Reduced Overclock Potential?

[StopSign]

I built my system in January and ran it at 3.19 GHz (456x7) for over 4 months. When summer started, I turned down the overclock to around 3.0 GHz and have been running that for about 2 months now. Last night, I tried setting it back at 3.19 GHz and it would not pass 1 minute of Orthos. All of the settings were exactly the same as the original 3.19 GHz settings. It was also on the same bios.

The only major hardware change during the 3.0 GHz period was replacing my OCZ Platinum (Elpida) with Buffalo Firestix (D9GMH). I'm aware of some of the compatibility issues with D9, but it ran perfectly fine while I was at 3.0 GHz. Memory stability is definitely out of the picture here.

I flashed the bios to F12 last night just to completely rule out the memory issue, since the major D9 issues have been resolved in bioses after F7. It would still not pass Orthos at 3.19 GHz. It's like there's some kind of FSB hole between 430 and 460 now.

I can't think of anything reason for this aside from undervolting. While I was at 3.0 GHz, I undervolted the chip even more from 1.22v to 1.20v. Yes, I was stable at 3.19 GHz with only 1.22v. Does the CPU "adapt" to lower voltages, thus reducing the clock potential?

 

[imported_miahallen]

Interesting, I've had the same issue with my Opteron 165. When I first got it (1.8GHz @ 1.35V stock) it would run all day at 2.7GHz with stock voltage. About 9 months later, I was having stability issues, and I ran Orthos again to get errors. I ended up having to back off to 2.5GHz, but was able to undervolt it to 1.25V....so I'm OK with that, it runs nice and cool. I was never able to get it back up to 2.7GHz though, not even with a new OCing friendly mobo. Not sure what the deal is.

 

[CTho9305]

My reply from this thread:

Originally posted by: CTho9305
<blockquote>quote:
Do cores just always gradually burn out after a year or so of heavy overclocking?</blockquote>
Read up on:
electromigration - weakens wires, so they conduct less well, slowing things down
hot-e effects - hurts transistor performance
nbti - hurts transistor performance

...and of course the simpler ones like dusty fans, dried up thermal paste, etc.

 

[Nathelion]

I think processors do reduce in OC potential over time. Electromigration or somesuch.

 

[Smartazz]

I had the same problem with my AMD X2 processor. It happens I guess.

 

[StopSign]

I see. Oh well, I can't complain too much about my current overclock. Runs nice and cool.

 

[IEC]

Originally posted by: CTho9305
My reply from this thread:

Originally posted by: CTho9305
<blockquote>quote:
Do cores just always gradually burn out after a year or so of heavy overclocking?</blockquote>
Read up on:
electromigration - weakens wires, so they conduct less well, slowing things down
hot-e effects - hurts transistor performance
nbti - hurts transistor performance

...and of course the simpler ones like dusty fans, dried up thermal paste, etc.

That sums it up pretty well. In my experience the dust bunnies really hurt cooling performance and can make a previously stable clock unstable. In any case an eventual decrease in clocking potential is almost unavoidable due to the aforementioned EM/hot-e/NBTI. Think of it as "wear and tear"