980X going bad?

[Xel_NaGa]

So for 2 years I've ran my system at 4500/1.45 vcore and the memory at 1800/1.6v with temps always in the @ 38-48C range. recently(last 30 days) the system has been hard resetting (no BSOD at all) and increasingly doing so to where now the system doesn't last 5-10 mins.

I tried dialing the system back to auto and the system will post but wont even make it to desktop.

Looking in the system logs I find the following entry
[
A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Cache Hierarchy Error
Processor APIC ID: 4
]

To me this says that physical core 3 (windows lists cores including HT cores as 0-11) is crapping bricks. To confirm this I went into the bios and disable cores 3-6 and here I am now 2 hours later system not showing any errors in the system log and stable.


So what do you guys think? Busted core? or maybe needs some voltage tweaking?

[AMDforME]

It could be many things. CPUs can and do deteriorate over time when subjected to high voltages. While core 3 may be the issue, I'd also run Memtest86+ overnight to confirm the RAM isn't borderline as I have personally seen a number of brand name DIMMs go south after a few weeks to a few months of use at their rated voltage and frequency. In all cases the problem was confirmed by Memtest86+ and the RAM supplier once the RAM was returned. If core 3 is the weakest then you might be able to re-tune the Vcore or RAM voltages to improve things but there are no guarantees.

Two years @ 4500 w/1.45V is a lot of stress on a CPU, especially one that is over-volted as delivered from Intel.

[subzero 1979]

i agree but test to raise the vcore a litthe plus vtt ,run one stick of ram and see if you are ok

[Xel_NaGa]

Just to update and inform. I chatted intel and told them the errors and they said it was faulty for sure. They asked for the S/N off the heatsink(wasnt required) then they asked for the markings from the top of the cpu. They asked what model ram I was using (had to be an approved model). They asked what memory speeds, clock speeds, and voltage I was running at. also asked what mother board I had and which bios version. Told him all my information and they approved the standard return. In order to get the advanced return you need the serial number of the CPU which on the 980x is only on the sticket you cut to open the box. Sent it off and I just received a brand new 980x. had hopes a 990x would arrive but a new 980x will do just fine for those that care it is batch 3045A708 and Ill post up some numbers I get

[pheonix_ace]

well at least they gave you the time of day i am happy for you got a replacement

[Ket]

My 2500k just died exhibiting very similar behaviour (hard locks and random BSODs even at stock), in my case I've had to buy a new CPU as I lapped the IHS when I first bought it some 2.5yrs ago but I'd have to agree recent CPU generations are very fickle about prolonged exposure to voltage even when run in safe regions. With this latest experience I'm seriously thinking 22nm CPUs shouldn't be exposed to more than 1.25v - 1.3v and temps shouldn't go above 65c. CPUs are supposed to be robust dammit

[andressergio]

Re flash your bios and put SAME SETTING by hand most bios get corrupted on time and start doing crap
Cheers
Sergio

[Ket]

I've reflashed the UEFI multiple times, my CPU really is half dead. I also tested the CPU on 2 other boards that have just sat in their box for months so definitely nothing wrong with them and the CPU did exactly the same thing. RIP 2500k