cancel
Showing results for 
Search instead for 
Did you mean: 

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

WillyK
Level 10
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!
2,641 Views
13 REPLIES 13

Deepcuts
Level 10
In the last 7 years the only component to ever fail me was a GTX 580 from Gainward, which was running at stock speed.
My old i7 2600k @ 4 Ghz still going strong.
My old i7 4770K @ 4.2 Ghz still going strong.
My new i7 8700k...is too new to say.
Then again, I take care of my rigs. Clean them often and no coffee or other liquids (they R on diet)
So as you pointed out, if proper cooling then no problem, unless bad luck/crappy hardware 🙂

WillyK
Level 10
Thanks pal! Congrats for your healthy CPUs! Interesting that I had a Gainward card that died on me but it was before GTX 580. Perhaps not the best brand CPUs... Taking good care of the rig is a very sound advise, especially cleaning the dust in radiators etc. and maintaining properly the entire cooling system. High temperatures would definitely kill any hardware!

i recently sold my asus P6T Deluxe V. 2. still in great working order!

EGGPLANT303 wrote:
i recently sold my asus P6T Deluxe V. 2. still in great working order!


Yeah, Asus Deluxe boards are great! Never heard about a failing one. Congratulations 🙂 I kept my P4C800-E for over a decade just for fun and it was still working! Then the IDE drives died and I didn't care to revive it. This MB was incredibly robust and stable.

Korth
Level 14
I'm still running a pair of X99 boards with 22nm E5-Xeons rated at 3.8GHz but overclocked to 4.4GHz. Almost 4 years now, warranty was 3 years.
These parts could run stable but hot overclocks at 4.5GHz and 4.6GHz, I bumped them down a notch or two because I didn't want a furnace in my chassis.
My CPU temps never exceed ~70C, ~75C tops. Still going as strong today as when new, never had to retweak volts or clocks, not the slightest sign of aging.
My VRMs never get too warm or struggle because I have a robust Platinum PSU and only needed to bump voltages up a little, not a lot.

I used to have a twin pair of GTX980 cards, primary GPU overclocked to 1377MHz/1515MHz/7800MHz, other at EVGA stock 1291MHz/1393MHz/7010MHz.
Guess which one burned out after about 30 months and which one still works in my sister's computer.

One of my buddies overclocks hard on everything he's got, all liquid cooling all the way, even uses TECs to pre-cool his reservoir and rads. Very overkill but hey engineers need hobbies too, lol. He seems to burn through CPUs and GPUs and mobos and RAM pretty frequently, something major in his system fails (and needs replacing) at least once per year.

Our IT guys run everything pure spec, not a single overspec or overclock or undervolt to be found anywhere on site. High duty cycle, the hardware is kept hard crunching, but hardware failure is almost unheard of, most of the blades rotate out when they've finally just become too obsolete and gutless and inefficient, not when they break.
"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams

[/Korth]

WillyK
Level 10
Thanks Korth! So what you're saying is that OC could be almost as durable as a stock setup, as long as you don't take the OC to the limit. Logical, but how far can you take it without too much risk? I usually run heavy P95 loads (just initially for a short while), just to verify the OC stability, and then take the OC a couple of notches down (CPU, GPU, RAM) for 24/7. However, I've had to retweak some of the rigs after 2-3 years, and usually it's the CPUs that want a little more voltage to get stable again. But they are still working and I guess they are not dying, just wearing out a little. However, I had 3 dead GPUs the past 5 years, and these just die all of a sudden without a warning. Maybe I was unlucky. But I never had any RAM problems.

WillyK wrote:
Thanks Korth! So what you're saying is that OC could be almost as durable as a stock setup, as long as you don't take the OC to the limit. Logical, but how far can you take it without too much risk?

First of course: overclock or undervolt is out of spec, not operating within rated design parameters, no promises. No piece of silicon is perfect, no two pieces have exactly the same quirks and flaws. You don't know how fast or how long they can run overclocks until you push them to the threshold. The only guarantee is the manufacturer's guarantee, only 3 years and only when operated completely within rated spec.

The big killers of semiconductor components are thermal stress and electrical stress. It turns out that most damage is actually chemistry-based, galvanic and thermal conditions shift equilibriums, accelerate the separation of metals, the breakdown of crystal lattices, the corrosion or erosion of materials from within - this damage is molecular-scale, inevitable, gradual, and utterly irreversible. Some stress tests and burn ins and benchmark brag-n-swag is fine, even expected - it's the sustained long-term day-to-day punishment which adds up, driving a little beyond the speed limit is a lot less damaging than flooring the engine into the red at every intersection.

So use robust cooling to keep things as cool as possible, always well below rated maximum temps. And remember that the CPU package might report 80C while the CPU internals are being baked in a smelting pot.

Use highest-quality PSU, quality not quantity, because smoother and tighter and steadier regulation means fewer spikes and ripples and fluctuations get passed downstream to the mainboard VRMs, which in turn passes cleaner gentler power downstream to the CPU (and other stuff).

I always "graph" out my overclocking parameters, clocks vs volts vs temps, it's a curve which goes exponential at some point, some people push up to the maximum "stable" extreme, I prefer more modest overclocks somewhere in the "sweet spot" where performance gains outweigh energy losses.

Yes, I spend "too much" on my PSU and cooling, but then again I don't have to replace any hardware in my system for many years at a time. Other guys buy "proper-sized" PSUs and coolers, and they push their clocks more aggressively because rawr moar faster must mean moar teh pwnage (even if you actually suck at playing games) ... but they end with system instabilities and finicky timings and constant voltage tweaking and dead parts within a couple years.
"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams

[/Korth]

Korth wrote:

I always "graph" out my overclocking parameters, clocks vs volts vs temps, it's a curve which goes exponential at some point, some people push up to the maximum "stable" extreme, I prefer more modest overclocks somewhere in the "sweet spot" where performance gains outweigh energy losses.


Can you please give an example of such a "graph" and explain how you produce it (the steps you take etc.)? Is it a single graph for each component, or do you combine the graphs for several components on the same sheet? Can you please also indicate where you mean is the "sweet spot"? Thanks!

I believe that sharing specific practical tips and solutions would be very useful to the community.

zoran
Level 9
Last seven years asus maximus iv gene-z and i2500k (oc 4.6ghz) still going strong and thats even after at least 100 power blackouts in bangkok where I live.
sold original gtx 580 , bought a gtx 780.

Just bought a new i2500k on ebay and will swap my coolmaster fan for a corsair water cooler and try to get to 5.0 ghz

I see no reason for a full upgrade yet.