Exploding CPU's

[Guydodge]

https://www.youtube.com/watch?v=kiTngvvD5dI

[Bad 💀 Motha]

It is OK and deemed "normal" to have your RAM dynamically change voltage within a small amount. It might not stay at a continuous static 1.2, 1.35, or 1.5 volts. Also the voltage for DDR4 differs as well; most DDR4 at 2666 or less is around 1.2 volts; where as DDR4 above that; such as 2800 and above are 1.35 volts. DDR3 is typically 1.5 volts

So lets say I'm using DDR3 1600 1.5V kit, it is normal to maybe see this 1.5 volts change from time to time to somewhere between 1.48 - 1.52 ~ something along those lines. But there is no reason when you set 1.5V in the BIOS or its locked in from an XMP Profile that this 1.5V should be able to hit like 1.55 or higher all on its own, NO SIR. There is something wrong there.

[Illusion of Progress]

It's not the RAM voltage that's involved here. It's the SoC voltage (part of the CPU), which the motherboard sets. When Expo is enabled, the board gives the SoC more voltage to remain stable now that the IF and IMC are running higher speeds. The problem is, this extra voltage is apparently beyond the limits of what the SoC can tolerate. It has nothing to do with RAM voltage.

[Bad 💀 Motha]

Sorry yes I understand I was simply using that as an example.
Especially with ASUS Motherboards because I usually have to set a voltage in the BIOS on ASUS Motherboard just a tad below what I really want it to be at. Otherwise it will be higher in reality if I see the proper even numbered values.

[Guydodge]

İlk olarak kingjames488 tarafından gönderildi:

heh, heard about this and instantly remembered all the people who are always recommending that people update their BIOS...

cuz ya, that's always a great idea... especially when it voids your warranty (again HA at people who tried to tell me this wasn't a thing).

pretty sure you can refer to these updates as a downgrade they are trying to figure out
how AMD can avoid a recall.blame the M.B. manufacturers all you want you know
this was all considered long before the m.b. were manufactured.what changed ? obviously
it was the processor.avoid this processor theres more to the story here.

[Bad 💀 Motha]

Sounds like what's happening with some games too, in a way. Steam wants answers about a games support, customers complaining... meanwhile the game dev is pointing the finger at the publisher and the publisher is pointing the finger at the game dev.... sigh

[Overseer]

If you follow the issue you will see that so far everything points at the motherboard manufacturers. Especially ASUS. They decide the numbers not AMD. And AMD can't lock them out of these numbers because of overclocking features, especially on those super expensive boards. And if security features would work, no matter how much a CPU would ask for, the board could deny it.

[AmaiAmai]

İlk olarak The Commendatore tarafından gönderildi:

İlk olarak Bad 💀 Motha tarafından gönderildi:

It could also be that what you see in the bios is wrong. And apps inside Windows are just going off of that incorrect data. It's hard to be sure of such things

Agreed. I had a motherboard and the voltages it showed were off by like 0.2 or something.

I identified the issue, it was a bug in the firmware implementation. The scaling was set with the wrong value on auto or manual. MB manufacturer released fixed version. Got 6000 at 1.26 SoC now properly...

Well, at least the board did power down and/or refuse to run at all once it went over the new settings, so that's something. There does not appear to be any damage to the CPU

[Bad 💀 Motha]

Well that's great news. Yes they have to do something and do it right; otherwise we're talking some lawsuit stuff up in here.

But now you see why myself and those I have come to me for suggestions have all avoided AM5 for now. It's like Ryzen 1st Gen all over again. This time around it's even worse it seems.

You would think AMD would actually get samples of the 3rd party Motherboards and run them through various tests themselves. They should start doing that if not already. Run them through a series of stress tests and check all the BIOS settings, etc. before the board is allowed to be approved for mass production.

[GOD RAYS ON ULTRA™]

I still wouldn't trust it, like leaving it on when you're not home or something similar. If Im sleeping I wouldn't leave it running either.

I would send it back personally and go for am4 or Intel. This is a catastrophe if I may be honest. My next build is going to be am4 I know that.

Intel 12th gen wasn't very impressive either with their F-ed up latching mechanism. Am4 seems to be the way to go if you want something solid.

[Bad 💀 Motha]

Intel 14th Gen should be out around Fall 2023

[chickabumpbump]

İlk olarak Illusion of Progress tarafından gönderildi:

İlk olarak chickabumpbump tarafından gönderildi:

Has this been seen in B550 motherboards, is it only " next " gen ones?

It's only a problem on AM5, and largely when using Expo. Expo is now AMD's equivalent of XMP (which prior generation stuff did as well, but the motherboard manufacturers just gave it their own name like DOCP on Asus or EOCP on Gigabyte in the BIOS).

The SoC, like everything, has a limit. In order to get the system running with Expo RAM speeds, the SoC voltage is raised when Expo is enabled. This was setting them to run around 1.3V to 1.35V from my understanding. Some would see brief spikes to 1.4V or even 1.5V I think? The latter was causing issues, and when the 7800X3D specifically released (wondering why this wasn't caught with the Ryzen 9 X3Ds, maybe too few sold?), it became apparent because the X3D chips have lower voltage limits.

My 5800X3D runs at much lower voltages (~1.25V) than my 3700X did (which would run up to 1.4 or even 1.5V on spikes, but again, these values were never used under 100% use so it wasn't damaging to it, but it COULD if it used those voltages under heavy loads).

With AM4 CPUs, the SoC tends to run around 1.05V to 1.1V, at least in my experience (I saw 1.1V on both my AM4 CPUs and that was on two different boards and with a heavy memory configuration, so I don't imagine the values these AM5 CPUs are seeing is normal on AM4).

I made a better thread before this one explaining the issue here.

https://steamcommunity.com/discussions/forum/11/3821910251182999565/

TL;DR; This isn't a thing to worry about on AM4. Check your SoC voltage if you want reassurance but it's probably ~1.1V at most. The new platform when using Expo is just causing the SoC voltage to get too high. Especially a risk for the X3D chips but also a lesser but still risk for the rest. If you're on AM5 and the SoC voltage is at 1.25V to 1.3V at most you're not at risk either (presuming your particular board doesn't spike it higher like some do; MSI is said to not being doing this but that was a claim put out by themselves so take it for what its worth).

Thank you, information and not condescending! Youre the 1% of PC gamers lol

[Raoul]

Intel has not been immune from this kind of thing either, last year I warned people that using a 10th gen i9 with Asus Z590E board and enabling XMP for ram kits above 3200mhz was auto setting crazy IO/SA voltages with a DDR4 4000 kit it was at 1.45/1.6v after training while in my own testing it was stable at 1.15v/1.2v and still works as that to this day. To be clear even at 3466/3600/3733/3866 it would auto set it way beyond what is required.