power issue for threadripper 1920x

[Diver1]

So I have an odd issue that I haven't been able to find anywhere. My computer will work just fine when I power it off and it is a soft power off. When I unplug the computer or turn the psu switch off and try to turn the computer back on is where I have issues. When I turn it on after unplugging or flipping the psu switch the computer will not boot and will have the cpu red led light on and flickers the ram led light. The error codes cycle through different codes and sits there and does not boot. The only solution that I have come up with is to repeatedly hit the reset button on the mobo and eventually will produce the 0d code and power off a few times then will boot just fine and I just have to reset all of my bios settings. Is it the psu or what could be causing this?

 

cpu: amd ryzen threadripper 1920x not over clocked

mobo: gigabyte auroras gaming 7

ram: 32 gb trident z 3200 

gpu: asus strix 1080 ti oc

psu: evga 750 g2 80+ gold

[Tabs]

Just now, Diver1 said:

So I have an odd issue that I haven't been able to find anywhere. My computer will work just fine when I power it off and it is a soft power off. When I unplug the computer or turn the psu switch off and try to turn the computer back on is where I have issues. When I turn it on after unplugging or flipping the psu switch the computer will not boot and will have the cpu red led light on and flickers the ram led light. The error codes cycle through different codes and sits there and does not boot. The only solution that I have come up with is to repeatedly hit the reset button on the mobo and eventually will produce the 0d code and power off a few times then will boot just fine and I just have to reset all of my bios settings. Is it the psu or what could be causing this?

 

cpu: amd ryzen threadripper 1920x not over clocked

mobo: gigabyte auroras gaming 7

ram: 32 gb trident z 3200 

gpu: asus strix 1080 ti oc

psu: evga 750 g2 80+ gold

I would ensure you have the latest firmware installed, since it may be a memory compatibility issue. DDR4 goes through something called training whenever it cold boots, and if your memory is not quite supported by the board at the full XMP settings, it may be failing to train.

 

You can sometimes offset this by increasing the training voltage, but updating your bios/firmware first is the best step.

[Diver1]

22 minutes ago, Tabs said:

I would ensure you have the latest firmware installed, since it may be a memory compatibility issue. DDR4 goes through something called training whenever it cold boots, and if your memory is not quite supported by the board at the full XMP settings, it may be failing to train.

 

You can sometimes offset this by increasing the training voltage, but updating your bios/firmware first is the best step.

Not running the brand new version but running a newer version that helps with compatibility. The cpu led light is lit most of the time though. Wouldn't a memory issue have the ram led lit. The ram led light only flickers infrequently. Technically the ram sticks that I am using are not on the list of validated sticks but others have used them without issue.

[Tabs]

Just now, Diver1 said:

Not running the brand new version but running a newer version that helps with compatibility. The cpu led light is lit most of the time though. Wouldn't a memory issue haver the ram led lit. The ram led light only flickers infrequently. Technically the ram sticks that I am using are not on the list of validated sticks but others have used them without issue.

I get that, but remember that the memory controller is part of the CPU. If the memory controller isn't able to get stable training done, it could show up as a CPU issue - even if it's the ram that's mainly at fault.

 

I'd still recommend trying it - you should still be able to downgrade back if it gives you a worse experience. 

[Diver1]

23 minutes ago, Tabs said:

I would ensure you have the latest firmware installed, since it may be a memory compatibility issue. DDR4 goes through something called training whenever it cold boots, and if your memory is not quite supported by the board at the full XMP settings, it may be failing to train.

 

You can sometimes offset this by increasing the training voltage, but updating your bios/firmware first is the best step.

 

1 minute ago, Tabs said:

I get that, but remember that the memory controller is part of the CPU. If the memory controller isn't able to get stable training done, it could show up as a CPU issue - even if it's the ram that's mainly at fault.

 

I'd still recommend trying it - you should still be able to downgrade back if it gives you a worse experience. 

Do you have a recommendation for what to set the training voltage at? Thanks for the help by the way.

[Tabs]

Just now, Diver1 said:

 

Do you have a recommendation for what to set the training voltage at? Thanks for the help by the way.

I'd recommend starting at 1.25 and going no higher than 1.35. The training voltage is somewhat tied with the voltage the modules themselves run at, and I'm assuming for 3200 your kit needs 1.35.

 

Note that this is based on my own experience, I have found zero conclusive evidence on whitepapers etc as to what safe voltages are for training. If you have the ability to set an offset instead of a static value, that is much more useful. I'd start with +30mv and go up from there, to no higher than +100mv. That means you'll be increasing from the baseline your motherboard already sets, rather than the arbitrary value I had to guestimate for mine.

[Diver1]

1 minute ago, Tabs said:

I'd recommend starting at 1.25 and going no higher than 1.35. The training voltage is somewhat tied with the voltage the modules themselves run at, and I'm assuming for 3200 your kit needs 1.35.

 

Note that this is based on my own experience, I have found zero conclusive evidence on whitepapers etc as to what safe voltages are for training. If you have the ability to set an offset instead of a static value, that is much more useful. I'd start with +30mv and go up from there, to no higher than +100mv. That means you'll be increasing from the baseline your motherboard already sets, rather than the arbitrary value I had to guestimate for mine.

I'll give it a go thank you.