how to damage a cpu?

[magictorch]

Hi, i was wondering how much damage overclocking a cpu does when you hold it at stock volts and keep the temperatures within recommended levels. I can overclock the opteron 165 to 2.7 or 2.9 stable on air stock volts-load, 42-44oc but how long will it last if i keep doing this?? if it lasts another year thats great , anything less is not good. Would the c2duos last longer when overclocked? hope you experts can help, cheers.

[b1lk1]

I would dare say you shave off about 6 months of life from punishing a CPU when you keep it under 60C full load and with a reasonable amount of Vcore. 6 months of life taken off of a 5-7 year life span. Chances are you'll kill the memory or motherboard long before the CPU. At stock voltages, I cannot even fathom that you do any harm at all to the CPU when overclocking.

[magictorch]

if the memory is not heavily overclocked, will a heavily overcloked cpu the memory? I noticed the memory was limiting the overclock. It runs at DDR500 or 5201T easily on lesser overclocks but when i push the cpu past 2.8 i have to run it at around DDR425-1T. That would be a pain to brake the memory or MB due to CPU overclocks. cheers,

[b1lk1]

Those parts (memory and motherboard) only break under their own stress, not by just ocing the CPU. Motherboards tend to break from high MHz speeds and memory tends to die from running it overvolted.

[magictorch]

I am running the MB a little higher from reference clock 200mhz to 323mhz to overclock the cpu 9xmulti..hopefully it will last another 8months or so till i upgrade...thanks

[pscout]

I would not claim to be an expert, but running a 724 folding farm with every rig oc'ed to the max stable folding speed:

- I have killed a couple in pentium D cpus ... and i am pretty convinced it was by running pretty high volts over a sustained >1 month period folding (ie 100% load) 724 after doing a few very high initial short oc/suicide tests. These were both initial retail steppings.

- I have also killed some lesser mobo's that couldn't provide very long term high power draws through their design (eg p5ld2's) - these were folding at max stable oc on smithfield pentium D's ( 830D's), 100% load for > 1 year 724. The cpus and memory went on to fruitful (OC'ed) lives in other mobos.

- I may have killed 2 sticks of ballistix memory through higher than normal volts but I have treated other memory (ocz, corsair) with same memory chips much more harshly without problem so I think it is weak pcb manufacturing or design by crucial that lead to these failures. I treat their rma replacements more gently (ie i don't run them with max mobo volts to get more speed/tighter timings out of them).

With the higher energy costs for running and cooling them, I now try to find the oc'ing sweet spot where minimal voltage increases yield a good oc and where more v doesn't increase the oc much. I used to try and squeeze every last bit out of every component but measuring the power draw for every oc'ed rig helps me find the sweet spot. CPU voltage or vid card v oc'ed have the only significant impact on power draw. Other mobo voltages including memory are insignifcant to draw but do affect temps/stability and potentially longevity of mobo and affected components.

Also, since i primarily build for folding, max stable oc's for folding are usually a lot lower than a stable oc for the 'normal' benchmarks. No errors are tolerated when folding, and are especially evident with the new SMP folding clients. I can run 24 hours or more on othos or 4xP95 yet error fairly quickly when folding.

Also, some mobos are just tanks and take any abuse i can throw at them. The best ones i have had so far are the p5wd2 family from asus. Sadly, soon to be retired as i replace 930d's @ 4-4.2 ghz for quads. These are the mobos that i killed both a 920d and a 930 d with and are still 100% stable.

Assumption throughout all this rambling is that temps are kept within normal/stable ranges via extra fans, upgraded HSF's, watercooling, etc.

[magictorch]

thanks for that seems like youve had your fairshare of burned out bits..ive been trying to resist the temptation of raising vcores as i know it will end in tears..id love to hit 3ghz, can get 2.93on air stock volts. But id like to see what the voltage draw is on the current overclock.....hoping the MB holds out till the new year anyway... when the new dx10 games eventually surface im going to give that a go with an 8800GTS 6**mb and if the systems not up to it ill move on to a new platform...i wonder if these crazy pricewars will run into the new-year i cant see how AMD can break the intel stranglehold-their relentless...

[pscout]

A conservative rule of thumb on intel cpus is to keep within 10% of max spec cpu v. Most oc'er exceed this including me but i like to stay < 20-25% for power cost, temps and longevity. Intels usually often have several oc'ing walls where to go past a particular fsb, you need to add ~ .15-.25 V. The higher the fsb the closer it gets between need v bumps. I would expect the amd's to be similar. I seldom oc past more than 1 or 2 of these walls anymore.

If you are almost at your oc goal at stock V I wouldn't be afraid of giving it a bit more to get there as long as your temps are good.

I think one thing that may hurt longevity more than high V is frequent on/offs.

[magictorch]

thanks pscout, just noticed your last comment!!!...i think i may push the vcore within the ranges you mention. one thing thats weird is that the mb is set to 1.425 on auto but cpuz says 1.44?? anyway will try for 3ghz but not expecting miracles. Am thinking about folding but use this system for work-running optimal energy allocation scenarios...like the idea of building proteins though!