• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Ryzen 7 7800X3D CPU Burns Up

VSOC uses auto rules. Not sure if there are any kits in the wild that have the values programmed into SPD, but it doesn’t change the outcome of what I said.
Sure, I'm aware EXPO/XMP doesn't set voltages, was more just a comment because you mentioned about them setting higher. Was just kind of clarifying I was referring to what people were getting when enabling EXPO. While people probably shouldn't be able to set unsafe voltages in a BIOS manually, it's slightly less concerning than something that is just set from a common overclock most people are encouraged to use.
 
Sure, I'm aware EXPO/XMP doesn't set voltages, was more just a comment because you mentioned about them setting higher. Was just kind of clarifying I was referring to what people were getting when enabling EXPO. While people probably shouldn't be able to set unsafe voltages in a BIOS manually, it's slightly less concerning than something that is just set from a common overclock most people are encouraged to use.

That gets into the same nonsensical ideological nod and wink that others in this thread seem to be trying to propagate regarding what was or wasn’t safe.

We can touch on that, briefly. Remember the fabric itself isn’t high current. You don’t need exotic cooling in order to be liberal with it. It will help, but it’s not needed. If 1.4v limit set by AMD is indeed too much, electromigration is probably still going to occur at a rate fast enough to become an issue within the CPUs useable lifespan with or without exotic cooling. Remember Roman killed his 7950X on LN2. I think even AMD were shocked at how little voltage it took, and this CPU failed in a similar fashion to the end user cases. Some of that is obviously speculative.
 
Last edited:
For those following along, this is not the place to probe for accurate voltage readings:

image.png



If the discussion was around Asus auto rules using more voltage than others, there'd be some merit but they're still within tolerance range set by AMD via Agesa rollouts.
 
That gets into the same nonsensical ideological nod and wink that others in this thread seem to be trying to propagate regarding what was or wasn’t safe.

We can touch on that, briefly. Remember the fabric itself isn’t high current. You don’t need exotic cooling in order to be liberal with it. It will help, but it’s not needed. If 1.4v limit set by AMD is indeed too much, electromigration is probably still going to occur at a rate fast enough to become an issue within the CPUs useable lifespan with or without exotic cooling. Remember Roman killed his 7950X on LN2. I think even AMD were shocked at how little voltage it took, and this CPU failed in a similar fashion to the end user cases. Some of that is obviously speculative.
Excuse me?

I don't even understand what you're trying to say if I'm being honest. AMD themselves are saying it's not safe now. What's the nonsensical idealogical nod? I'm just using that as a basis.
If the discussion was around Asus auto rules using more voltage than others, there'd be some merit but they're still within tolerance range set by AMD via Agesa rollouts.
Which is exactly what I said.
 
when i finally get around updating my asus hero board (i'm away) to the new bios and able to run my memory from stock 4800 to 6000mhz what kind of performance % increase can i expect? like for cinebench/some other sythentic tests and games like microsoft flight sim which is quite cpu dependent?
 
Excuse me?

I don't even understand what you're trying to say if I'm being honest. AMD themselves are saying it's not safe now. What's the nonsensical idealogical nod? I'm just using that as a basis.

Which is exactly what I said.

I was referring to others saying AMD’s 1.4v limit and certain individuals trying to suggest vendors shouldn’t have been using that much as if there was some kind of gentleman’s agreement. Almost by the by, regardless as some of these cases would have been lower than this.
 
Last edited:
For those following along, this is not the place to probe for accurate voltage readings:

image.png



If the discussion was around Asus auto rules using more voltage than others, there'd be some merit but they're still within tolerance range set by AMD via Agesa rollouts.

That image is measuring current. And zero current at that.
 
Follow up question: who sets the voltage thresholds for the auto rules? Asus on their own accord or guidance from chip manufacturer?
How auto works should be individual to each board. However I would have thought they all do take what AMD says into mind. Like what voltage is the max safe voltage and so on when setting what auto will do.
 
Pre-flashed my Asus X670E-A before fitting a 7800X3D. It actually booted first time and didn't explode or catch fire or anything! :cry:

With memory at 6000 c30, it autos the vsoc to 1.25v (1.27 displayed) and the vddio to 1.4v (1.43 displayed).

For now while I install Windows, I've tweaked them to be 1.25 and 1.3 displayed, but idk if I need to worry about the vddio? 1.43 seemed like a big number after the past few weeks.
 
Pre-flashed my Asus X670E-A before fitting a 7800X3D. It actually booted first time and didn't explode or catch fire or anything! :cry:

With memory at 6000 c30, it autos the vsoc to 1.25v (1.27 displayed) and the vddio to 1.4v (1.43 displayed).

For now while I install Windows, I've tweaked them to be 1.25 and 1.3 displayed, but idk if I need to worry about the vddio? 1.43 seemed like a big number after the past few weeks.
I have SOC manually set to 1.20v and VDDIO manually set to 1.25v, no issues, dont forget to set SOC voltage manually in the AMD overclock menu too just to be sure, so if you have it set to 1.25v in your tweaker menu, set it to 1250mv in the AMD overclock menu too, this pretty much assures it wont go over what you've set.
 
I have SOC manually set to 1.20v and VDDIO manually set to 1.25v, no issues

Oof, well I was doing ok at 1.25 and 1.3, but I tried lowering them, bluescreened out of Linpack, then got stuck having a hard time even getting to bios - and if it tried to get to Windows it would complain about corrupted files.

Reset cmos, boots fine on defaults. I guess I will put the previous voltages back and be happy :cry: As far as I know, nothing has said that VDDIO being over 1.3 is dangerous, it's all been focused on the SoC, it just felt like the auto 1.43 was "very high" - but it is driving 2x32gb 6000 c30, which might need a bit more oomph.
 
Last edited:
Oof, well I was doing ok at 1.25 and 1.3, but I tried lowering them, bluescreened out of Linpack, then got stuck having a hard time even getting to bios - and if it tried to get to Windows it would complain about corrupted files.

Reset cmos, boots fine on defaults. I guess I will put the previous voltages back and be happy :cry: As far as I know, nothing has said that VDDIO being over 1.3 is dangerous, it's all been focused on the SoC, it just felt like the auto 1.43 was "very high" - but it is driving 2x32gb 6000 c30, which might need a bit more oomph.
Ahh yes you will probably need more running 64gb, im running 32gb (2 x 16gb), mine are single sided whereas yours will be double sided, pretty much the same as running 4 x 16gb sticks.
 
VDDIO is not on the cache chiplet, so voltage is different.

I run it 1.4v, same as the RAM voltages

Tbh I just knee-jerk reacted to a big number :cry: It is indeed 1.4v memory.

Having trouble finding info on what VDDIO actually is/does/should be however... all google shows is vSoC information. It's probably totally fine at 1.4v, but this being on an Asus board and knowing it was happy at 1.3 during initial Linpack testing, I will run 1.3 if possible because I bet Asus just threw the maximum safe voltage at every setting they could see.

I'll reset it all when I get home tonight, maybe Linpack it for a few hours, then look at whether I can tweak it down any with Curve Optimiser. Not sure it'll make much difference, but it'd be nice to have the thing all set up for maximum efficiency before I swap to it as my daily runner. I don't want to be doing stability testing when I could be doing actual fun things :)
 
Last edited:
I now have my VDDIO running at 1.162v, same as my SOC voltage for 6200C26.

That's interesting... I'm currently testing with Linpack at 1.2v soc and 1.27v vddio, but if it passes 100 loops I will try lowering the latter just for experiment's sake. (And if it doesn't, I know it's happy and stable well below the dreaded 1.3v anyway.)

I have discovered that the crash I got before was Curve Optimiser in Ryzen Master; even a -10 causes a blue "windows cannot verify this file" screen (possibly not a BSOD?) the moment Windows starts to load and the only way out is lowering memory (clearing cmos) to let it start and remove the CO setting. It's so sensitive that I'm a little sus it's a software problem rather than the CPU itself giving up due to the missing milivolts... every other undervolting BSOD I've ever had would take it's time to show up and was more likely to be a system lock than a BSOD.

But I kind of can't be bothered fighting with it when it's stress testing at 78 degrees with all fans on a very-quiet 60%. It is never going to get even that warm under a normal workload and I suspect it will game in the 50s-60s. It is quieter under stress test than the old rig is at idle, and pulling 120 watts less while it does it. I don't feel like I really need to fight for the last 1% of performance :cry:
 
Back
Top Bottom