Hidden dangers of over clocking.

[UltraTaco]

Few days ago, I was bored and decided to do an overclock. Plugged in numbers and after few trials, found what I thought was a stable overclock, or was it?

Cinebench r15 passed clean, so did p95 small and blend runs, next was occt small and large data sets. I ran tests half an hour-1.5 hours. Everything checked out. I figured it'll do for me for now. I'm not doing anything intense anyways.

This evening, I went into graphics configuration and changed color range for video playback from 16-135 to 0-135 and clicked apply. It took longer than usual to apply and in the end, I got a BSOD!! 101 error, but....but I wasn't even encoding or playing!! It was just a simple settings change!!

How could it be?! Yes OCF men, this can happen to anyone, even yiu!

Be safe and run those tests longer! Better yet, don't push your overclocks hard!

 

[Alaric]

The real hidden danger is the need for more MHz. You get antsy, it takes more megs for that same high. Before you know it you're broke, have a stack of motherboards, and you're asking strangers on the internet "Hey, man. You got any chips?" Then someone like johan45 says "Yeah, I got chips, mobos, what you need brother?"

"Got any of that AM3+? I need some big numbers, man!"

Now you're hooked and there's no going back. You'll spend a week on RAM for a few MHz. Just say "YES!" to overclocking! LOL

 

[UltraTaco]

Rehabilitation Center
There's hope. Freedom. Remission.

 

[Mr.Scott]

Resistance is futile. You will be assimilated.

 

[ihrsetrdr]

It was just a simple settings change!!

You have not only altered your time continuum(from which there is no return) you have also altered the molecular structure of your spatial coordinates, causing an inter-dimensional rift.





The only hope for redemption is to configure your system stable. Maybe relax those timings a bit....perhaps bump that voltage a tad.

 

[Alaric]

 

[Tekred]

"Got any of that AM3+? I need some big numbers, man!"

I got some of them AM3+ boards and them fx chips whatcha need (as I scratch the neck and look around to see if anyone is watching)

 

[Woomack]

when I'm reading threads like this then I'm not sure what to say ...

 

[EarthDog]

I typically dont post in hopes it will just go away.

I mean, it tells us is when overclocking there is instability. Which every soul here knows. Amd then says not to push limits. Lol, son this is overclockers.com... youve decided to hole up in the wrong place if giving advice like that out!


Par for the course for taco...

 

[caddi daddi]

it sounds like the idle vcore sagged so give it a bump.

 

[Alaric]

I got some of them AM3+ boards and them fx chips whatcha need (as I scratch the neck and look around to see if anyone is watching)

On a serious note,
Have any quad core FX? They don't seem to have ever been real popular and I was wondering if points might be easier to come by with them. And I'm a fan of the road less traveled. I'm still trying to get familiar with the ins and outs of competitive benching. As opposed to my current "Oh, look! Magic Smoke!" method that seems get the girlfriend a little unhinged. LOL

I typically dont post in hopes it will just go away.

when I'm reading threads like this then I'm not sure what to say ...

You guys really don't want to go near the Flat Earth thread, then. Not after what I just posted.

 

[UltraTaco]

Well, I should explain. This is a public service announcement! Sometimes someone might come in and say, "hey my system crashes opening screensaver options or saving a word file", and then you ask, "well is it over clocked?", And someone goes, "yeah dude, it's at 6.3ghz rock stable, I tested with prime 15 minutes", and you go, "oh okay, it's probably not stable".

It's all in the name of:
Unstable overclock can do wonders things and can cause chaos under light loads as well.

 

[mackerel]

You're not in too deep until you start cutting your own mixes of liquid metal...

I do find it difficult to prove something is stable, perhaps because we can't... we can only say how confident we are that it isn't unstable. You are only stable until you are unstable. When prime number hunting with my first 6700k, I had all sorts of problems with it, eventually reaching error rates of about one detected every 3 months. Normal people wouldn't have noticed it. Is that stable? Is that not stable? Before anyone puts it down to "one of those random software things", my Haswell or earlier era systems didn't error at all. Other Skylake and newer systems were ok too, it was some combo of that particular CPU sample, on that mobo, in that system.

For any 24/7 OC, I typically now find what I think is the best balance between voltage and frequency, and once I'm happy it seems stable, I reduce the clock another step for additional headroom. If I still get instability I might even go as far as to run stock. Yes, I said it!

 

[Alaric]

I might even go as far as to run stock. Yes, I said it!

Oooh! You said the "S" word! I'm telling!



I ran my RAM at stock speeds for years after some bad experiences with PNY memory murdering my boot.ini every time I bothered it. Gotta do what you gotta do.

 

[mackerel]

I ran my RAM at stock speeds for years after some bad experiences with PNY memory murdering my boot.ini every time I bothered it. Gotta do what you gotta do.

I have to say, I find ram OC far more scary than CPU. If you push a CPU too far, you generally get pretty obvious crashes or otherwise easily detectable errors. My limited experience with ram OC is that even if you throw a mix of tests at it, it is a higher chance of it coming back to bite you later. Ram errors, unless very frequent and severe, can more easily go undetected.

Like my manual timing B-die attempt... it was aida64 stable, memetest stable, I could run Prime95 small/large/blend until the cows came home. But when running actual prime number tests (they are double checked as routine) I got varying degrees of error rate. There must be some kind of access pattern where it wasn't stable, yet other tests didn't find it. One of the differences might be in Prime95 itself, as the stress test in that tends to run multiple single-thread tasks. I was running a single multi-thread task instead.

 

[Alaric]

And unstable RAM can bork your entire OS install, and for some reason it seems to take it out on boot files. That's why I love independently bootable memtests now. I can go in the BIOS, screw things up, and not have to expose my OS to vindictive memory ICs.

 

[EarthDog]

LOL, the IC's...

It's the settings users input that are the issue.

 

[Alaric]

If the ICs don't like those settings they seek out your boot.mgr, beat it up, and take its lunch money. Then its afraid to leave the house and *POOF*, boot failure and it's borked. True story, bro.

Just a little harmless anthropomorphism.

Seriously, I know it's the settings. And in my case, I wasn't aware that PNY wasn't the best stuff to try it with, so ignorance and the wrong tools for the job conspired to teach me a lesson. Well, several lessons, as I was a slow learner. LOL

 

[cchase33]

The real hidden danger is the need for more MHz. You get antsy, it takes more megs for that same high. Before you know it you're broke, have a stack of motherboards, and you're asking strangers on the internet "Hey, man. You got any chips?" Then someone like johan45 says "Yeah, I got chips, mobos, what you need brother?"

"Got any of that AM3+? I need some big numbers, man!"

Now you're hooked and there's no going back. You'll spend a week on RAM for a few MHz. Just say "YES!" to overclocking! LOL

LOL

 

[Johan45]

If the ICs don't like those settings they seek out your boot.mgr, beat it up, and take its lunch money. Then its afraid to leave the house and *POOF*, boot failure and it's borked. True story, bro. LOL

OS is child's play, when really pushing, especially in Win10 I can take out my BIOS. Also have a couple boards with a blown RAM channel upstairs somewhere they still work but single channel only .