At what point does everyone stop OCing CPU?

[CurseTheSky]

The overall cost of the chip, and how much risk I'm taking.

If the chip is relatively expensive to me (like my i7 860), I'll generally see how far it goes on stock volts or with only a minor voltage bump. If it was cheap (like an E2180, Celeron, etc.) I love to ramp up the voltage and see how the temperatures do.

When I had my watercooling setup intact, I was a lot more daring. Now that my funds are going into other things (girlfriend, 5 year old, pets, classic car, etc.) I have less to spend on my computer hobbies and generally can't afford to replace something that blows. I can't even afford to get a new case, waterblock, and tubing to start watercooling again.

The last chip that I really put a decent overclock on was my E6600 - 3.6GHz, and I bought it back to about 3.4GHz for 24/7 stability. My E8400 sadly never saw 4.0GHz mainly due to lack of time and money.

 

[n7]

Usually i try for as high as possible & generally temps are what holds me back.

I then step a bit down from that to a number i like.

Or if temps aren't an issue, then i stop wherever it starts taking a lot more voltage for only a bit higher speed.

 

[soccerballtux]

I go until it breaks

 

[Ryland]

I OC'd my Q6600 to 3.0 Ghz (from 2.4) and stopped because it was "good enough". I may be able to push higher but don't have a reason to.

 

[Painman]

Sweet spot. I could squeeze another 200-300 MHz out of my CPU, but old Quads like mine really suck down the juice and bake after a certain point. I'm on water, but I value silence and simplicity more than absolute maximum performance.

That, and I don't really feel the need for more kick. Q6600 @ 3.2 is still keeping me happy for almost a year now.

 

[dguy6789]

I prefer to overclock as far as I can without increasing processor voltage and without overclocking any other components beyond their rated values such as any motherboard components or memory. Of course, I make sure it can run any of the most demanding stability tests and never crash. If there's any program anywhere that can make it crash, then I tone it down until that can't happen. The machine needs to be stable through all the seasons, be it winter time or summer time with the air conditioning turned off.

 

[MrK6]

Almost every CPU I've overclocked has a "sweet spot" that I've overclocked to using minimal voltage, after which the speed gained per voltage added ratio is not so good. Most recently, for my Q6600 it was 3.6GHz @ 1.375V BIOS and now my i5 750 @ 3.8GHz 1.33125V BIOS.

 

[MagickMan]

I push and tinker for months, but I usually try and stay within the CPU's voltage specs. That said, I can get my e7200 to 4.2GHz @1.45v, but it gets a lot hotter. So, unless I spring for a better HS (for some big $$$) I'm staying where I am @4GHz.

 

[Elfear]

Almost every CPU I've overclocked has a "sweet spot" that I've overclocked to using minimal voltage, after which the speed gained per voltage added ratio is not so good. Most recently, for my Q6600 it was 3.6GHz @ 1.375V BIOS and now my i5 750 @ 3.8GHz 1.33125V BIOS.

This.

Once the voltage-to-Mhz ratio starts climbing drastically I back it off a bit. I also try to stay within the old rule of thumb of stock volts +10% for good air cooling or +15% for good watercooling. System must be OCCT/Game stable for 6+ hours.

I usually enable C1E and Speedstep too so I don't worry about the processor running over stock when I surf the web.

 

[Tweakin]

I leave the vcore at stock, run the chip up till it has issues, back it down a little and give it one bump in vcore...normally very stable and heat is not an issue.

 

[bigriggg]

well i usually try to get a full Ghz out of every processor I get and leave it there. so i like leaving my q9550 at a nice 3.8, that should last a while. i like to leave my machine running 24/7 and dont want it to crash because of my OC. luckily i got the better stepping. i only notice a big difference on encoding and some gaming load time.

 

[BlahBlahYouToo]

Depends, I overclocked my i5 quite high but found it got quite toasty, so just leave it at stock and let the Intel built in OC kick in when needed. With speedstep enabled it sits at 0.8v in Windows, so a lot cooler and less leccy.

yeah, with built in turbo mode on 1156 i5/i7's, i'd leave it alone...
after a certain point, the instability, heat, decreased life span, headaches, voided warranties, extra power consumption isn't worth the extra couple hundred MHz.

seriously, do you notice a different going from 3GHz to 3.8GHz in day to day PC use?

 

[Elfear]

yeah, with built in turbo mode on 1156 i5/i7's, i'd leave it alone...
after a certain point, the instability, heat, decreased life span, headaches, voided warranties, extra power consumption isn't worth the extra couple hundred MHz.

seriously, do you notice a different going from 3GHz to 3.8GHz in day to day PC use?

It depends on what cpu you're talking about. Most i7 920s can hit 3.9-4.1Ghz so you're talking about a 1.1-1.3GHz increase, not just a couple hundred Mhz. I can definitely tell the difference.

 

[JimmiG]

Well, I haven't even overclocked my Ph-II 955BE yet - it's GPU limited by my 4850 in most games, and it's already fast enough for everything else... I might give it a shot at some point - it would help a little when playing flight sims or rendering audio from Ableton Live and Reason, for example. I overclocked my previous Phenom 9650 2.3 GHz to 2.7 GHz, which made enough of a difference to be viable.

There are a number of things I watch:
Temps: A few degrees C below the max specified by the CPU manufacturer at all times. This means even during the hottest summer months, and when running stress tests like Prime95 or OCCT. So for my Ph-II, I would have tried to stay below 60C under those conditions. Temps will usually be lower than that when running actual programs and games, increasing the margin.

Voltage: You can usually find "nominal voltage" range if you search for your CPU. For the Ph-II 955BE, that's 0.850V - 1.5V. So I would not have any worries running the CPU at 1.5V 24/7. Above that, I would feel uncomfortable leaving the system overclocked 24/7.

Stability: 1 hour Prime95 before I'll even consider the overclock successful. If I get stability issues of any kind later, the first thing I do is further ensure that the overclock is stable.

 

[Suspicious-Teach8788]

Prime stability in 1hr? Prime in 8? The general rule is 24. I told my friend that and he's like "Nah doesn't matter unless you're doing crazy stuff." So he goes Prime 8. Not even Prime 8 under terrible conditions. I've had Prime work for 24 hours and then a heat wave came and Prime crashed at 35 hours.

Well anyway guess what happened to my Prime 8 hrs friend? He crashed out of half his DOTA games the first week with his system. I know he dialed stuff back too and even then his testing wasn't rigorous enough.

Shrug, I OC to find the max. Then I dial that back a little. If my max is 4ghz, I'll dial back to 3.8 for 24/7. I'll minimize the voltages and what not to find my Prime 24 hours. After 24 hours I'll shoot for 36 hours just to check. Then I'll nudge the voltage up a small bump. So if 1.30v got me prime 36 stability, I'll run 24/7 at 1.31. I'll make sure to do this in closed case and to bump the ambient temp as much as I can to simulate a hot summer day.

Edit: OCCT is a quick way of finding out. I've had Prime take 4 -5 hours to fail but OCCT fail within 5 minutes. 3D Mark on loop is also a great traditional way. You heat up your GPU AND your CPU so you get a toasty inside...

Edit2: Also a lot of people fail to really stabilize their systems. While I passed memtest and everything I didn't figure out that my OCZ VX chips were messing up my system time and time again. I had to reformat every month due to a corrupted OS. I replaced the HD. I even dialed clocks back down to stock and it still failed. Memtest and Prime and OCCT everything.... passed. You may call me stupid for going Prime 36 hours, but I tested everything at max timings. OCing will cause errors if you don't test enough. Even then I couldn't identify what was going wrong. I swapped them out for a totally different type of RAM and just went to lower timings. No problems today.

 

[fffblackmage]

Also a lot of people fail to really stabilize their systems.

True for me at least. I only run Orthos for an hour at most, and, more often than not, only for 15 minutes or so. I just start playing games as a test for stability (I only overclock for gaming and not normal use). If it crashes, then I know it isn't stable and that I need to go back into bios to change something (usually voltage, I don't typically use aggressive timings).

On my old comp, an Athlon XP 2600+ (Barton) on a Shuttle nForce 2 mobo, I finally decided to overclock it and ran Prime95 to test it (1.9GHz -> 2.3GHz, 1.6V -> 2.0V, had some pretty insane temps). I got errors immediately. Still, that ran fine for a month until a spontaneous crash.

 

[LoneNinja]

This.

Once the voltage-to-Mhz ratio starts climbing drastically I back it off a bit. I also try to stay within the old rule of thumb of stock volts +10% for good air cooling or +15% for good watercooling. System must be OCCT/Game stable for 6+ hours.

I usually enable C1E and Speedstep too so I don't worry about the processor running over stock when I surf the web.

Same for me except I'm using AMD machines and K10stat to manually downclock when I'm doing nothing intensive.

 

[myocardia]

On my old comp, an Athlon XP 2600+ (Barton) on a Shuttle nForce 2 mobo, I finally decided to overclock it and ran Prime95 to test it (1.9GHz -> 2.3GHz, 1.6V -> 2.0V, had some pretty insane temps). I got errors immediately. Still, that ran fine for a month until a spontaneous crash.

I'm surprised that Barton didn't explode, melt down, or both. Seriously. Back when I had my Barton XP-M 2600, it only required a BIOS setting of 1.825V for 2.5 Ghz completely stable.

 

[Avalon]

I used to overclock as high as I could stably go, but now I overclock as high as possible with only slight increases of voltage over stock, in order to keep temps low.

 

[potato28]

If I have the funds to replace parts, I'll OC until something burns out. But, I'm not rich and just run with a modest OC.

 

[BlahBlahYouToo]

It depends on what cpu you're talking about. Most i7 920s can hit 3.9-4.1Ghz so you're talking about a 1.1-1.3GHz increase, not just a couple hundred Mhz. I can definitely tell the difference.

that's why i said 1156 i7's.
they have auto on-demand "oc'ing".

i don't need to surf the web .000001s faster day to day.

 

[Drakcol]

that's why i said 1156 i7's.
they have auto on-demand "oc'ing".

i don't need to surf the web .000001s faster day to day.

I could be wrong but I thought the x58 based core i7's also had the on demand "ocing" feature, it's just implemented better on the newer cores being the 1156 cores. At least that's from my understanding

As for Overclocking I OC till my cpu till it hits 85 degrees (and stable) in prime then step it back a notch or two for stability purposes .. I'd rather not have a dangerously hot core. That and my case acts like a space heater if I let my hardware get to hot.

 

[fffblackmage]

I'm surprised that Barton didn't explode, melt down, or both. Seriously. Back when I had my Barton XP-M 2600, it only required a BIOS setting of 1.825V for 2.5 Ghz completely stable.

I'm actually surprised I didn't just outright kill it as well. Especially since I was using a friend's stock AMD heatsink meant for his Athlon XP 2800+ (T-bred B). My original stock HSF was all aluminum. His had a copper core with the aluminum fins. Then I replaced the fan with a faster/noisier one I randomly had.

Needless to say, my room was very warm that month. I tried to compensate by turning the A/C on lowest temp/highest fan setting, but I don't think my apartment-mates appreciated that.

 

[Fardringle]

I'm surprised that Barton didn't explode, melt down, or both. Seriously. Back when I had my Barton XP-M 2600, it only required a BIOS setting of 1.825V for 2.5 Ghz completely stable.

My XP-M 2600 did 2.4 Ghz completely stable at stock voltage but it got really hot, particularly in the summer, so I usually ran it at 2.0 or 2.2 in the summer, and 2.4 in the winter.

 

[myocardia]

My XP-M 2600 did 2.4 Ghz completely stable at stock voltage but it got really hot, particularly in the summer, so I usually ran it at 2.0 or 2.2 in the summer, and 2.4 in the winter.

Well, I was using an all copper $60 Thermalright SLK-900a heatsink (their best, and most expensive at that point) with a 92mm 119 cfm(!!) Tornado fan. It was extremely loud, and used a full 1A of 12V power, but it kept my XP-M 2600 a hair above 40°C running Prime95 during air conditioner weather, and well below 50°C at up to 80°F ambient, all while running @ 209x12 @ 1.825V. Not that I regularly let it get that hot in my house, but I did experiment once, to see how well that all copper Thermalright and noisy Tornado fan could cool it.

And before you ask, I have no idea how accurate the reported temps with my Epox 8RDA3+ happened to be. The actual temps may very well have been higher, although it's hard to imagine, with that heavy heatsink, and a 119 cfm fan running wide open sitting right on top of it. I started a thread about it here, 5.5 years ago!