Ryzen temperatures and voltages

OK..I've read a lot of threads covering recommended temperature and voltage for Ryzen cores so I'm not really asking about that that.

When I first started overclocking my FX processor (FX-6300) I remember people were super skittish about getting over 1.4V and keeping temperature around 55C so I took my guidance from that. But now I'm seeing people with1.55V / 80C are saying that's 'sustainable'...crazy sounding to me!

So I'm curious where Ryzen is when being run on the extreme high side, and this is mainly for curiosity. I think Ryzen is considered hard thermal limited. If mine climbs very much over 70C it's very likely to crash, for instance, regardless of frequency, load or voltage. So I don't think it'll ever see 80's and that 90C limit where it protection-throttles seems mythical.

But what voltages? I KNOW AMD recommends 1.45 as a voltage limit above which life is shortened, but that's all very theoretical and based on their models. What's happening in practice?
 

Dunlop0078

Titan
Ambassador
Well we don't know yet. It would take long term testing to know what voltages and accompanied temps will degrade a ryzen cpu long term. 1.55v certainly wont kill a ryzen cpu outright if you can keep it cool, we don't yet know how that will play out long term. I certainly wouldn't recommend pushing 1.55v for say a gaming PC, I would stay at 1.45v preferably lower.
 


With my board voltage is only a problem at low loads as the LLC it offers at best a joke. At low load it's around 1.39V but once loaded up in Prime95 it drops to 1.31. This is at 3.9G (a 1700). It will hold at 3.95 but low load voltage then is 1.44-1.45 and at P95 load it drops to 1.36.

Temp is OK...high 60's to low 70's on the processor under a 240mm AIO. It's the low-load voltage that worries since the only time it sees a drop that extreme on all cores is in Prime95. Playing games only OCCASIONALLY does one core come close to 100% load and even then it's hardly as well optimized as P95 so temperature never exceeds 50C. Even encoding, when all cores will go to 90+% with fairly well optimized code temps are well controlled, peaking into the 60's briefly.

It's strictly that low-load voltage that bothers so I'm trying to get a feel whether letting it peak at/over 1.45V is OK, long term.

Also: I am using what's called STI2 - TFN core voltage, the actual voltage at the core of the processor as reported by telemetry and made available by HWInfo64. The VCore readings provided by the controller on my board are right at the output of the VRM and not reflective of board losses getting to the socket. That's a major reason that LLC is so ineffective too, i imagine, if that is the VSense input to the VRM controller.

EDIT ADD: I THINK it will hold at 4.0G but then I'll have to bump voltage to 1.5 which I'm not willing to do. Also, VRM temps in P95 would definitely be excessive and throttle even with a fan on the heatsink. But in routine use, including gaming, it would probably work OK.