AMD Releases Second Official Statement Regarding Ryzen 7000X3D Issues

[trparky]

I remember review sites calling out certain Intel boards that enforced factory recommendations by default for being "too restrictive", "too slow" and "lying to the customer", and I just thought those boards should be praised, not the ones that break those recommendations.

Yeah, but enthusiasts demanded faster and faster performance and damn what may happen in the long run. We essentially did it to ourselves.

And I'm guilty of the same thing. I've been around PCs for a long time now and yeah, I want to squeeze every last drop of performance out of my hardware.

 

[Mussels]

Wow, that's cool! What are the chances of an Australian class-action lawsuit against manufacturers that break OCP, for example?

they'd have to cause a fire or something for major action, but products could be recalled
the companies that certified them would have questions asked of them, like how are these products passing testing - does just soldering on a chip quality them, even if its not actually doing anything?

 

[mikehcl]

Hi, I upgraded to the latest F10C bios my Gigabyte X670E Aorus Master, Ryzen 7950x3d and EXPO enabled, yesterday after 5 minutes of testing on Cyberpunk and 15 minutes of gaming on world of warships I checked Hwinfo64 finding these crazy voltages! But that's not possible come on, the CPU should be dead instantly and anyway even the average voltages are strange, if indeed I reached such high peaks shouldn't the averages be higher?

 

[HD64G]

Hi, I upgraded to the latest F10C bios my Gigabyte X670E Aorus Master, Ryzen 7950x3d and EXPO enabled, yesterday after 5 minutes of testing on Cyberpunk and 15 minutes of gaming on world of warships I checked Hwinfo64 finding these crazy voltages! But that's not possible come on, the CPU should be dead instantly and anyway even the average voltages are strange, if sffectively or reached such high peaks shouldn't the averages be higher?

For sure the hwinfo doesn't show the correct values or else the board and the CPU would have been fried. Check with OCCT.

 

[mikehcl]

For sure the hwinfo doesn't show the correct values or else the board and the CPU would have been fried. Check with OCCT.

the problem is that now I am afraid to enter the bios, activate EXPO and test with OCCT xD
Update:
When I entered the bios to reactivate the EXPO to run the tests, I checked some parameters and to my amazement I found the SoC/uncore OC voltage option activated ( I never activated anything OC), I updated the bios a few days ago with the F10C version and after that I only activated the EXPO, nothing more. But how is it possible that the bios set me to enabled this option? probably this is the cause of these crazy voltages? then I loaded the default bios settings and enabled EXPO then I checked the SoC/uncore OC voltage option again and now the setting is on Auto.

 

[HD64G]

the problem is that now I am afraid to enter the bios, activate EXPO and test with OCCT xD
Update:
When I entered the bios to reactivate the EXPO to run the tests, I checked some parameters and to my amazement I found the SoC/uncore OC voltage option activated ( I never activated anything OC), I updated the bios a few days ago with the F10C version and after that I only activated the EXPO, nothing more. But how is it possible that the bios set me to enabled this option? probably this is the cause of these crazy voltages? then I loaded the default bios settings and enabled EXPO then I checked the SoC/uncore OC voltage option again and now the setting is on Auto.

My humble advice is: Do not leave the SOC Voltage at auto setting. Set it at 1,2V and check RAM stability with OCCT. If not stable enough get the voltage at 1,25V and check again.

 

[mikehcl]

My humble advice is: Do not leave the SOC Voltage at auto setting. Set it at 1,2V and check RAM stability with OCCT. If not stable enough get the voltage at 1,25V and check again.

ok thank you, would you tell me from where to set the SoC voltage from the Gigabyte bios? where can i find it? thank you.

 

[trparky]

ok thank you, would you tell me from where to set the SoC voltage from the Gigabyte bios? where can i find it? thank you.

I’d love to know that too.

 

[Mussels]

Hi, I upgraded to the latest F10C bios my Gigabyte X670E Aorus Master, Ryzen 7950x3d and EXPO enabled, yesterday after 5 minutes of testing on Cyberpunk and 15 minutes of gaming on world of warships I checked Hwinfo64 finding these crazy voltages! But that's not possible come on, the CPU should be dead instantly and anyway even the average voltages are strange, if indeed I reached such high peaks shouldn't the averages be higher?

Do you have snapshot polling enabled? It does help with accuracy of some values

Since only "CPU die" reached that temperature, either only one spot on the CPU got that hot - or it's a false reading.

I’d love to know that too.

It's always in the same area as CPU and DRAM voltages

 

[mikehcl]

Cpu: Ryzen 9.7950x3d
Motherboard: Giagabyte X670E Aorus Master
Ram: 32gb, G Skill, Cl 30, 6000Mhz
I have already proceeded to write to gigabyte, who replied immediately that they forwarded everything and as soon as they know more they will let me know. There is something strange in their latest F10C bios, last night I entered the bios, set the recommended bios configuration and then activated the EXPO, saved and tested the voltages with OCCT, so far so good, even the tests seemed correct, the voltages of the SoC had not exceeded 1.245v, but after a few hours while the PC was idle I looked again at the voltage values on OCCT and I saw very high values, 1.800V of the SoC. I have now deactivated EXPO.

 

[Mussels]

I have already proceeded to write to gigabyte, who replied immediately that they forwarded everything and as soon as they know more they will let me know. There is something strange in their latest F10C bios, last night I entered the bios, set the recommended bios configuration and then activated the EXPO, saved and tested the voltages with OCCT, so far so good, even the tests seemed correct, the voltages of the SoC had not exceeded 1.245v, but after a few hours while the PC was idle I looked again at the voltage values on OCCT and I saw very high values, 1.800V of the SoC. I have now deactivated EXPO.

Holy shit
Did the PC enter a sleep or hibernate state in that time?
Can you edit the exact motherboard CPU and RAM you have into your post, as i'm sure a lot of people are going to quote that post or screenshot it in the coming days

 

[mikehcl]

Holy shit
Did the PC enter a sleep or hibernate state in that time?
Can you edit the exact motherboard CPU and RAM you have into your post, as i'm sure a lot of people are going to quote that post or screenshot it in the coming days

I did the various tests in the afternoon, the Expo was activated and everything went well, then I went for 30 minutes to have dinner, the PC went into suspension after 15 minutes, when I came back I moved the mouse, it came out of suspension, I logged into discord and later went on twitch to watch a friend of mine who was live, shortly after i checked OCCT and found the voltages went crazy and immediately rebooted and entered the bios to disable expo.
I don't know what to think anymore.
Anyway, I edited the above post.

 

[Mussels]

I did the various tests in the afternoon, the Expo was activated and everything went well, then I went for 30 minutes to have dinner, the PC went into suspension after 15 minutes, when I came back I moved the mouse, it came out of suspension, I logged into discord and later went on twitch to watch a friend of mine who was live, shortly after i checked OCCT and found the voltages went crazy and immediately rebooted and entered the bios to disable expo.
I don't know what to think anymore.
Anyway, I edited the above post.

sleep and hibernate have been mentioned as the cause of the high voltages by other users - disable sleep and hibernate on the PC in the meantime

 

[trparky]

It's always in the same area as CPU and DRAM voltages

But what setting? I tried messing with VCORE SOC settings and it appeared that it did nothing.

 

[HD64G]

So, the board vendors seem to have become too lazy to do their part in the quality control of the UEFI. With such high pricing this is totally unacceptable.

 

[trparky]

sleep and hibernate have been mentioned as the cause of the high voltages by other users - disable sleep and hibernate on the PC in the meantime

I could understand sleep doing that but hibernate? Hibernate is simply a more involved process of shutting down.

 

[mikehcl]

Here is Gigabyte first response, which says absolutely nothing new.
As you can see in the ticket I answered them, now I'm waiting for them to give me a little clearer explanation

 

[AusWolf]

I could understand sleep doing that but hibernate? Hibernate is simply a more involved process of shutting down.

As I see it, hibernate is the same as sleep, just instead of keeping the RAM alive, its image is saved onto your boot drive.

 

[trparky]

As I see it, hibernate is the same as sleep, just instead of keeping the RAM alive, its image is saved onto your boot drive.

That’s what I mean. It’s a full shut down after a full memory dump to disk.

 

[Mussels]

But what setting? I tried messing with VCORE SOC settings and it appeared that it did nothing.

That's the setting, you just need to type the value you want

As an example on my 5800x3D, 0.95v let me run 3800MT/s easily at load with zero problems - but at lower loads, especially transitions like starting up a 4K youtube video, i'd get PCI-E dropouts and black screen crashes randomly. 1.10v worked but had random USB dropouts, most noticeable with my keyboard where holding W would just keep me running forward in-game after letting go, and my mic (which defaults to muted on a power up)
1.13v+ solved all of that

As I see it, hibernate is the same as sleep, just instead of keeping the RAM alive, its image is saved onto your boot drive.

Hibernate uses sleep at the same time on some setups - it sleeps, but on power loss resumes from hibernate.
IIRC, the BIOS option for that was 'ACPI sleep/suspend state' and listed it as "S4" or "S3 + S4" (Hibernate only, or suspend + hibernate)

I can't find this at all in my B550 board, so i cant tell what the modern default is
Without knowing what systems default to, it's safest to advise against hibernate in the meantime

 

[trparky]

That's the setting, you just need to type the value you want

According to HWInfo, VCORE SOC runs at a maximum of 1.188 volts and fluctuates between that and 1.176 volts. However, what really confuses the hell out of me is that VDDCR_SOC is locked at 1.25 volts. Why are the two SOC voltage numbers so different?

 

[Mussels]

According to HWInfo, VCORE SOC runs at a maximum of 1.188 volts and fluctuates between that and 1.176 volts. However, what really confuses the hell out of me is that VDDCR_SOC is locked at 1.25 volts. Why are the two SOC voltage numbers so different?

BIOS settings are a VID, a requested voltage - under load, they drop down to actual voltage. Not all boards have measurements for both, and various overclocking features like voltage offsets cheat on this - they tell the CPU it's 1.10v VID, but actually send 1.15v or 1.05v depending on what you chose to do (like AMDs curve undervolting, this is how it works)

Mines set to 1.15v in the BIOS (VID) but the actual voltage comes out lower after some is turned into heat by resistance

LLC settings in the BIOS counter this, but can also lead to the over volting problems people have experienced.
The entire concern people have with this current issue is that they could see 1.1v in the BIOS/software and 1.8v with a voltmeter probing the contacts

 

[trparky]

BIOS settings are a VID, a requested voltage - under load, they drop down to actual voltage. Not all boards have measurements for both, and various overclocking features like voltage offsets cheat on this - they tell the CPU it's 1.10v VID, but actually send 1.15v or 1.05v depending on what you chose to do (like AMDs curve undervolting, this is how it works)

Mines set to 1.15v in the BIOS (VID) but the actual voltage comes out lower after some is turned into heat by resistance
View attachment 294399

LLC settings in the BIOS counter this, but can also lead to the over volting problems people have experienced.
The entire concern people have with this current issue is that they could see 1.1v in the BIOS/software and 1.8v with a voltmeter probing the contacts

Not really putting my worries to rest here dude.

 

[AusWolf]

BIOS settings are a VID, a requested voltage - under load, they drop down to actual voltage. Not all boards have measurements for both, and various overclocking features like voltage offsets cheat on this - they tell the CPU it's 1.10v VID, but actually send 1.15v or 1.05v depending on what you chose to do (like AMDs curve undervolting, this is how it works)

Mines set to 1.15v in the BIOS (VID) but the actual voltage comes out lower after some is turned into heat by resistance
View attachment 294399

LLC settings in the BIOS counter this, but can also lead to the over volting problems people have experienced.
The entire concern people have with this current issue is that they could see 1.1v in the BIOS/software and 1.8v with a voltmeter probing the contacts

I never knew what VID voltages meant. Thanks for explaining it.

 

[R0H1T]

As I see it, hibernate is the same as sleep, just instead of keeping the RAM alive, its image is saved onto your boot drive.

Pretty much except there's also hybrid sleep, which is similar to what Mussels said, it doesn't power off the rig but saves your OS' current state in case of power failure & you can resume using mouse/keyboard etc.