Came home to a dead computer - MB/GPU lights on but hitting the power switch did nothing. Trouble shooting revealed that it would power up with the GPU (GTX780) out. Would also power up with an old 7600GT so I don't think it's the MB.
Note that during these tests there were no peripherals and I didn't let it boot completely, just enough to know it was working and then I would hold the power switch till it turned off.
After setting it back up and using the integrated graphics (i7-4770K) I got an error message saying "Overclock Failed" and hit F1 to enter setup or something like that. Went into the UEFI but couldn't think of any reason why the OC that I've been running for several years would fail so I Exited without Saving.
Well it seems to have reverted everything back to stock anyway - CPU is at stock clocks and the RAM is at default settings instead of the XMP profile.
Not going to mess with it right now, just wondering if anyone has heard of this before.
Running a chip at higher voltages and frequencies will gradually degrade the silicon. After years of running it as you have, the silicon can no longer support your prior overclock at the prior voltages. You can either raise the voltages or lower the frequency. Lowering the frequency and maybe the voltages as well would be advisable, since the chip has obviously started to fall off. This is normal behavior to observe from a chip that has been overclocked and pushed for years, this just comes with overclocking. I had the same thing happen with an FX 8350 that I pushed too hard, and even my graphics card overclock became unstable after prolonged use. It just happens because silicon doesn't hold up forever.
That is an unknown at this point. I haven't gone back in to try and re-do the OC.
I thought maybe you couldn't clock as high when using the integrated graphics or something. It was only running at 4.2GHz (my top OC was 4.7 lowered it to keep the temps down for the 24/7 BOINC crunching).
I will try re-doing the OC eventually and post my findings.
Sucks that a $500 GPU only lasted 3 years and 5 months (and I didn't even OC it).
I've had systems not work for completely unknown reasons when OCing and I take it all apart again and put it back together and for some reason it suddenly works. Also with a system that has been together for a long time. Probably is electron migration but hey, you never know and it's totally worth just rebuilding cause it won't cost you anything, and if it doesn't work, it doesn't work, and it was probably a good time to blow the dust off everything anyway....
Given the history of constant OC on CPU (aforementioned CPU lifespan reduction) + enabling additional heat and power usage on the same die, would be for me, enough explanation for your situation after removing GPU.
Also when there is a component changer in the comp, the UEFI/BIOS detects it and does longer than "fast boot". The normal boot actually does more testing which simply might detect issues that were very close appear in your normal work on that comp.
One more thing. I think, in general and regarding current CPUs/chips, upping voltage by some % contributes more to the lifespan reduction of microelectronics than the same % of GHz change. Especially at some point of voltage value, the negative effect (also occurring as temperature) gets more and more apparent (the effect is exponential).
Strange? - the overclock came back on it's own. I normally keep the computer running 24/7 so it's most likely not been restarted since the "Overclock Failed" warning. I'd been reluctant to go back in and try re-doing the OC since it's been running (relatively) quiet while crunching BOINC projects, however, I got my new monitor today and rebooted while trying to figure out why I wasn't getting any signal. Turns out I needed to switch the input in the OSD since I'm using HDMI off the MB until the new GPU arrives and it defaults to Display Port.
So it looks like it was just a hiccup either due to switching to the on-board graphics, multiple forced shutdowns while I was troubleshooting or a combination of both.
Isn't this issue irrelevant though if you have adequate cooling? I heard that it's the heat specifically that degrades the hardware. Not necessarily increased clock speeds and voltage. I've had my i5 4670k overclocked to 4.5Ghz since 2013 without having to change any multipliers or voltage. 3 graphics cards later and it's still running like a champ. I do have an absurdly large Noctua NH-D15 on it though...
Yes, and no. Heat outside the package can be easily dispatched because the whole chip package is relatively large to actual elements that produce that heat. I believe that the ratio of mass of active elements (generating heat) to the whole chip mass alone would mean that if package got hotter then those active elements got hotter even more.
There is also electro migration:
So good colling helps, but it pushes the negative effects only so far.
Depending on your buying habits it's mostly irrelevant regardless. The degradation being talked about is something like a 25%-33% reduction in the life of the CPU. These chips will easily run for a decade at stock speeds and most people who overclock will upgrade every 3-5 years so... it doesn't really matter.
Say what? Pretty sure you didn't up the BCLK to get to 4.5 and don't recall seeing anyone get 4.5 on stock voltage. Run a stress test and see what CPU-Z is showing for multiplier, bus speed and core voltage.
Haha no I meant that I haven't had to change my voltage or clocks since I overclocked my CPU in 2013. Sorry bad wording. And yes its stable, I run cpu-z tests every once in a while. I also use handbrake quite frequently to encode h.265
Ahhh, that makes more sense... and now that I went back and re-read everything I think I should have picked up on that to begin with. That's what I get for browsing forums at work.
