly reboots itself. No blue screen, no Windows Server memory dump, event log shows says "The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly". Just like someone pulled the power plug and re-plugged it and it boots back up. No, I haven't had any power failures either

So, I'm getting impatient a bit and I need to get this server shipped out because of the money invested, time spent, and the dust it has been collecting, I need to find another CPU to get this this stable and out the door.

The attached file is from STH review of the X570D4U-2L2T of supported CPUs. Problem is, I honestly don't know which one would be the best for a server configuration. I do not want to overclock, I just want speedy, reliable server build that will do database, file encoding and some other cloud services that are not too CPU intensive. Also, need this to be an APU as I do not need a GPU in the build. Also I have 64GB ECC memory installed and what I've read, I need a PRO CPU to utilize this.

Also, I'd like to buy it from a vendor that allows warranty returns as the 4750G I bought was from an independent seller on Amazon and AMD OEM warranty says any returns or warranty should be done from the seller.

Looking for advice on this. Thank you in advance.

 

[Wasmachineman_NL]

Why not get a 5750G?

 

[ScottN]

Why not get a 5750G?

Thanks for the response. A Ryzen 7 PRO 5750G? Looks to be pretty good. Not sure where I can buy one though right now

 

[CyklonDX]

this doesn't sound like CPU issue.
Would you happen to run your system off nvme?
Care to record temps on all parts; like VRM, CPU, Mem, GPU, Disks (i.e. to a file/db or even use telegraf with influxdb on another system?)

Are you sure your psu is good; and power cables are firmly attached?

 

[ScottN]

this doesn't sound like CPU issue.
Would you happen to run your system off nvme?
Care to record temps on all parts; like VRM, CPU, Mem, GPU, Disks (i.e. to a file/db or even use telegraf with influxdb on another system?)

Are you sure your psu is good; and power cables are firmly attached?

Hello, thanks for your response.

I am on NVME in mirrored configuration for boot OS. The server is completely idle in a temperature controlled space and without warning reboots after about 7 days of just idle, nothing happening, no services running, so temps are not increasing from ambient values. I've ran Prime95 for 24 hours stressing CPU and RAM and no reboots as well. I disabled overclocking option in BIOS that I found because with it on, one of the cores was getting a computational error in Prime95. With it off, no errors. But maybe that core is still something that is causing the instability here.

I've also done long duration file copies with robocopy in Windows under multi-threaded file copy switches maxing out file transfers on the RAID configuration. I can try to do the same on NVME for a day possibly. But I have no controller errors in event log. Just normal event logs then after the reboot the normal errors of "The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly "

So yes, I believe it is CPU from testing I've done.

 

[CyklonDX]

try running it from normal ssd/hdd for a while - see if you get crash again.
I've had same issue once with bad nvme's.

 

[ScottN]

try running it from normal ssd/hdd for a while - see if you get crash again.
I've had same issue once with bad nvme's.

I could take out the NVMe drives, boot to a Linux live disk and let it run idle. Frustrating I have to wait 7 days for this to crop up again.

 

[Allan74]

You don't need an APU. The VGA output will serve up video output via built-into the motherboard. The HDMI is there to output from an APU.

Next, APU memory controllers are weak. Try run 2 sticks of memory for a short time (if you are running 4) and run lower Mem clocks with 4 sticks.

Lastly, Use the Motherboard's VGA output and disable the CPU's video output.

Ps. You don't need a PRO Chip to run ECC UDIMMS. All Ryzen 2 and newer are confirmed to officially support ECC UDIMMS (not registered).

 

[SilverZelos]

I have the X5704DU and was experiencing similar issues. It turned out the X570 chip was overheating. Maybe that is the issue?