How durable is an OC build? What is your experience?

[WillyK]

Hi guys. I just wonder... What is your experience with aging OC components, i.e. when your rig is getting older than 2+ years?
Which is first to fail? CPU, GPU, RAM, MB... and how do you prevent this from happening? (typically in terms of a BSOD 124)

I suppose most of us expect our builds to last for more than an year or so. My personal experience don't point out any clear "winners", and it partly depends on how much you OC different parts (assuming proper cooling) and of course, of the quality of the parts. But I believe it's still a valid question. Are there certain parts more exposed to a failure and what could be a clever strategy to prolonging the life of your OC rigs as much as possible before having to start replacing parts? Sudden BSOD 124 on a long-working rig is usually a good indicator that your HW is starting to fail. Of course, BSOD 124 won't tell you much about which component is failing, so you are back to pulling your rig apart trying to identify which part is the sinner etc... Not much fun!

So I hope that you would find a few minutes to share your valuable experience about aging OC components and your approach to preventing/fixing such issues on the longer run.
Many thanks!

[WillyK]

Thanks again! Best PSU is a truly great advice. I've had PSU troubles twice and those cause very nasty errors! I follow the rule that your PSU should be never loaded more than 80% (at peak!) and I buy only Platinum class, top brand PSU. Yes, it's expensive but not a lot, and compared to the risk reduction and well-being of your rig is definitely worth the bucks!

Finding the sweet OC spot for 24/7 is the ultimate goal. I just wonder if you do this by component, i.e. for each component individually? I mean the combination of OC on CPU/GPU/RAM/etc. influence each other, and this is the hard part. How do you find the best configuration everything included? Do you have any specific tips about this?

[Korth]

Good overclocking always isolates, measures, and tests one component and one parameter at a time. You push it faster by an increment, if it's stable you push it faster another increment, if it's not stable you bump your voltages an increment, if it's stable you move forward if not then backward. A lot of people keep pushing until it finally can't get pushed any further, then if it worked once they assume it's stable and start fiddling around with other settings to tweak away their BSoDs, crashes, lockups, boot fails, and error msgs. They also suffer from weird seemingly unrelated issues like malfunctioning USB or excessively long boot times, etc. I recognize that no overclock is truly stable and any faults of any kind indicate some kind of hardware instability, so I isolate and test and confirm/deny the cause of the fault, I pull back when necessary. I also stop the process when I see temps ramping up, that last few hundred MHz (which is like ~2% or so performance gain) just isn't worth ~100W more power and ~20C more heat and maintaining more finicky finely balanced tweaks.

It's nice to push your parts right up to their limits, see what they can do, feel happy, play with your toy, brag a little. We all do it. But I usually buy computers with the expectation that they'll continue working "perfectly" for at least 2-3 years, really I expect them to keep working indefinitely until I decide to upgrade. Aggressive overclockers (and reviewers) replace their hardware frequently, literally every time the next New Thing comes out, they're far less concerned about the impacts of overclocking on longevity because they only brush the dust off last-week's tech a few times a year for comparison purposes.

[WillyK]

Korth wrote:
Good overclocking always isolates, measures, and tests one component and one parameter at a time. You push it faster by an increment, if it's stable you push it faster another increment, if it's not stable you bump your voltages an increment, if it's stable you move forward if not then backward...

Sure, I get all this. I've build a few OC rigs in my life too. The question is about this "sweet spot" you mentioned before and how to get there (not above, not below). I agree that components' OC capabilities do vary from sample to sample, especially CPUs, but those differences are minor in the big picture. All electronics is manufactured within strict standards and specifications.

My hypothesis is that some components tend to be more robust than others and I was hoping that sharing relevant experiences on this topic across builders would give a better picture of which components tend to be the most vulnerable, and which can take a lot of heat before they crack... or something in this alley 🙂

[Korth]

Well, a sample graph (not mine, from this article, just meant to illustrate the general idea) -
https://images.anandtech.com/doci/5763/Temperature.png

On this part the "sweet spot" is 4.5GHz. The point where things stop being linear and start increasing geometrically.
This part can actually overclock up to 4.8GHz but doing so requires a voltage bump from ~1.1V to almost 1.35V and increases temps by ~28C.
To my mind >4.5GHz places ever-increasing strain on this part, pushing the silicon beyond its "optimized" capabilities.

Another article about overclocking vs CPU longevity here. Basically what I said but supported with fancy diagrams and graphed data.

I should mention that high-quality mobos with strong VRMs extend overclocked component lifespan. Hard to say how much impact this has, but I've seen Supermicro workstation boards with server-grade VRMs outlast low-end gaming boards with lesser components. It turns out the high end ASUS ROG stuff (basically anything TUF and any mobo with Q-Code display, lol) tends to use very similar (essentially identical) industrial-grade components as high-end server boards, though I'm a little apprehensive about the low end ASUS stuff (like STRIX mobos) offering the same long-term uptime and extended component longevity.