My new build is Power-Cycling at first boot, but then seems to work fine once it gets going...

[deephungermusic]

My new build is power-cycling at first boot, as soon as the logo appears and it says to press F2 for bios the system immediately powers off and tries again. It loops anywhere from 2-8 times, but DOES eventually boot up. Once on, it seems perfect, everything works, all memory is seen, CPU/GPU operating correctly all day long and under full load. And... once it has been running I can do a full shutdown and turn it back on right away, and it does not power cycle. But if I leave it off, say overnight, it does the power cycling thing again when I first turn it on.

I have tried re-seating the RAM, CPU, and GPU... with no change.

Where should I begin, given that it seems to work most of the time? Does this sound like a Power Supply issue, a faulty motherboard or maybe a bios issue? Would a new battery on the MB help?

Thanks for your time!

My setup...
PSU: EVGA 1000 GQ
MB: ASROCK x399 Taichi
CPU: AMD threadripper 1950x
RAM: Gskillz 4X16
SSD1: Samsung 960 Pro M.2

 

[velocityg4]

I tried explaining this as well as I could. It could be any of the above. I’d normally say update the BIOS first. But it’s risky on computer with dodgy POST behavior. Unless you can establish a state where it POSTs reliably. I wouldn’t risk it.

All I could say is to start at the most basic level. Perform a CMOS reset. Only have the CPU (plus heatsink and CPU fan), PSU, Motherboard, GPU, one RAM module installed and monitor. Nothing else, no other fans or cables. Besides the cables absolutely necessary for this basic task. See if it will POST beep and stay on a minute or two.
1. If it fails. Try a different RAM module in another slot.
1A. If it then works. Try that good RAM module in the first slot you used.
1A1. If it then fails. It sounds like a faulty motherboard or at least bad RAM slot. Retest later when cold to see if results are consistent.
1B. If it still fails. Odds are either a motherboard, PSU, GPU or CPU problem. You’ll need a known good PSU and GPU to test with. In order to rule your current GPU and PSU as the culprits.
2. If it works add a keyboard and mouse. Then try again.

If consistently reliable on POST. Plug in a thumb drive and update to the latest BIOS

Next I'd add the SSD. Making the minimum changes necessary to the BIOS settings to boot. Then boot off it.

From there on I'd add devices one at a time and try again. With any luck you'll find the suspect. Since it only seems to happen when your PC is cold. This may take multiple attempts over time until you zero in on the culprit.

I assume by full load on the CPU and GPU. You did stress tests and left them running overnight.

Did you run memtest86 on your memory?

 

[deephungermusic]

First, thank you for the help and the strategy for moving forward!

(I hadn't heard about the memtest86 so I went ahead and ran that and got no errors back.)

I had a friend tell me that it could be software related, and I had to spend some time double checking the UEFI boot situation. It did turn out that Windows had setup my 960Pro nvme m.2 as a Master Boot Record and not a GUID partition, which could be confusing the boot if the Legacy bios option was not enabled, as MBR does not allow UEFI. So I did get that straightened out (gotta love mbr2gpt.exe which allows you to switch your drive to GUID without erasing content and starting over). This sped up my boot time significantly, being able to enable "UEFI only" in the CMS section of the bios. But still had problems power cycling when machine was cold.

So I began the process for diagnosing the hardware, and I think I found the culprit in one of the RAM sticks. When singled out, 3 of the sticks would POST but 1 would just give an error code. I put in two of the good sticks, left it over night, and it booted fine in the morning. I will be returning the four pack of RAM sticks and getting a replacement batch soon hopefully.

 

[velocityg4]

Hopefully, that clears it up.