AMD Releases Second Official Statement Regarding Ryzen 7000X3D Issues

[mama]

I'm taking from this discussion that I probably should not update my Gigabyte bios just yet. I have been monitoring my VCORE SOC with HWInfo64 since I heard about the issue and it seems stable on F6h BIOS. Jayz Two Cents suggests that software monitoring doesn't work. How true is that? Should I move on to the new BIOS? Screenshot attached for those who care.

 

[AusWolf]

I'm taking from this discussion that I probably should not update my Gigabyte bios just yet. I have been monitoring my VCORE SOC with HWInfo64 since I heard about the issue and it seems stable on F6h BIOS. Jayz Two Cents suggests that software monitoring doesn't work. How true is that? Should I move on to the new BIOS? Screenshot attached for those who care.

As long as everything is fine, and VSoC is below 1.3 V, I'd leave it as it is. My new MSi BIOS, although made "Memory Context Restore" finally work, added 5 W to my idle power consumption and heat for no reason.

 

[Mussels]

Not really putting my worries to rest here dude.

Which worries?
If you have an AM5 Asus board, i'd worry.
And you do. I'd keep PBO disabled, enable EXPO but keep RAM clocks 'low' (5600 or lower) and manually lock SoC to something safe, like 1.20v


You could get answers and a BIOS fix in a week, your board may not even be affected - but keep the hardware alive by running the DRAM and SoC lower until you know for sure.

I'm taking from this discussion that I probably should not update my Gigabyte bios just yet. I have been monitoring my VCORE SOC with HWInfo64 since I heard about the issue and it seems stable on F6h BIOS. Jayz Two Cents suggests that software monitoring doesn't work. How true is that? Should I move on to the new BIOS? Screenshot attached for those who care.

It's been confirmed that the software readings are not accurate on asus boards at least, when they failed they showed 1.10v to the SoC while a hardware probe read 1.80v

The newer BIOSes have limits that should prevent the over-voltage happening, even if the software reading is bugged out

 

[mikehcl]

Which worries?
If you have an AM5 Asus board, i'd worry.
And you do. I'd keep PBO disabled, enable EXPO but keep RAM clocks 'low' (5600 or lower) and manually lock SoC to something safe, like 1.20v


You could get answers and a BIOS fix in a week, your board may not even be affected - but keep the hardware alive by running the DRAM and SoC lower until you know for sure.


It's been confirmed that the software readings are not accurate on asus boards at least, when they failed they showed 1.10v to the SoC while a hardware probe read 1.80v

The newer BIOSes have limits that should prevent the over-voltage happening, even if the software reading is bugged out

I don't think only asus has inaccurate readings, gigabyte too for sure!

Gigabyte has released the F10d bios

 

[trparky]

I've loaded the latest BIOS on my board, version F5c and yes, I encountered some of those bugs that Buildzoid encountered in my BIOS as well. It's confusing as hell when I set my SoC voltage in one area of the BIOS and have it not appear to have been set in another area of the BIOS. These bugs in BIOS are frustrating to say the least.

 

[Mussels]

I don't think only asus has inaccurate readings, gigabyte too for sure!

Gigabyte has released the F10d bios

Anyone on AM5 should be avoiding automatic voltages and playing it safe for now.

Asus Gigabyte and MSI have all had problems with high end motherboards, Now we wait and see if its an AMD issue, the board makers, or a hardware problem (specific VRM's they all had in common with a fault, etc)

I've loaded the latest BIOS on my board, version F5c and yes, I encountered some of those bugs that Buildzoid encountered in my BIOS as well. It's confusing as hell when I set my SoC voltage in one area of the BIOS and have it not appear to have been set in another area of the BIOS. These bugs in BIOS are frustrating to say the least.

I've seen this on AM4 in the past too, you can set voltages in the "AMD overclocking" section of the BIOSes, but the 'vendor' sections (asus, giga etc) always has priority and their setting is chosen

LLC settings could also be involved here, and might be a way to keep things safer by avoiding automatic LLC and setting it to the droopiest values, if it's stable for you

 

[Klemc]

And, about using two CPU 8pins when available (mobo and PSU), can it make the CPU use more power ?

 

[Klemc]

 

[trparky]

OK, I was playing in BIOS again today to try and get the SoC voltage even lower. For some time, I've been running it at 1.2 volts, but I was hoping to get it to be even lower; say like 1.15 or even 1.1 volts. However, it appears that it doesn't like that value at all. I can set it in BIOS but when I come back up into Windows, the SoC voltage according to HWInfo is set at 1.2 volts. I can't get it to be any lower than that with EXPO enabled.

 

[Klemc]

ASUS says :

//

So, is ASUS hardware crappy or only their BIOS ??

 

[Mussels]

And, about using two CPU 8pins when available (mobo and PSU), can it make the CPU use more power ?

no

So, is ASUS hardware crappy or only their BIOS ??

The hardware, the BIOS, their response, and their RMA process are all implicated as being utter trash at this point, between JayZ and GN's videos on the issues.

OK, I was playing in BIOS again today to try and get the SoC voltage even lower. For some time, I've been running it at 1.2 volts, but I was hoping to get it to be even lower; say like 1.15 or even 1.1 volts. However, it appears that it doesn't like that value at all. I can set it in BIOS but when I come back up into Windows, the SoC voltage according to HWInfo is set at 1.2 volts. I can't get it to be any lower than that with EXPO enabled.

One of the exposed problems on gigabyte boards was that a bug with 'previously set' voltages was occuring so that you could reset the BIOS or go to auto, and it would still use the previous value. Manual settings should work, but beware of auto.

According to GN, if you set the SoC voltage to auto in the asus side of things and then manually set it on the AMD overclocking section, you get accurate locked SoC voltages and not random overvolting based on the RAM speed/EXPO being enabled.
This may apply to your gigabyte board, as well.


The software readings are often in more than one place - you could be finding a software reading for the requested voltage, vs the actual voltage.
Try Zentimings as it tends to show a live voltage that will vary if you fire up benchmarks etc - then use that value to find out which reading is accurate on your board in HWinfo, from the CPU and motherboards readings.

For example the HWinfo value on my board has a voltage for the SoC that seems to be the 'original' voltage, VIN6
1.15V in the BIOS becomes 1.136V here, somehow. This value does not vary at all.

Zentimings varies between 1.1313v and 1.375v while running CPU-Z's benchmark

HWinfo is more or less the same, and shows that it at least gets close to the 1.15v i set under certain conditions

 

[tanaka_007]

Voltage is constantly fluctuating.
When a load is applied, the LLC counter causes the voltage to rise.
[A] fluctuates greatly, and [C] hardly fluctuates.
is the composite voltage of [A] and [C], not the actual sensor voltage.
If you want to know the true voltage, use a tester to measure the voltage across the capacitor legs on back of motherboard.
*Since voltage drop varies depending on number of capacitors and routing of wiring pattern, etc. characteristics differ for each PCB product.
*If you want faster transient response, set a faster SOC VRM switching frequency. Instead, set the voltage and LLC weaker. (ASUS/MSI)

Digital Multi Meter detects [A].
M/B usually shows .
CPU Sensor usually shows [C].

Example
A=1.400v # SOC VRM capacitor voltage (VID_SOC + Offset + LLC)
B=1.350v # (A+B)/2 # Calculated Virtual SOC Voltage
C=1.300v # SOC Internal Voltage (CPU_vSOC)