AMD Ryzen 7000X3D Power Consumption Spiking Beyond 100 W in Idle Mode

[phubar]

AGESA/SMU is very much at fault for reasons he explained in the video,

If the volt tables that were wrong were in the AGESA then yes AGESA is at fault.

If they're not then its the mobo vendors who were setting the volt tables wrong.

AMD's screw up in that case would be a failure to communicate and QC, not the AGESA, which is what Steve has said in the vid.

 

[AusWolf]

If the volt tables that were wrong were in the AGESA then yes AGESA is at fault.

If they're not then its the mobo vendors who were setting the volt tables wrong.

AMD's screw up in that case would be a failure to communicate and QC, not the AGESA, which is what Steve has said in the vid.

Based on my motherboard (MSi), I'd say board vendors were lazy by defaulting to VDIMM = VSoC, which is unnecessary for normal 7000 CPUs, and even dangerous for X3D chips. I'm not saying that AMD couldn't have provided clearer guidance, though.

 

[Mussels]

EXPO has nothing to do with voltages.

It does, because more than one issue has occured at this time.
One is that boards are ignoring their own OCP safeties, the other is that the boards are raising SoC voltages to stupid levels - some get worse over time, as high as 1.5v in these reviewers tests.


If the CPU is in a 'crashed' state because the system cant post, the CPU's safety features cant be activated - and if the motherboard doesnt use any of its OVP or OCP protections at this stage that is 100% on the motherboard.

As their video showed, 400W in the same error state as a board with no CPU detected is insane and should never have happened.

Based on my motherboard (MSi), I'd say board vendors were lazy by defaulting to VDIMM = VSoC, which is unnecessary for normal 7000 CPUs, and even dangerous for X3D chips. I'm not saying that AMD couldn't have provided clearer guidance, though.

It's not guidance, they chose to do that themselves.
They chose to do that, to help overclock RAM higher on their boards when EXPO was enabled - not because AMD ever said it was a good idea.

The faults with voltages being higher than expo or the user sets are the biggest issue here, and that needs more investigation.

 

[watzupken]

AGESA/SMU is very much at fault for reasons he explained in the video, it seems everyone shares a largely equal part of the blame, more so ASUS for their excessive voltage and loose OCP, AMD not providing guidelines, and also the CPU that exploded on first boot without any EXPO applied...

With multiple vendors exhibiting the problem, it displays at the very minimum poor communication and lack of certification, as well as poor quality control by AMD. Launching a new brand to compete with XMP, alongside newer CPUs and a new memory standard, AMD should have had the foresight to establish a certification board and publish detailed guidelines. Gigabyte's board surviving shows that they've been the closest to right in a situation that is so wrong that there's no excuse for this.

Karmic rebalance complete, though. I just hope that all parties affected have their products replaced ASAP, with as little bureaucracy and questions asked as possible.

I feel AMD is at fault here by not providing strict guidelines. If motherboard makers are doing what they want, it is a big problem. It is no secret that motherboard makers will surreptitiously try to "overclock" your system so that when it comes to reviews, they will look better. So by ignoring this risk, it is a matter of time something will get fried.

 

[AusWolf]

It's not guidance, they chose to do that themselves.
They chose to do that, to help overclock RAM higher on their boards when EXPO was enabled - not because AMD ever said it was a good idea.

The faults with voltages being higher than expo or the user sets are the biggest issue here, and that needs more investigation.

I'm not saying that AMD said it was a good idea. But they could have said that it's not a good idea.

I feel AMD is at fault here by not providing strict guidelines. If motherboard makers are doing what they want, it is a big problem. It is no secret that motherboard makers will surreptitiously try to "overclock" your system so that when it comes to reviews, they will look better. So by ignoring this risk, it is a matter of time something will get fried.

This. AMD probably thought that board makers have some common sense and test things themselves, but it's apparently not the case.

 

[phubar]

It does, because more than one issue has occured at this time.

A) that doesn't prove EXPO is the issue.
B) Steve pretty clear cut that EXPO isn't the issue since it litterally can't set volts that are being effected, he says as much in the vid

Basically the mobo vendors could've easily screwed up multiple times and that would explain things just fine.

We already know for instance that Gigabyte had a bug where user settings were being ignored. MSI had a different one too. Steve also pretty clearly stated there are other bugs at play and they didn't have time for them in the video too.

If the CPU is in a 'crashed' state because the system cant post, the CPU's safety features cant be activated - and if the motherboard doesnt use any of its OVP or OCP protections at this stage that is 100% on the motherboard.

Sure but that has nothing to do with AGESA, AMD, EXPO, or the idle power consumptions spikes (the latter of which is the actual topic of this thread) so I'm not sure what you're getting at here.

 

[Mussels]

A) that doesn't prove EXPO is the issue.
B) Steve pretty clear cut that EXPO isn't the issue since it litterally can't set volts that are being effected, he says as much in the vid

Basically the mobo vendors could've easily screwed up multiple times and that would explain things just fine.

We already know for instance that Gigabyte had a bug where user settings were being ignored. MSI had a different one too. Steve also pretty clearly stated there are other bugs at play and they didn't have time for them in the video too.

Sure but that has nothing to do with AGESA, AMD, EXPO, or the idle power consumptions spikes (the latter of which is the actual topic of this thread) so I'm not sure what you're getting at here.

EXPO itself isnt the issue, some boards behaviour when EXPO was enabled was the issue

Multiple threads on multiple related issues on AM5 right now, and people are mixing them up all over the place.

As people are testing and finding out WTAF is happening, more and more comes down to some specific motherboards being the problem - you enable EXPO, they run DRAM voltage as SoC voltage.
Enable PBO, they disable OCP and OVP.

Finally found the relevant thread here on TPU, of an example of this happening on AM4 as well

5950x and PBO overheated something or busted custom loop? | Page 2 | TechPowerUp Forums


The board makers have been cheaping out. They've been relying on the CPU safeties and not using their own.

 

[phubar]

EXPO itself isnt the issue, some boards behaviour when EXPO was enabled was the issue

Glad we can agree on that part.

As people are testing and finding out WTAF is happening, more and more comes down to some specific motherboards being the problem - you enable EXPO, they run DRAM voltage as SoC voltage.
Enable PBO, they disable OCP and OVP.

On some of them yeah. There are several bugs, some of which are mobo specific, as you note some apply to AM4 too.

1usmus mentioned something about knowing of this and other problems but being asked not to talk publicly about themnot to long after this story broke on twitter. Can't find the link right now. Point is its been insider knowledge for a while though.

Amazes me they let this problem sit and fester for so long. Especially since apparently it can be either fixed or made safe through a BIOS update that only takes a week or 2 to whip up and doesn't seem to have significant negative impacts on anything!

Someone, probably several people since there different bugs at play and they can differ depending on the mobo, have been acting real stupid for a long time here.

 

[Dr. Dro]

Glad we can agree on that part.

On some of them yeah. There are several bugs, some of which are mobo specific, as you note some apply to AM4 too.

1usmus mentioned something about knowing of this and other problems but being asked not to talk publicly about themnot to long after this story broke on twitter. Can't find the link right now. Point is its been insider knowledge for a while though.

Amazes me they let this problem sit and fester for so long. Especially since apparently it can be either fixed or made safe through a BIOS update that only takes a week or 2 to whip up and doesn't seem to have significant negative impacts on anything!

Someone, probably several people since there different bugs at play and they can differ depending on the mobo, have been acting real stupid for a long time here.

Don't you understaaaaand? Each and every party involved is highly ethical! AMD and the motherboard partners are all happy little friends, of which you, the consumer, is the besto and the most loved of all of their important friends.

They'd never do something as ask an enthusiast who found a high severity problem lingering through generations of hardware to keep it hush. They're our FRIENDS and we should TRUST them, because Jensen Huang likes leather jackets (and he probably eats meat, the horror!) and Pat Gelsinger is literally Darth Vader.

Whoever even brings the mere notion that AMD and motherboard partners did anything shady up is a traitor and an Intel shill, with their fate sealed in blood. That this issue suddenly came to light was literally a coordinated and paid act of sabotage by Pat Gelsinger and has JHH's direct involvement.

Pure evil, how is poor little AMD ever going to recover from such coordinated acts of sabotage against it? Yuri should be banned from the industry for daring accuse anyone asked him to keep that hush.

...yes, this is obviously satire with an unhealthy dose of sarcasm.