Question Can PBO make Zen3 CPUs crash / BSOD? Is it enabled by default?

[VirtualLarry]

5900X on an Asus ROG STRIX B450-F Gaming. Sill Win10.

 

[naukkis]

Yes it can. Mostly it's up to MB VRM. I have seen reviews that imply that particular MB has extremely poor quality VRM( from test The Stilt did for finnish hw-site io-tech) IO so it might be better to run it at defaults instead of PBO.

 

[VirtualLarry]

I currently have all of the PBO stuff set to "Auto". Maybe I should change it to "Disabled" then.

 

[DrMrLordX]

PBO is not active by default. It's not hard to check anyway, just load up something like Prime95 and monitor PPT. If it's 142W then you're running bone stock.

 

[UsandThem]

PBO is not active by default. It's not hard to check anyway, just load up something like Prime95 and monitor PPT. If it's 142W then you're running bone stock.

I believe on my Asus ROG Strix X470 and B550 motherboards, it was enabled by default.

I'm pretty sure I had to manually go in and disable it.

 

[PJVol]

I currently have all of the PBO stuff set to "Auto". Maybe I should change it to "Disabled" then.

PBO auto = disabled. And even if it is enabled in some way, it basically do nothing, other than expand power/current limits for your cpu, assuming other settings stay untouched (boost, curve, scalar and thermal limit).

The reboots itself are actually two separate issues. And the reasons for it depends on whether it happens at idle or under some load (i.e. games or rendering).

IMO, most probable are:

1. Defective sample. If reboots happen with the default CPU settings - RMA it ASAP.

2. CPB and c-state boost reboots, that occur with PBO tuned - might be fw bugs or, as previously mentioned, mb voltage supply issues. Usually happen when idle.

3. Reboots under load are the result of messed OC or tuning.

 

[pakotlar]

PBO auto = disabled. And even if it is enabled in some way, it basically do nothing, other than expand power/current limits for your cpu, assuming other settings stay untouched (boost, curve, scalar and thermal limit).

The reboots itself are actually two separate issues. And the reasons for it depends on whether it happens at idle or under some load (i.e. games or rendering).

IMO, most probable are:

1. Defective sample. If reboots happen with the default CPU settings - RMA it ASAP.

2. CPB and c-state boost reboots, that occur with PBO tuned - might be fw bugs or, as previously mentioned, mb voltage supply issues. Usually happen when idle.

3. Reboots under load are the result of messed OC or tuning.

reboots under load can also be the result of messed up oc, especially cache memory hierarchy.

 

[PJVol]

reboots under load can also be the result of messed up oc, especially cache memory hierarchy.

Sorry, don't get the meaning of this comment.

 

[DrMrLordX]

I believe on my Asus ROG Strix X470 and B550 motherboards, it was enabled by default.

I'm pretty sure I had to manually go in and disable it.

Strange. PBO isn't supposed to be enabled by default on any board, not even the enthusiast ones. It certainly wasn't enabled on my X570 Aorus Master.

 

[pakotlar]

Sorry, don't get the meaning of this comment.

>Reboots under load are the result of messed OC or tuning.

Reboots at idle aremost commonly caused by instability at low load due to improper application of PBO2, which is overclocking.

 

[PJVol]

I thought tech forum is not the right place to discuss marketing terms and definitions, and how PBO2 tuning should be considered. Technically, with ryzens, OC is "oldschool" manual overclocking, i.e. when cpu explicitly switched to OC-mode, while tuning PBO2 settings implies automatic clock control and not necessarily results in higher operating frequency.
Anyway, both cases are concidered by AMD as "operating outside of specs or outside of factoty settings".

As for the reboot at idle, I don't think it has much to do with PBO2 tuning, based on my experience.

 

[Accord99]

As for the reboot at idle, I don't think it has much to do with PBO2 tuning, based on my experience.

It can if you also use the curve optimizer to set a negative voltage offset, which would allow for a higher clock at load but potentially result in too low of a voltage at idle or low load situations.

 

[DrMrLordX]

It can if you also use the curve optimizer to set a negative voltage offset, which would allow for a higher clock at load but potentially result in too low of a voltage at idle or low load situations.

How low of a voltage would you have to set for that to happen? When I manually set low voltage offsets on my Matisse (3900X) all it would do is downclock like crazy, or "false clock" (read a higher clockspeed but rapidly downclock to retain stability, killing performance).

 

[pakotlar]

How low of a voltage would you have to set for that to happen? When I manually set low voltage offsets on my Matisse (3900X) all it would do is downclock like crazy, or "false clock" (read a higher clockspeed but rapidly downclock to retain stability, killing performance).

There's no hard threshold; it depends on the quality of the chip. PBO/PBO2 are a major cause of reboot at idle on Ryzen chips, rather than at load. Ryzen stability testing is whether your core can survive sitting at idle, when voltage drops to its lowest point, without rebooting.

 

[PJVol]

It can if you also use the curve optimizer to set a negative voltage offset, which would allow for a higher clock at load but potentially result in too low of a voltage at idle or low load situations.

PBO/PBO2 are a major cause of reboot at idle on Ryzen chips

Out of curiosity, are these conclusions purely theoretical or based on your personal experience?

 

[fleshconsumed]

B450-F has really crappy VRMs/VRM Heatsinks that are not built to handle 12/16 cores. 5900X will eat 140W of power, that is too much for B450-F VRMs without active cooling. It is entirely possible that is the source of instability.

I would reduce PPT to 100W and see if that improves stability.

 

[Accord99]

How low of a voltage would you have to set for that to happen? When I manually set low voltage offsets on my Matisse (3900X) all it would do is downclock like crazy, or "false clock" (read a higher clockspeed but rapidly downclock to retain stability, killing performance).

It depends on the particular CPU though in general for the 5000 series, the 5900x and lower appear to have quite a bit of headroom while the 5950x is already optimized by AMD (probably to make it shine even more). Here's a video about using it to overclock:

Out of curiosity, are these conclusions purely theoretical or based on your personal experience?

I could only get -9 on my 5950x. Even -10 would cause a random restart every few days or weeks when the system was idle.

 

[PJVol]

I could only get -9 on my 5950x. Even -10 would cause a random restart every few days or weeks when the system was idle.

Well, if you mean -9 for the best core, then you've got a pretty good sample. My 5600x is not bad at all (to say the least), but it barely handle less than -5 on its best core. 2nd best is -10.
Not considering the fact that the 5950x silicon itself is quite different from the rest of the 5000's family.

But back to idle restarts. The first ones have occured to me long before people have heard the term Curve Optimizer, even before 5000 launch.
To be precise, first idle reboots were noticed when first vermeer-ready bioses starts to appear. Me personally have experienced it first with taichi x370 + 3600x, then with b550 extreme + 3600x and finally with a current config, until it was fixed in agesa 1.1.0.0 based bios.
I still have a png with a words "still alive" somewhere in a Pictures folder, which I used to put on the screen when leaving the pc on for a long time (power saving was always off) to let me know, if PC has restarted, when I got back to it

That is basically my experience with a sudden idle reboots. Haven't seen it since then, no matter what I do with cpu settings in bios.

PS: have to agree with a poster above, that a poor mb VRM design might be the culprit in the OP's particular case.

 

[fleshconsumed]

Well, if you mean -9 for the best core, then you've got a pretty good sample. My 5600x is not bad at all (to say the least), but it barely handle less than -5 on its best core. 2nd best is -10.
Not considering the fact that the 5950x silicon itself is quite different from the rest of the 5000's family.

But back to idle restarts. The first ones have occured to me long before people have heard the term Curve Optimizer, even before 5000 launch.
To be precise, first idle reboots were noticed when first vermeer-ready bioses starts to appear. Me personally have experienced it first with taichi x370 + 3600x, then with b550 extreme + 3600x and finally with a current config, until it was fixed in agesa 1.1.0.0 based bios.
I still have a png with a words "still alive" somewhere in a Pictures folder, which I used to put on the screen when leaving the pc on for a long time (power saving was always off) to let me know, if PC has restarted, when I got back to it

That is basically my experience with a sudden idle reboots. Haven't seen it since then, no matter what I do with cpu settings in bios.

PS: have to agree with a poster above, that a poor mb VRM design might be the culprit in the OP's particular case.

I've had horrible luck with AM4 ASrock motherboards. They were great when I had their Z87/Z97 motherboards, but both of my Asrock x370's had issues, and so did B450's. Asus B450/X470 don't have VRM's built to handle 12/16 cores but at least they're stable. I'm waiting for some kind of sale on gigabyte/msi b550m motherboard to replace my last asrock b450 motherboard.

 

[scannall]

I've had horrible luck with AM4 ASrock motherboards. They were great when I had their Z87/Z97 motherboards, but both of my Asrock x370's had issues, and so did B450's. Asus B450/X470 don't have VRM's built to handle 12/16 cores but at least they're stable. I'm waiting for some kind of sale on gigabyte/msi b550m motherboard to replace my last asrock b450 motherboard.

I had 2 Asrock motherboards, x470. Both died a horrible death and I got no resolution from Asrock, so they are on my don't buy or recommend list.

 

[Markfw]

I had 2 Asrock motherboards, x470. Both died a horrible death and I got no resolution from Asrock, so they are on my don't buy or recommend list.

I had a X399 Taichi board go, but all my EPYC are Asrock.

 

[eek2121]

Strange. PBO isn't supposed to be enabled by default on any board, not even the enthusiast ones. It certainly wasn't enabled on my X570 Aorus Master.

I currently have all of the PBO stuff set to "Auto". Maybe I should change it to "Disabled" then.

PBO is always set to “auto”, which means “disabled”.

If you are getting BSODs, select Load Optimized Defaults”, save, exit, reboot, and test again. If all works well, enable XMP or whatever your motherboard calls it and test again. If things work well at that point, you are good to go.

With my Zen 2 chip things worked fine. With Zen 3, I get BSODs if I am too aggressive with PBO. I keep it disabled for the most part. My board is okayish (aorus x570 elite wifi)

 

[PJVol]

I'm waiting for some kind of sale on gigabyte/msi b550m motherboard to replace my last asrock b450 motherboard.

I'd choose msi from these two. Have a cheap b450 at the office PC and tested it with 2600, 3600X, 5600X and currently running 5700G. Nothing to complain, except not so great RAM OC potential. Max I've managed to run stable is 4400CL17, but for its price its far above what was declared by specs.

And you're right, on my current board despite having way more robust VRM module (compared to office MSI B450) 5600X running less stable with high FCLK (>1900), whereas on B450 it is fine upto 2033