Is a Seasonic Focus GX 850W good enough for my system?

[tabascosauz]

@Glaceon disable Low Current Idle for your power supply. It's a setting in BIOS that should be specific to Ryzen. Sometimes it can be related to idle reboots and C-state woes. Something to do with the super low idle voltages on Matisse CPUs. The safer option should be called Normal Current Idle or something. Should be all in the same dropdown menu on both Asus and Gigabyte.

I haven't had these issues with either my Focus Plat or Focus SGX, but everyone's PC is different.

It's a scratch fix sorta along the same lines as C-states. With both, as long as you verify with some benchmarking, you shouldn't be losing any performance - only maybe more idle power consumption. Which would probably be a deal breaker on a laptop, not so much a desktop.

Lastly make sure you check event viewer and verify the actual WHEA error causing the reboot.

 

[Glaceon]

Typical Current Idle/Auto doesn't fix it.

What do you mean by WHEA?

 

[tabascosauz]

Typical Current Idle/Auto doesn't fix it.

What do you mean by WHEA?

Go to your event viewer and look through recent logs for Warnings and Errors. Specifically, WHEA-Logger labelled events in relation to both, and Kernel-Power events for Errors.

Generally with a hardware error-induced reboot, you'll find a Warning and an Error occurring with the same timestamp. The Error details should tell you what exactly went wrong, as in the screenshot above. Lots of different types of errors can end up here - WHEA can log errors caused by an unstable CPU overclock, the infamous CPU cache hierarchy error, Infinity Fabric errors, etc. just to mention a few.

Above is an example of an Infinity Fabric error that I was struggling with for a long while, running my memory at 3733. The memory was completely stable, but I couldn't quite find a set of voltage settings that the memory controller liked at 1866MHz, evidenced by it being a CPU Bus/Interconnect Error. I made some changes to my SOC Voltage, VDDP and VDDG and it's finally gone away. For what it's worth, I had very similar symptoms to what you're reporting, with occasional idle reboots when I walked away from my computer.

I don't know if you use hibernate at all - I completely disabled hibernate functions and quick startup (which also uses hibernate to give you the impression that your computer boots super fast when in reality it's just resuming) a while back.

 

[EarthDog]

@EarthDog
should have kept reading a bit more:
... some boards (at least on the Gb i had) dont delete properly and you have left overs.
you might have to use dos to update, with flag to delete (previous) bios before flashing update (even if its the same version)...

i had multiple ryzen rigs build and was able to exclude any other hw, and after spending some time,
found others have issues too (not just me or Gb boards), but i knew the board was fine,
as all tests/benches from memtest to prime/occt ran fine, same for light/med/full load no matter if gaming,
only happened under idle/low load (browsing/office).

having flashed the bios under dos, setting flag to delete bios/update both chips fixed it.
runs now for almost +3 month without sudden reboot, before avg of 1 week, max 3.

Thats fine and all, but... the bios just starts the car. Once the motor turns over (POST), it runs on the OS(?).

Anyway..a bit OT..

 

[Glaceon]

View attachment 174287

Go to your event viewer and look through recent logs for Warnings and Errors. Specifically, WHEA-Logger labelled events in relation to both, and Kernel-Power events for Errors.

Generally with a hardware error-induced reboot, you'll find a Warning and an Error occurring with the same timestamp. The Error details should tell you what exactly went wrong, as in the screenshot above. Lots of different types of errors can end up here - WHEA can log errors caused by an unstable CPU overclock, the infamous CPU cache hierarchy error, Infinity Fabric errors, etc. just to mention a few.

Above is an example of an Infinity Fabric error that I was struggling with for a long while, running my memory at 3733. The memory was completely stable, but I couldn't quite find a set of voltage settings that the memory controller liked at 1866MHz, evidenced by it being a CPU Bus/Interconnect Error. I made some changes to my SOC Voltage, VDDP and VDDG and it's finally gone away. For what it's worth, I had very similar symptoms to what you're reporting, with occasional idle reboots when I walked away from my computer.

I don't know if you use hibernate at all - I completely disabled hibernate functions and quick startup (which also uses hibernate to give you the impression that your computer boots super fast when in reality it's just resuming) a while back.

This is the only WHEA thing I got, filtered results to only show me WHEA events:

There was a shutdown 9 minutes before this event was logged.
There are no corresponding WHEA events to every other shutdown.

No, I do not use hibernate.

 

[tabascosauz]

This is the only WHEA thing I got, filtered results to only show me WHEA events:

There was a shutdown 9 minutes before this event was logged.
There are no corresponding WHEA events to every other shutdown.

No, I do not use hibernate.

If your first post accurately reflects your CPU setup at the moment, methinks you need more Vcore. Especially if that per-CCX OC is done from Clocktuner; I have full confidence in its abilities to determine an ideal OC, but much less faith in its ability to exhaustively determine the complete stability of that overclock.

Are your Ripjaws CJR (16-19-19-39) or B-die (16-16-16-36)? In any case, make sure your SOC Voltage is set at 1.1V, do it manually if you need to. Asus should be pretty good with the minor voltages, so other than that you should be okay on that front.

 

[Glaceon]

This happens in stock CPU settings too.

My RAMs are 16-19-19-39.

Kinda scared to change anything RAM related after the Gigabyte incident.

Ryzen Master reports SoC voltage is at 1.1

 

[tabascosauz]

This happens in stock CPU settings too.

My RAMs are 16-19-19-39.

Kinda scared to change anything RAM related after the Gigabyte incident.

Ryzen Master reports SoC voltage is at 1.1

The Gigabyte incident?

Monitor your CPU core and SOC voltages using HWInfo. Ryzen Master wack af for monitoring and often just gives you static values you set in BIOS.

If you're running the stock XMP 3600 16-19-19, go up to 1.38V-1.4V on the DRAM to see if that eliminates the problem. Will be a safe daily voltage for CJR. The problem with Asus boards is that they don't provide DRAM voltage monitoring in Windows, you're completely in the dark about what your actual DRAM voltage is as you use the computer. Setting to actual voltage on each board will be different.

I can't remember if you've already gotten the Seasonic, and how old your CX650 is (a PSU of that tier I would be fine with indefinitely in an office PC, but would be replacing it after 3 years tops in my main PC), but if you're on the new PSU already and all else fails, it might be time to reach out to AMD to see about a CPU RMA. The cache hierarchy issue comes to mind as one of the potential issues with Ryzen that is sometimes best remedied with a CPU RMA.

 

[Glaceon]

The Gigabyte incident?

Monitor your CPU core and SOC voltages using HWInfo. Ryzen Master wack af for monitoring and often just gives you static values you set in BIOS.

If you're running the stock XMP 3600 16-19-19, go up to 1.38V-1.4V on the DRAM to see if that eliminates the problem. Will be a safe daily voltage for CJR. The problem with Asus boards is that they don't provide DRAM voltage monitoring in Windows, you're completely in the dark about what your actual DRAM voltage is as you use the computer. Setting to actual voltage on each board will be different.

I can't remember if you've already gotten the Seasonic, and how old your CX650 is (a PSU of that tier I would be fine with indefinitely in an office PC, but would be replacing it after 3 years tops in my main PC), but if you're on the new PSU already and all else fails, it might be time to reach out to AMD to see about a CPU RMA. The cache hierarchy issue comes to mind as one of the potential issues with Ryzen that is sometimes best remedied with a CPU RMA.

Yes, tightened timings too much on Gigabyte board and instead of having it reset, the BIOS got corrupted.

HWiNFO says the SoC voltage is 1.080V. I'm still not sure how changing RAM settings would fix this as I tried different modules and it persists.

I didn't get the Seasonic, still on this PSU. This PSU was bought brand new back in December 2019, but I tried with another CX 650 that was bought only a few months ago new and the issue persists.

CPU RMA out of the question as the warranty is void, not gonna get into details about that.