Hi all, I have stumbled on a fix that works for me for the most part for getting 5900X stable and avoiding WHEA bluescreen (Bus/Interconnection) errors.
However, I have questions as it involves the PBO Curve Optimizer which is extremely lacking in details, and I don't know if there might be a better fix until (crossing fingers) an updated AGESA gets released.
PROBLEM: Like others, I upgraded to 5900X and immediately got WHEA_UNRECOVERABLE_ERROR bluescreens that would crash windows within seconds or up to a minute of start up. Disabling PBO in BIOS made the system stable, but now you've lowered performance and turned the 5900X into a $100 processor.
FIX: I went into the PBO curve optimizer in the BIOS and set the voltage magnitude adjustment for all cores to a positive value offset. This has made the system much more stable and allowed me to enable PBO, memory 3600Mhz XMP, and 1800Mhz Infinity Fabric (see my specs below). I set this value to +6, and ran heavy benchmarking and lower usage stuff like web browser that still puts PBO into effect fairly often. It ran for 12 hours, but bluescreen WHEA crashed after running a GPU benchmark with high and erratic single thread CPU usage for a couple of minutes. I'm now testing it out with a +10 setting. Still, it's pretty much a night and day difference as far as stability goes, if not yet perfect.
QUESTION #1: I am not a heavy overclocker and not an expert voltage tweaker, so don't know if what I've done is OK for long term usage. Documentation on the PBO curve optimizer is especially lacking. For example, the curve optimizer accepts a "magnitude" positive/negative value (integer between 0-9999), but you can't find anywhere explaining what unit this "magnitude" value is. I found an overclocker who set the value to -15, and it caused voltage on PBO usage to drop by -27 mV. So it might be related to mV?
QUESTION #2: I also don't know if the stability I get from this setting narrows what the problem is, and if someone can recommend a different voltage/BIOS setting I can use that will provide a better fix. I like the curve optimizer because it uses an offset value, versus one of the voltage settings that you enter in an absolute value, which I don't know what I should type in. In the overclocker article I linked to, he mentions "With Curve Optimizer set to -25 it was necessary to add a voltage offset of +50 mV in order to keep the system stable." I could not clearly find in BIOS where I can enter in a voltage offset. Everywhere it seems to want absolute voltage values, not relative offsets. It's possible I've just missed it and it's right under my nose.
Manually setting PBO Curve to +10 on all cores... genius. good idea. I think working it up slowly if needed at high IF/fclk is a good idea. Because the differences in CCDs on package.
just for giggles, in case something is borrowing current from the PCI bus.. try giving a touch of volts manually to the PCIE bus. but not much , check your mobo, just a touch.