Overclocking rant!

zuludawn

zuludawn

zuludawn

Associate
Joined
20 Nov 2002
Posts
1,405
Overclocking tools Prime and IBT are not the be-all and end-all of proving a system is stable, they are a quickfire route to stress testing only.

They produce 100% load which in normal usage never happens (note normal usage, browsing forums games etc)
They produce high temps with load , again with normal usage never happens.

Taken if you pass with low temps brilliant but, like my cpu and many others if the temps are high THIS DOES NOT MEAN YOUR SYSTEM IS UNSTABLE BECAUSE IT GETS HOT AT 100% STRESS LOAD.

Im sick of seeing posts of run prime for 6 hours IBT for an hour, just why, just because some systems will and some wont who cares.

For years before these stress tests it was done by trail and error, BSOD IS YOUR FRIEND OR ENEMY depends how you look at it.

Its simple no BSOD, no freezing up then i would call that a stable system however you use it.

Plz stop if its to hot , if it dont run xxx, i wouldnt, it will melt, its getting boreing

My system according to OCUK experts will melt in a day, but im still here, still play BF3 etc, system on at least 18 hours a day, has been for a month.

Hmm i must be missing something (must be most post high oc then throttle back for 24/7, we all do it we all want the best OC we can get. You have to go to the limit then decide if you like it there or not). Seen with with Ivy, try for 4.8ghz then run at 4.4ghz 24/7.

Over and out! Yes Bass is my friend.

No ONE answered my question either in a previous post ALL TEMP RELATED.
Why can a GPU run at a constant 95c but it madness for CPU to do the same according to the majority. Read post i would love to no why, plus with such small heat sinks compared to CPU custom monsters.



http://forums.overclockers.co.uk/showthread.php?t=18399428

Who would in ther right mind oc to the point where you get a BSOD every now and again, it is what it is..... Takes another good swig and waits!
 
Last edited:
Depends on what you use your PC for TBH. Mine is frequently running at 100% on all cores for hours or days at a time as I use it for CG rendering - meaning I want it to be as stable and as fast as humanly possible, and prime gives a good example of what that will be like.
 
Rhodri2311 said:
Depends on what you use your PC for TBH. Mine is frequently running at 100% on all cores for hours or days at a time as I use it for CG rendering - meaning I want it to be as stable and as fast as humanly possible, and prime gives a good example of what that will be like.
Click to expand...

Yes point taken, this is meant for the more general games user etc.

As a high OC mine will pass Prime and IBT but at very temps, i have run system for a month gaming use etc and only ever hit 70c as a max (minus the run stress test requests for hours and hours to prove it stable, which was done but still frowned upon)
 
Last edited:
I overclock both cpu and my gpu's. Stability can never be 100% guaranteed. Ive had cpu's that were 24hrs prime small fft/blend/ibt stable, yet fire up a game and boom, (bfbc2 being a great example) even moreso than the newer bf3. Ive seen a few of your posts mate where youve clocked your 920, but tbh enable hyperthreading when stress testing, eg at 4ghz with ht enabled on my current cpu, i hit 80c in ibt, 1.3125 vcore. But my current case has very poor airflow compared to my last one. Cooler is a k2, 73c max in my old case. But im never gonna see temps like that in my daily use, which primarily is gaming.:)
 
setter said:
I overclock both cpu and my gpu's. Stability can never be 100% guaranteed. Ive had cpu's that were 24hrs prime small fft/blend/ibt stable, yet fire up a game and boom, (bfbc2 being a great example) even moreso than the newer bf3. Ive seen a few of your posts mate where youve clocked your 920, but tbh enable hyperthreading when stress testing, eg at 4ghz with ht enabled on my current cpu, i hit 80c in ibt, 1.3125 vcore. But my current case has very poor airflow compared to my last one. Cooler is a k2, 73c max in my old case. But im never gonna see temps like that in my daily use, which primarily is gaming.:)
Click to expand...

I got to 4.5ghz with Ht enabled which is mad and what im running now. IBT still passed but a stupid temps which this chip still took. Buts that's stress testing and this a gaming rig.

You right for stability as well ran BF3 fine for a month, new 12.4 drivers crashes in 4 games costing me hours of lost kills, back two 12.3 all ok again, so many variables.

Like i keep saying a very special cherry picked i7 920 4.5ghz with Ht but no one cares these days its old lol. Three years ago with this batch number i would have been rich, but i got it from Canada, now this guy is a lot better off, all his Sandys are 5.2ghz guaranteed and unlocked Ivy`s. He certainly knows his batch codes.

But i still was faster an a 3570k at 4.5ghz, more threads or not. Even at 4.4ghz it was 7.59 in Cinebench, but games don't use Threads so pointless i guess, unless its a benchmark for something else like encoding?

http://forums.overclockers.co.uk/showthread.php?t=18395882&page=10

45x multiplier try 215 BCLK x a tiny 21x

Goes to hide yet again
 
Last edited:
it does get very annoying i find with my E6600 i run perfectly stable oc'd from 2.40 to 3.40ghz untill i hit 80% load or above but i never hit that load on any game only when i run the likes of ibt and prime 95
 
I own the 920 in sig, it's done 4.2ghz. I also own a 930 on a p6 x58de, ive had it prime/ibt stable at 4.3ghz under a TRUE with very loud fans. But it's a barely used second rig. Even 4ghz on the rig in sig is plenty for my use, it's not a bottlekneck with gtx 470's in sli for games. Yes, ive heard others mention amd/ati and driver issues, cant say myself as ive only used nvidia gpu's. The i7 9** are still very powerful cpu's for a gaming rig, and can be had very cheap now. For example, the board, cpu and my original ocz reaper ram kit cost almost £600.00 when bought a few years ago. I picked up my 930 and a p6 x58de with 6gb of ram for £200.00. I call that a bargain.:)
 
Ollie1132 said:
Most people just fail to understand how to actually OC and what is stable and why to be fair.
Click to expand...

Surly what is stable is pretty obvious your rig does not crash normally. Expect for.

1. New GPU driver release and most games crash
2. New games crashes anyway BF3 and Skyrim or example on release.
3. Oc in the Winter and ambient temp rises by 10c in the summer.
4. Bloody thing stops working anyway after years of good service.
Please feel free to add more....

Good OC just comes from an understanding of the basic Bios, some modern Bios are easy (Asus) older ones have more settings than it took to launch Apollo 11. Sometimes its just trail and error but it all about perseverance.

Even though IBT is a bitch for heat it is very good guide for lowering v-core watts (because at just 10 runs its quick), i only use it now to lower watts, i find if it passes IBT it will run anything.
 
Last edited:
You CPU doesn't need to be at 100% to crash.

You overclock your CPU, you run prime 95 for 2 hours and then it fails. This doesn't mean 100% CPU caused it to fail. This means that approx 1 in every 10 billion (for example) calculations will fail. 100% CPU just means it is processing those calculations faster so has found that 1 fail in 10 billion quicker then normal usage. Normal usage might take a month to find that fail/crash but it will still happen.
 
Personally I dislike IBT since as you said earlier it's quite unrealistic having all 4(6) cores running at full load. IBT generally creates a lot more heat over Prime however if you don't BSOD as you said then doesn't make a difference which way you check for stability. If you wanted you could check for stability doing every day tasks... would probably never BSOD that way however.
 
no_1_dave said:
You CPU doesn't need to be at 100% to crash.

You overclock your CPU, you run prime 95 for 2 hours and then it fails. This doesn't mean 100% CPU caused it to fail. This means that approx 1 in every 10 billion (for example) calculations will fail. 100% CPU just means it is processing those calculations faster so has found that 1 fail in 10 billion quicker then normal usage. Normal usage might take a month to find that fail/crash but it will still happen.
Click to expand...

Lol does that mean a crash each month or once it finds the fail total crash, then you reset and wait a month (for example) for the next fail?
Plus in Prime is it not a set of certain calculations that in real usage my never not be used.

Help me out Dave im struggling give me the sesame street version plz
 
In simple terms if it fails prime there is a chance it will fail anytime. It has nothing to do with the % of CPU usage. It may mean on average 1 in every 10 billion calculations will fail. It could take a week, a month or a year, who knows. It all depends how much you use your PC. The 10 billion figure is hypothetical, It could be any large number, I have no idea how many calculations a CPU makes every second.
 
Ollie1132 said:
Personally I dislike IBT since as you said earlier it's quite unrealistic having all 4(6) cores running at full load. IBT generally creates a lot more heat over Prime however if you don't BSOD as you said then doesn't make a difference which way you check for stability. If you wanted you could check for stability doing every day tasks... would probably never BSOD that way however.
Click to expand...

Spot on.
 
no_1_dave said:
You CPU doesn't need to be at 100% to crash.
Click to expand...
This, i said earlier. full load in prime95, ibt/linx and memory tested on memtest for 24 hours. A game, bfbc2 ran on stock gtx 470's in sli, a few minutes into a game, bsod. One of the most demanding cpu games ive played in recent years. Even the newer bf3 isnt as sore on cpu usage.

Stability is a very important consideration for people who run programs such as folding at home. This demands a stable system to generate accurate results. And it can be as demanding as some of the stress test programs such as prime95.
 
Last edited:
zuludawn said:
Spot on.
Click to expand...

Indeed, well nearly.

A combination of operations occurring simultaneously while the chip is at or near an unstable state will cause it to fail. Running it faster will simply increase the number of operations happening.

It is not the single operation that causes the chip to fail, it is a combination of what the chip was doing at the time it was doing that operation that causes the unhandled exception. Increasing the frequency simply increases the number of operations being performed.

I like to think of it like cars on a motorway. A low number of cars means that there is a lower risk of accident. More cars means more risks of accident. It doesn't necessarily follow that there will be an accident, just there is a greater risk of it. As the speed of the traffic increases the chance of collisions also increases.

The quality of the CPU is analogous to the size/quality of the road. The better quality the CPU, the wider and straighter the road. :)

It's probably a rubbish analogy, but it works for me.
 
Tuppy_Glossop said:
If it passes prime there is still a chance that it will fail anytime. The key question is "what is the difference between the two chances ?". IMO this can not be quantified accurately.
Click to expand...

Prime is effectively worst case conditions if it can survive that it should not have any problem in any usage, of course there will always be software bugs that can throw up errors regardless of how stable the CPU is.

There have also been cases where a Prime stable machine crashes in games but that will usually be down to the added heat from the GPU, extra stress on the PSU and other external factors like overclocked pci/qpi bus etc.

If AMD/Intel released cpu's that were not Prime stable most of the modern world would cease to function properly, for example the OCUK forums would rarely be up with messages getting constantly corrupted etc.

I've just never understood why people all willing to sacrifice stability for 5-10% extra performance which won't be noticed outside of benchmarks.
 
it's about finding a balance - performance vs cost. In order to keep that CPU stable at those higher frequencies usually involves buying higher quality components, and I for one like to use that kit optimally - FOR PERFORMANCE. My rigs folding 24/7 so mine needs to be stable 24/7. If it's not, then the points that I lose for my downtime outweigh the benefit of winding the dial up past 11, so I simply stick there @ 4.2Ghz. Other users don't hammer their rigs quite so hard, so they can push it a little further, but for shorter period.

It's all about balance, and finding yours... each scenario is quite different.

mmj_uk said:
Prime is effectively worst case conditions if it can survive that it should not have any problem in any usage, of course there will always be software bugs that can throw up errors regardless of how stable the CPU is.

There have also been cases where a Prime stable machine crashes in games but that will usually be down to the added heat from the GPU, extra stress on the PSU and other external factors like overclocked pci/qpi bus etc.

If AMD/Intel released cpu's that were not Prime stable most of the modern world would cease to function properly, for example the OCUK forums would rarely be up with messages getting constantly corrupted etc.

I've just never understood why people all willing to sacrifice stability for 5-10% extra performance which won't be noticed outside of benchmarks.
Click to expand...
 
You must log in or register to reply here.
Back
Top Bottom