Skrivet av Yoshman:
Om kretsen inte tål Tjunction (som är 100°C för hela Kaby Lake serien) är kretsen defekt för det är en temperatur som Intel garanterar.
En anledning att kretstillverkarna (inte bara Intel) försöker skapa kretsar som tål att högre temperaturer är en direkt konsekvens av att man hårdsatsar på mobila enheter. Ju större delta man kan ha mellan kretsytan och omgivningen, desto mindre kylfläns, lägre luftflöde etc kan man ha för att kyla bort en specifik mängd utvecklad värme. Det är ren termodynamik, inte någon form av bias.
Förkortar hög spänning och hög temperatur livslängden? Det gör den absolut.
Men normala livslängden på en CPU är i normalfallet något man får mäta i decennier, minskar den till hälften är det fortfarande längre än vad som har någon praktiskt relevans. Jämför med allt gnäll kring Samsungs fix av 840 EVO. Ja fixen minskade livslängden, men den var långt över tio år med rätt hård användning så att den minskade vara i praktiken totalt irrelevant.
Så länge man bara kör med en V-core som är lika med eller marginellt över det max tillåtna enligt spec's kan man väl förvänta sig att det inte påverkar livslängden så att det har någon nämnvärd betydelse, det kommer att ta så lång tid att man har uppgraderat långt innan problemen inträffar. Men börjar man köra med ännu högre spänningar är det ett rent lotteri OM det öht. kommer att fungera och hur länge. Det finns väl normalt en säkerhetsfunktion i CPU'n som stänger av den om den blir för varm men något liknande skydd för V-core finns väl inte?
Jag minns den här gamla godingen från HardForum där man diskuterade 'Max vCore for Q6600 on Water ?'....
"
Heat does not kill Core 2;s or Quads or the MCH/northbridge These chips have a programed Catastrophic Temp Trip point calibrated/set at the factory and the chip will just shut down to preserve itself. If it gets too hot it will just turn itself off.
Transistor junctions are basically two conductive layers separated by an partialy insulating layer. If you put too high of a voltage across the conductive layers it will short through the insulating layer damaging the transistor junction permanently. Voltage stress causes the transitor junction to no longer work.
The Intel data sheet specifies the Absolute Maximum Vcore as 1.55V. In the engineering world Absolute Maximum means just what it says. If you are building widgets you make damn sure under any conditions encountered you stay well under that. It is not heat that kills, its voltage, your temps have nothing to do with it.
But where did this Absolute Maximum Voltage come from ? Typically it is a function of the technology used in making the chip. The layers of semiconductor material used in making the chip, as well as the interconnects and other physical parts of the chip are controlled with unbelievably accurate manufacturing methods and their characteristics (thickness, what they are made of, impurities, etc.) are precisely known. It is then easy to calculate the voltage at which it is likely something will fail. Some destructive part testing is done as well but in general because of the long term expertise in making semiconductors this kind of calculation is not difficult to do and is very accurate when you are considering making 200 million widgets and you do not want any of them to fail. Also it is likely that some "worst case" or "marginal" values are tossed into the calculation as engineers tend to be conservative people and get annoyed when marketing types start screaming about widgets failing in the field and they also know there are customers like you out there that just have to push the widget as hard as they can. It is true that keeping the chip cool helps with the physical properties that tend to reduce junction failure due to voltage stress. But just a little and under your conditions, very little. However it does explain why if you dumped liquid nitrogen on the chip it could handle much higher voltages that otherwise would kill it quickly.
My point is, there are CPUs out there that can take more than 1.55V but its like running through a minefield blindfolded and with snowshoes on, twice. Sooner or later, and probably sooner, boom!. You are already pushing it.
This does not address the long term effects of running close to the absolute max, lifetime could be shortened but we typically don't keep these CPUs for too awful long. I have noted a few people indicating their chip, that was heavily overvolted when the c2ds first came out last year, will no longer overclock well but its not hard proof of anything.
This does not address the overshoot of the voltage requlator circuitry due to a rapid change in load current creating spikes that will, if you are already close to a limit take you over or under it.
My opinion (since you kinda asked for it, well maybe not but here it is anyway )
So a lot of the decision hinges on your situation, you are closer to max voltage than I would run my chips at. If you can afford and do not mind having to replace the chip who cares. otherwise you are already too close to trouble for my taste. A 50mV transient will take you at or over the spec and that is within the amount of ripple your power supply can put out under normal conditions. The accuracy of your voltage readings is also an issue, without a calibrated Oscope you really have no way of knowing exactly what voltage your CPU is seeing. Normally this is not a problem but when you get to the edge things that do not normally matter, matter.
"
https://hardforum.com/threads/max-vcore-for-q6600-on-water.12...