Long term overclocking with 100% cpu 24/7

[plonk420]

not sure if this is a better place to ask opposed to the distributed computing forum, but here goes:

so, my noob ass likely fried a pretty much new X58 mobo*. best i can figure is that it was either "default settings" for everything but vCore .. or just continuous use at a high BCLK, i'm not sure. this is my first time OCing.

anyone running distributed computing (or something else to load it 100%) 24/7/365 on an OC? if yes, what percentage of an OC?

only reason i ran most of the settings at "default" was that the "default at stock" voltages (other than CPU) seemed to not post. i can't say i made like 50-100 attempts at combinations, but i did do ~20 or so.

probably the worst i did to it was running a 920 at 1.48v for a week on a distributed project. i pulled back to ~1.38-1.40v after that week.

temps were ~50-60 for "CPU" but cores were reading as high as 80C (when i'd left for work, it was *only* 70C). after that week of hell (at least for the hardware), i pulled back to temps that maxed at ~72-74C


*black screen for post, no beeps, nothing at all. 1 stick of mem (tested all 3 sticks 1 at a time), diff power supplies, diff GPUs, no other hardware plugged up, cmos cleared, all power sources removed, incl battery.

 

[joutlaw]

I've been running a E4300 @ 3.0Ghz which is a 66% OC for around 2 1/2 years.

I don't do distributed computing, but I leave my PC on 24-7.

I'm using a gigabyte GA-965P-DS3 mobo and it's been pretty solid.

What case are you using? I have the Antec Nine Hundred which has a lot of fans... 4 -5 120mm and 1 250mm on the top.

 

[SickBeast]

I've done that before on a wide variety of overclocked processors. So long as you make sure it's Prime stable for 24 hours, you should be fine. Most people consider 8 hours acceptable. Most bad overclocks will fail within 1 minute of running Prime95.

Based on your temps I would say you should look into a better HSF for your CPU.

 

[ShawnD1]

Originally posted by: joutlaw
I've been running a E4300 @ 3.0Ghz which is a 66% OC for around 2 1/2 years.

This really pisses me off. I paid probably $100 more than you did in order to get the E6600, and it hits the same 3ghz wall. Lesson learned, I guess. The binning means almost nothing.

Mine's an E6600 at 3ghz, that's about 25% overclocked. It has been doing folding on both cores (2 single-core clients) since late 2006 and it's still going strong. I do prime95 tests every now and then just to check up on it and it passes every time. All voltages are auto, memory timing and frequency are auto.

Doing folding on an overclocked processor shouldn't be a problem at stock voltage. Increasing the voltage is very hard on the processor and probably shouldn't be done if you're going to be running the CPU at 100% all day and night.

 

[ochadd]

This post is going to bounce around a bit.

Im running SETI and GPU grid on the system in my sig. Using an Asus P5ke wifi the CPU running 1.4 vcore @ 4.2ghz e7400 is rock solid but I choose to run at 4ghz so my core temps never go above 60c.

On the same motherboard I had an e6750, stable @ 3.6ghz 1.45 vcore with 1700mhz FSB, running at 3.4ghz long term, same 1.45 vcore, and is still running perfectly after three years in the machine I'm typing this on.

Warning though from past experience. My AMD X2 3800+ running 2.3ghz vs 2.0 stock deteriorated over the course of three years to completely lose the ability to OC and was only stable under clocked to 1.8 even though temps were unchanged. Before that an AMD 64 chip went unstable on me after two years OCed hard but I can't remember the numbers. All of my chips run BOINC 24/7.

If you overclock to the max the chips will not last long. Find the max, get the verification and happy screenshots, then dial it back a bit if you want them to last. It may be all in my head but giving up that last 5% seems to get me a chip I can resell in the end. Yet to fry a motherboard.

 

[dbcooper1]

Originally posted by: plonk420


probably the worst i did to it was running a 920 at 1.48v for a week on a distributed project. i pulled back to ~1.38-1.40v after that week.

temps were ~50-60 for "CPU" but cores were reading as high as 80C (when i'd left for work, it was *only* 70C). after that week of hell (at least for the hardware), i pulled back to temps that maxed at ~72-74C


*black screen for post, no beeps, nothing at all. 1 stick of mem (tested all 3 sticks 1 at a time), diff power supplies, diff GPUs, no other hardware plugged up, cmos cleared, all power sources removed, incl battery.

Too much voltage; likely did some latent damage running that long at that voltage and temp.

 

[plonk420]

Originally posted by: joutlaw
What case are you using? I have the Antec Nine Hundred which has a lot of fans... 4 -5 120mm and 1 250mm on the top.

case is Antec 300

Originally posted by: SickBeast
Based on your temps I would say you should look into a better HSF for your CPU.

dark knight. probably perfect for stock speeds...

Originally posted by: ochadd
Warning though from past experience. My AMD X2 3800+ running 2.3ghz vs 2.0 stock deteriorated over the course of three years to completely lose the ability to OC and was only stable under clocked to 1.8 even though temps were unchanged. Before that an AMD 64 chip went unstable on me after two years OCed hard but I can't remember the numbers. All of my chips run BOINC 24/7.

If you overclock to the max the chips will not last long. Find the max, get the verification and happy screenshots, then dial it back a bit if you want them to last. It may be all in my head but giving up that last 5% seems to get me a chip I can resell in the end. Yet to fry a motherboard.

Originally posted by: dbcooper1
Too much voltage; likely did some latent damage running that long at that voltage and temp.

good to know.

i THINK i've been doing DC for 5-8 years. i've had 2 mobos die just from wear and tear... (well, athlon XP boards, so insert theories there; my opinion was probably 60% heat (occasionally let dust bunnies build up), 40% capacitor uncertainty. they were ~5-7 years or so old)

i haven't "believed" in OCing up til now, and this will probably put me back in my original mindset.

either that or stock voltage at a mild 21x140/147. it's an agonizing decision >_<

Originally posted by: plonk420
temps were ...

the weird thing is that 2 "cores" would maybe register 68C while 2 other cores would register 72C

 

[joutlaw]

Originally posted by: ShawnD1

Originally posted by: joutlaw
I've been running a E4300 @ 3.0Ghz which is a 66% OC for around 2 1/2 years.

This really pisses me off. I paid probably $100 more than you did in order to get the E6600, and it hits the same 3ghz wall. Lesson learned, I guess. The binning means almost nothing.

Mine's an E6600 at 3ghz, that's about 25% overclocked. It has been doing folding on both cores (2 single-core clients) since late 2006 and it's still going strong. I do prime95 tests every now and then just to check up on it and it passes every time. All voltages are auto, memory timing and frequency are auto.

Doing folding on an overclocked processor shouldn't be a problem at stock voltage. Increasing the voltage is very hard on the processor and probably shouldn't be done if you're going to be running the CPU at 100% all day and night.

Yep it's a good little chip. I've been wanting some more performance since I got a 4890, but it's going to have to do for now.

I keep the chip at 1.40V in the BIOS ... I don't believe I've ever seen it go higher than 42c under full load. It idles around 30c, but I have a lot of air moving through case.

I may try F@H since it installed with my ATI software and see what my temps get too.

 

[aigomorla]

check your ram.

thats usually the first thing that dies in a failed DC system.

 

[plonk420]

even if i never hit the rated speed of the ram? (DDR3-1333)

 

[Markfw]

I have all the rigs in my sig 24/7/365@100% at the listed speeds. However, I have cooling that never allows any temps over 60c at any time. Most are 55c or less.

 

[aigomorla]

Originally posted by: plonk420
even if i never hit the rated speed of the ram? (DDR3-1333)

the act of loading and putting voltage though it, is part of the damage.

its not just voltage and speed.

Even if you ran ram stock, kept it under full load, you are pushing more voltage then someone who has faster clocked ram at higher voltage idling.

DC computing in general stresses out a lot of computer parts, thats why the guys who do DC computing, and have high overclocks gets my upmost respects.

In order of things on how they die is as follows:

1. its always first the ram.
2. The board
3. The cpu.

 

[Zenoth]

I've had my E8400 at 3.8Ghz and 1.32v for a good eight months non-stop, 24/7, only rebooted the system when necessary such as re-installing drivers or games that needs rebooting, but other than that it was stable enough. But I suspect there's been some degradation, because I need 1.408v for the exact same settings to be stable now. It's part of the risk of over-clocking anyway, even if it's rare, if it needs too much voltage (it already needs way too much for 3.8Ghz, but hell, it's a bad OC'er that's it, it happens) then I'll just buy myself a E8600 until the next Intel architecture (not refresh) is released, that's when I'll jump in the Quad cores wagon.

 

[Markfw]

Originally posted by: aigomorla

Originally posted by: plonk420
even if i never hit the rated speed of the ram? (DDR3-1333)

..snip...

DC computing in general stresses out a lot of computer parts, thats why the guys who do DC computing, and have high overclocks gets my upmost respects.

...snip...

So you mean me ?????

 

[yh125d]

Originally posted by: Markfw900

Originally posted by: aigomorla

Originally posted by: plonk420
even if i never hit the rated speed of the ram? (DDR3-1333)

..snip...

DC computing in general stresses out a lot of computer parts, thats why the guys who do DC computing, and have high overclocks gets my upmost respects.

...snip...

So you mean me ?????

Probably. No matter how many times I see your sig I still go "oh wow" at your rigs

 

[faxon]

Originally posted by: yh125d

Originally posted by: Markfw900

Originally posted by: aigomorla

Originally posted by: plonk420
even if i never hit the rated speed of the ram? (DDR3-1333)

..snip...

DC computing in general stresses out a lot of computer parts, thats why the guys who do DC computing, and have high overclocks gets my upmost respects.

...snip...

So you mean me ?????

Probably. No matter how many times I see your sig I still go "oh wow" at your rigs

yea i have to say the same on that one lol. why are you running so many of them, other than the fact that it's for a good cause, anyway? you one of those guys that knows someone with a disease that they are researching? i saw an article a while back where some guy's dad got diagnosed with cancer right around i7 launch, so he bought like 30 GTX285s and built i7 systems around them to run F@H 24/7/365 because they were doing research on the type of cancer he had been diagnosed with on the F@H network

 

[aigomorla]

Originally posted by: yh125d

Originally posted by: Markfw900

Originally posted by: aigomorla

Originally posted by: plonk420
even if i never hit the rated speed of the ram? (DDR3-1333)

..snip...

DC computing in general stresses out a lot of computer parts, thats why the guys who do DC computing, and have high overclocks gets my upmost respects.

...snip...

So you mean me ?????

Probably. No matter how many times I see your sig I still go "oh wow" at your rigs

Yeah your definitely one of them with the power bill your forking over each month.

But if u look at raw computing power, i think i top ya, but i dont DC anymore.

3 i7's 975 + W3580 + W3570 are clocked at 4.4ghz with HT on.

my 965 is at 4.1 and my NAS box aka sammy is at 2.0.

so that makes a total of 36 threads at my disposal.

 

[Markfw]

Well, I have 38 cores, not counting GPU power (22 cards) or the X2 that I don't use, or any of my 4 laptops, but yours are better (the I7's)

 

[aigomorla]

well as i said mark your forking over your cash on power bill. Probably enough for you to buy several computers each yr alone.

That alone deserves respect!

 

[plonk420]

yeah, i haven't looked at any threads, but i'd assume i7s are the most efficient. most points per watt. as soon as i get mine back up and running, i'll cease operations on my others (except maybe my phenom ii based one .. and probably only on occasion like the cold cold winter).

 

[ShawnD1]

Originally posted by: plonk420
yeah, i haven't looked at any threads, but i'd assume i7s are the most efficient. most points per watt. as soon as i get mine back up and running, i'll cease operations on my others (except maybe my phenom ii based one .. and probably only on occasion like the cold cold winter).

If your DC project of choice supports graphics cards, a higher end graphics card gives the most points per watt. An 8800GTX video card does about 4x as much processing as a Phenom 9600. Even a $300 CPU gets its ass kicked by a $100 video card.

 

[plonk420]

separate from this thread, i just was (trying) to look up points per watt, but was only getting results for GPU-supported projects...

i guess i'm mainly looking for non-GPU projects. mostly medical (wcg, rosetta) , with a small diversion in to the historical Enigma@home and CG-related BURP...

 

[Markfw]

You will get much more informed responses if you post in the Distributed computing forum.

 

[VirtualLarry]

Originally posted by: faxon
yea i have to say the same on that one lol. why are you running so many of them, other than the fact that it's for a good cause, anyway? you one of those guys that knows someone with a disease that they are researching? i saw an article a while back where some guy's dad got diagnosed with cancer right around i7 launch, so he bought like 30 GTX285s and built i7 systems around them to run F@H 24/7/365 because they were doing research on the type of cancer he had been diagnosed with on the F@H network

atlasfolding?