B550-E w/ 5900X delivering more voltage than it should - why?

[Piers]

This build is only a few weeks old. Loading the optimised defaults in the BIOS (latest version as of writing) makes the VID higher, which makes the Vcore go above 1.5v, according to HWINFO64. Even with an undervolt and power limits on stock (PBO -> disabled -> curve optimiser -> -20 on 'low cores' and -7 on four 'high cores' results in slightly less voltage being delivered, but it's still too much.

Again, with optimised defaults loaded, AMP/XMP/DOCP enabled, and everything else on stock, the VID goes to 1.55v and the SVI2 TFN goes to 1.52v-1.53v. This is absolutely outside of the stated specification for the CPU.

The first screenshot is from a 15 hour run with CoreCycler - a SSE single-threaded workload. Vcore (SVI2 TFN) is normal. Vcore is abnormal. VID is abnormal.


The second screenshot is from playing Civilisation VI with the values reset during the game (I played for 3 hours). Vcore (SVI2 TFN) is abnormal. Vcore is normal. VID is abnormal.


The final screenshot is from running Cinebench R23 single core on a loop. Vcore (SVI2 TFN) is abnormal. Vcore is normal. VID is abnormal.


Could this be an LLC issue (it's set on auto)?
Is this simply a defective motherboard (apart from the still-defective I225-V_REV03)?
Could there be another cause?

Any help is appreciated. I've never regretted any purchase more than this motherboard...

[Nate152]

Hi Piers,

I don't have an AMD motherboard but I agree 1.50v is way too high, especially at stock settings.

LLC is a good setting to start with, I use LLC to help keep the voltage (Vcore) close to the voltage I set in the bios. A lower value should help get the maximum Vcore to a lower voltage.

[Piers]

Nate152 wrote:
Hi Piers,

I don't have an AMD motherboard but I agree 1.50v is way too high, especially at stock settings.

LLC is a good setting to start with, I use LLC to help keep the voltage (Vcore) close to the voltage I set in the bios. A lower value should help get the maximum Vcore to a lower voltage.

Hello Nate, and thank you for replying.

After spending nearly a week with this issue, I've reset the BIOS again using the optimised defaults in case something went wrong (the BIOS is buggy - an example is that disabling LN2 mode changes voltages and doesn't appear to do anything else. Another example is that Chinese characters show when going to save or load a profile. A third example is the I225-V only sometimes appearing in the BIOS under Advanced, but it seems that's a joint Asus and Intel problem). It's certainly odd coming from a Z170 Maximus Hero (6700K) with its rock-solid and well-programmed BIOS to this..... mess.

As of typing, 30 mins of CB R23 single core shows the expected AMD values - a VID of up to 1.5000v is considered acceptable, and brief SVI2 TFN spikes of up to ~1.5000v - with stock settings (unlike last time), although it's only been 30 minutes. I wish CB R23 didn't have the 30 minute limit so I could leave it running to complete a long test.

[Nate152]

Stock values are usually set pretty high to cope with underwhelming cpu's.

Try manually setting a lower cpu voltage and playing with LLC, you may have to try adaptive or offset mode.

This is about all I'm going to have as I am not familiar with your bios.

[Piers]

Nate152 wrote:
Stock values are usually set pretty high to cope with underwhelming cpu's.

Try manually setting a lower cpu voltage and playing with LLC, you may have to try adaptive or offset mode.

This is about all I'm going to have as I am not familiar with your bios.

Again, thank you for replying. The problem is that the images above show the CPU with Curve Optimiser set to negative per-core values.

In case you're not aware (that sounds patronising - that's certainly not my intent and I apologise if it comes across that way), negative values can be set per-core (and all-core) within AMD's Curve Optimiser in the BIOS. This can go from 0-30, with each value representing 3mV-5mV (-30 all-core would equal a maximum undervolt of 150mV). The best my CPU is able to handle is a negative value of ~8 for the fastest cores (cores 0,5,7,8), with the remaining eight lower-performing cores being stable with a negative value of ~20. This translates to the fastest four cores having voltage reduced by 24mV-40mV, and the remaining eight cores having voltage reduced by 60mV-100mV.

[Silent_Scone]

Hello,

1.5v is normal for this platform and is intrinsically related to how Zen handles idle voltage and parks cores. Many Intel users fall trap to these levels when crossing over to team red.

What's important is the sustained voltage when the CPU is under high current loads. Around 1.325v is a good ceiling to keep for these types of workloads.

[Piers]

Silent Scone@ROG wrote:
Hello,

1.5v is normal for this platform and is intrinsically related to how Zen handles idle voltage and parks cores. Many Intel users fall trap to these levels when crossing over to team red.

What's important is the sustained voltage when the CPU is under high current loads. Around 1.325v is a good ceiling to keep for these types of workloads.

After speaking with The Stilt, whose name is featured in most ROG motherboard BIOSes, he disagrees that 1.5500V is acceptable VID. He's manually confirmed that my CPU is factory set to 1.50000V and the Asus B550-E is doing something to force that to 1.5500V, which in turn causes higher Core SVI2 TFN readings. I imagine he's registered here and can confirm the public and private conversation I've had with him.

The Stilt spent an entire day looking through 9 BIOS version over two Asus motherboards and couldn't work out what you've done to violate AMD's factory limit. Can you, or someone from Asus, please explain? I don't mind showing proof of this conversation with The Stilt privately with an Asus employee and providing more information.

tldr; At entirely stock settings in the BIOS, Asus is doing something to force vmax over 1.50000V to 1.55000V. This does not meet AMD's stock specifications. BIOS version is 2423 "AMD AM4 AGESA V2 PI 1.2.0.3 Patch C".

[Silent_Scone]

Piers wrote:
After speaking with The Stilt, whose name is featured in most ROG motherboard BIOSes, he disagrees that 1.5500V is acceptable VID. He's manually confirmed that my CPU is factory set to 1.50000V and the Asus B550-E is doing something to force that to 1.5500V, which in turn causes higher Core SVI2 TFN readings. I imagine he's registered here and can confirm the public and private conversation I've had with him.

The Stilt spent an entire day looking through 9 BIOS version over two Asus motherboards and couldn't work out what you've done to violate AMD's factory limit. Can you, or someone from Asus, please explain? I don't mind showing proof of this conversation with The Stilt privately with an Asus employee and providing more information.

tldr; At entirely stock settings in the BIOS, Asus is doing something to force vmax over 1.50000V to 1.55000V. This does not meet AMD's stock specifications. BIOS version is 2423 "AMD AM4 AGESA V2 PI 1.2.0.3 Patch C".

Hello,

How are you measuring the core voltage? Are you solely basing all this on the SVI2 TFN readout?

Can you forward me any information you have, please.

[Piers]

Silent Scone@ROG wrote:
Hello,

How are you measuring the core voltage? Are you solely basing all this on the SVI2 TFN readout?

Can you forward me any information you have, please.

It's more VID, which The Stilt pointed out is going against AMD's specification and the CPU's factory-set vmax. I'll send you a private message with how The Stilt confirmed the voltage and with a couple of threads explaining his testing.

[Silent_Scone]

Piers wrote:
It's more VID, which The Stilt pointed out is going against AMD's specification and the CPU's factory-set vmax. I'll send you a private message with how The Stilt confirmed the voltage and with a couple of threads explaining his testing.

Right, so this isn't indicative of what the CPU is receiving. If you could, please! 🙂