[Computers] Overclocking

[montyben101]

Wow. I have just overclocked my i5 4690K to 4Ghz and the improvement is incredible.

I then had a look and found that i could overclock my GPU too, I turned the fan up to 100% permanently and they overclocked it from 700hz to 934hz... The improvement was stunning..

Seriously, if you can overclock anything and have the cooling to do so, do it. You will not regret it. (unless your computer cooks itself)

btw I accidently set my GPU clockrate too high and werid things happened, it then crashed and reset itself...

[The-Bean]

Your GPU should automatically adjust the fan speed based on temperature. Running it at 100% permanently shouldn't be necessary and will likely wear the fan out faster.

[montyben101]

Yeah I just found out that you can edit the speed slope, previously it would never go to 100% even if it was at 85C !

[MartGonzo]

I used to overclock too but came to the conclusion it was too much messing about.

Some engines really don't like overclocking, when I tried most games improved but a few, most noticeably Oblivion really didn't like it and the performance dropped from 40fps to around 5fps.

[Linear]

I used to overclock too but came to the conclusion it was too much messing about.

Some engines really don't like overclocking, when I tried most games improved but a few, most noticeably Oblivion really didn't like it and the performance dropped from 40fps to around 5fps.

I don't think using Oblivion as a example of what an over clock can do is fair.

[K^2]

You guys overclock, and we are then left staring at game crash reports that look impossible. Keep in mind that just because the system looks stable, does not mean it will not totally flip out every once in a while.

[MartGonzo]

I don't think using Oblivion as a example of what an over clock can do is fair.

I just used it as an example that overclocking doesn't always make things better, some engines really don't like it

I'm not trying to dissuade anyone from trying it if they know what they're doing.

But certainly wouldn't recommend messing with clock speeds without a good understanding of the relationship between voltage, cpu clock speed, bus speed and cooling.

Edited February 15, 2015 by MartGonzo

[K^2]

I wonder why, though. Was it only a problem with older CPUs? I can see timing problems arising from TSC desync between cores, but modern CPUs use invariant TSC.

[sgt_flyer]

Some components handle overclocking better than others (when they are not outright locked away from overclocking)

However, when going high over the reference clocks, even without damaging the components, you can increase calculations errors - so your system can become more unstable the further you push your overclocking (because of eletric noise) ex : you overclock your processir, so it needs more power to run-thus gets hotter. To limit the increase in temperature, you reduce the voltage (and thus the power needed) - but your voltage levels are that much closer to the ambient noise (which is heightened by the higher clocks)

Now, when CPU/GPU manufacturers build their processors, they create a large number of dies on a single wafer - however, at the sizes they work with, there will be dies with more or less damage on it. The dies in the best shape becomes the high end processors, while those in lower shape sees some of their damaged sections disabled and sold as medium / low end products. Hence why, some cpu / gpu handle that better than others (even with the exact same processor, if one circuit was slightly damaged during manufacture, so it works at reference clocks, but using higher than reference creates increased resistance in the slightly damaged part - which suddenly stops responding - and your cpu becomes unstable

Overclocking is not something to be done without having done some research / having knowledge of it

[DaveofDefeat]

My motherboard overclocks my processor and memory automatically, do I win?

[Elthy]

No, auto overclocking is realy bad for your hardware. A big part of overclocking is searching for the right voltage to sustain a specific clock and finding the sweet spot where you only increase the voltage a bit but are able to raise the clock quite far. All auto OC solutions ive heard of simply raise the voltage to extreme values and the clocks only by a little bit since the working values are different for every single CPU and the mainboard just asumes the worst.

[Kerbin Dallas Multipass]

My good ol' Core 2 Duo is running on 2.8 (1.86 original) since 2007.

Guess I would have bought a new one years ago without the OC boost. So, yea, good investment I'd say

[78stonewobble]

You guys overclock, and we are then left staring at game crash reports that look impossible. Keep in mind that just because the system looks stable, does not mean it will not totally flip out every once in a while.

To be honest, you can have completely stock systems, that crash for no apparent system.

Substandard components sometimes find their way into systems. Laptops that apparently can't handle a heavy software task while sitting on a table. Odd combinations of hardware and software, that for completely obscure reasons wouldn't run properly ie. I was unable to run farcry 2 and 3, without looping sound and in the 2nd case looping sound to bsod, over multiple different versions of vista and win 7 32 bit and 64 bit, 2 different sets of ram, 2 different motherboards, 2 different gpu's and 3 cpu's. Never found out exactly what the problem was.. guestimating it was some weird timing error.

[SpaceExplorer]

Overclocking has and has not worked for me. Although I have been able to run majority of my software without the need of overclocking, I really refrain from using it to get better FPS or running games my computer is not suited for. I guess I am just scared of damage, or unnecessary heat from either my CPU or GPU.

[linkxsc]

Used to overclock years ago all the time. Every machine, an AMD athlon not being overclocked was like a crime against humanity.

Nowadays, not so much. Most games are so much more optimized, and processors and gfx chipsets are so robust in their stock states, you really dont need to bother.

And it saves on stupid crashes, jittery gfx and framerate drops when the cooling weakens for a split second because your cat jumped up on your desk.

[Kerbin Dallas Multipass]

You guys overclock, and we are then left staring at game crash reports that look impossible. Keep in mind that just because the system looks stable, does not mean it will not totally flip out every once in a while.

Funfact: Maxmaps runs an OCd machine

[Red Iron Crown]

I <3 overclocking. The way processors are binned means that it's very likely a lower end processor will work well at the clocks of a higher end one, if not more. Modern Intel cores will do 4+GHz easily and reliably without any exotic cooling. All that is really required is to do some stress testing at the higher clocks to check for stability.

(Currently running a nominally 3.2GHz processor at 4GHz, I've tested it stably up to 4.8GHz. Best overclocker since the old Celeron 300.)

[Weegee]

I've had good experience with overclocking... My last PC had a Phenom II X2 565, which OC'd from 3.0 to 4.0. On my new PC, I have an Athlon 750k (which is basically an AMD APU whose GPU is fried), and it overclocked to 4.4 with my motherboard's auto-overclocking utility. I will look into the voltage, but it seems to be fine so far, and I've had no instability due to the overclock. My GPU (MSI R9 270) is also a good overclocker, straight out of the box it OC'd to 270x speeds.

[Hrv123]

I used to overclock too but came to the conclusion it was too much messing about.

Some engines really don't like overclocking, when I tried most games improved but a few, most noticeably Oblivion really didn't like it and the performance dropped from 40fps to around 5fps.

If your temperature didnt go berzerk this sound to me as lacking NB voltage or to low freq. Whenever I had stabile OC but running applications (games) had issues, to slow, not working etc, it was NB not able to cope with OC CPU.

Happily runing CPU with stock 3,5+ mhz at stable 4,7 mhz OC, for that +3 fps at KSP ship lanch

[Camacha]

Some engines really don't like overclocking, when I tried most games improved but a few, most noticeably Oblivion really didn't like it and the performance dropped from 40fps to around 5fps.

Just an overclock is never going to cause this result - it means you have gone too far. I have seen similar behaviour when the GDDR5 (video) memory got pushed too far, which meant the error correcting had to step in to keep things in check, slowing everything down. In CPU's different things happen, but in pretty much all cases the story is the same - you have gone too far one way or another. It is even possible your cooling was not up to par, causing the processor to throttle.

Pretty much all Intel Core chips can take a lot of abuse before they start being troublesome. If you know what you are doing (or are able to follow instructions properly) you can tease rather big gains out of a chip without practical drawbacks. Even chips that inherently dislike overclocking, like mobile (ARM) chips, can be pushed to give a little more - though I would advise against that unless you have some miles under your belt.

Edited February 21, 2015 by Camacha

[K^2]

To be honest, you can have completely stock systems, that crash for no apparent system.

Substandard components sometimes find their way into systems.

Of course. But OC-ed machines crash exponentially more frequently than the non-OC-ed ones, whether parts were good or bad to begin with. And typically, on computationally expensive tasks. So when I see a crash dump from a game with an "impossible" state, I can pretty much disregard it. Unfortunately, digging through the dump to confirm that it's a bad state is time-consuming.

[Camacha]

Of course. But OC-ed machines crash exponentially more frequently than the non-OC-ed ones, whether parts were good or bad to begin with. And typically, on computationally expensive tasks.

That, pretty much, is the reason enthusiast overclockers only consider an overclock stable when pushed through a number of intensive, long term and often varying workloads. If your chip conks out when it has to strain it was never stable to begin with. Some folks go through somewhat excessive lengths to prove to themselves and others that an overclock is stable. Running full load torture tests for a number of days is no exception. That is still not enough for proper product validation or use in mission critical hardware, but plenty for almost all other intents.

Modern chips have a lot of headroom for various reasons, among which product diversity, lowest common denominators and power efficiency. Though you never have any guarantees and exceptions exist, this means that most of them will work a lot harder without much trouble and do it happily.

Of course, there are a lot of people out there that just fiddle about. I guess most problems you read about are caused by those folks.

[K^2]

Oh, yes, certainly. I'm not saying that one couldn't significantly OC a chip and get stable performance. Only that there are a lot of people out there who do it haphazardly. And, well, anyone who just starts out is probably going to make mistakes and add to the pile of "impossible" crash reports.

[Camacha]

Only that there are a lot of people out there who do it haphazardly. And, well, anyone who just starts out is probably going to make mistakes and add to the pile of "impossible" crash reports.

Yes, no doubt

I am pretty sure I started out this way too, though I do not think crash reports were a thing back then. Maybe in a physical pen and paper kind of way.

[K^2]

Crash dumps via e-mail go way back. But yeah, it's way better these days. A piece of software can have an exception handler within its own code that registers what's about to become a fatal crash, does a stack walk, generates an error report, and if time/connection allow, sends a full memory dump of the process back to the mothership.