Overclocking help

[montanaman]

Hello,

I built a new system last year and everyone has been telling me to overclock it for best performance. Since the wife going to be working this weekend I thought I would watch football and start the process. (Beer, football, and overclocking...good weekend huh!)

I really don't know where to start, what programs would be beneficial, websites to use..etc..

Computer specs:

Motherboard: ASRock Z77 Pro4 LGA 1155 Intel Z77 HDMI SATA 6Gb/s USB 3.0 ATX
Cooling: CORSAIR Hydro Series H50 Quiet Edition Water / Liquid CPU Cooler. 120mm
Memory: CORSAIR Vengeance 16GB (2 x 8GB) 240-Pin DDR3 SDRAM DDR3 1600 Desktop Memory Model CMZ16GX3M2A1600C10R
CPU: Intel Core i5-3570K Ivy Bridge 3.4GHz (3.8GHz Turbo) LGA 1155 77W Quad-Core Desktop Processor Intel HD Graphics ...
GPU: NVidia GeForce 660 ti
PSU: Corsair 750w

Any help would be greatly appreciated!

[Majesticii]

Did you even attempt to google this?

[noteworthynerd]

Did you even attempt to google this?

Super helpful!

OP: Check out this guide: http://www.overclock.net/t/1198504/c...asrock-edition That's the exact same guide I used when I first started overclocking, it is great.

[Cyrops]

Exactly your mobo and cpu:
http://www.youtube.com/watch?v=IAnjq6W-UCk

[Swampmoose]

Super helpful!

OP: Check out this guide: http://www.overclock.net/t/1198504/c...asrock-edition That's the exact same guide I used when I first started overclocking, it is great.

I like how he says that if you pass P95 after an hour, you're good to go, unless you plan to torture your CPU 24/7. Gets old hearing people say you need to run P95 for ridiculous amounts of time like 24 hours to make sure it's stable when all they do is play games for 2 hours a night.

[noteworthynerd]

I like how he says that if you pass P95 after an hour, you're good to go, unless you plan to torture your CPU 24/7. Gets old hearing people say you need to run P95 for ridiculous amounts of time like 24 hours to make sure it's stable when all they do is play games for 2 hours a night.

Yeah, I agree. Everything I had seen up to that point said to run P95 for 8 hours or more to prove stability, which seemed ridiculous to me. I followed that guide to the letter and maintained a 4.6GHz clock on my 2500K for months without issue.

[montanaman]

Did you even attempt to google this?

Must be early for you too huh?

Yes I did in fact. Lots of results, but some had different ways to do this. I came here and posted hoping to get tried and proven methods from people who, for the most part, are a lot more interactive and helpful than a youtube post or a blog from a site I have never heard of. They usually post great links/guides/methods that they have used. Then others, who are way smarter than me, can weigh in on those answers. As you can see, this has already happened.

And thanks to all of you who do post on these threads and give helpful and productive advice to noobs like me.

- - - Updated - - -

Super helpful!

OP: Check out this guide: http://www.overclock.net/t/1198504/c...asrock-edition That's the exact same guide I used when I first started overclocking, it is great.

Great! This will be very helpful.

That is the scariest look Ewok avatar I have ever seen!

- - - Updated - - -

I like how he says that if you pass P95 after an hour, you're good to go, unless you plan to torture your CPU 24/7. Gets old hearing people say you need to run P95 for ridiculous amounts of time like 24 hours to make sure it's stable when all they do is play games for 2 hours a night.

OT: Swamp... Did you used to hang out at Retpaladin.com for awhile? Thanks for your input on the overclocking too!

- - - Updated - - -

Exactly your mobo and cpu:
http://www.youtube.com/watch?v=IAnjq6W-UCk

Oh.. Not sure why I didn't see this on youtube. Thanks!

Saturday is looking better and better!

[Swampmoose]

OT: Swamp... Did you used to hang out at Retpaladin.com for awhile? Thanks for your input on the overclocking too!

Sheesh, I can't even remember. I think so though. I was always on a number of sites back when I played regularly. Could have been an impostor though! I was the only Swampmoose on the Armory back then too, but now there are 30 Swampmeeses!

[chazus]

While your system might work fine 100% in gaming the stress test is there for a reason. And 8 hours is the recommended minimum because... for whatever reason, sometimes it takes a long time to weed out the bugs.

In P95, @ 4.7ghz, my system will crash at the 4 hour mark, reliably, every time. It works fine in everything else. Games, whatever. But there is some kind of inconsistency that causes it to crash, that I wouldn't have run into unless I ran it that long.

[Swampmoose]

While your system might work fine 100% in gaming the stress test is there for a reason. And 8 hours is the recommended minimum because... for whatever reason, sometimes it takes a long time to weed out the bugs.

In P95, @ 4.7ghz, my system will crash at the 4 hour mark, reliably, every time. It works fine in everything else. Games, whatever. But there is some kind of inconsistency that causes it to crash, that I wouldn't have run into unless I ran it that long.

I recently watched a video of an Asus employee not only saying that it's completely unnecessary to stress a CPU like that under something like Prime 95 for that long, but that you can also do "stress testing" relative to how you will use the PC. Example would be testing under something like Cinebench. I'll look for the video.

This bolded text above is the reason why. Your having an issue 4 hours into stress testing that you would never encounter on a day to day basis, yet your are most likely scaling back your OC to account for this error. Sacrificing performance in what you actually use the computer for to pass an unnecessary test. Heck I even heard a guy once say that his test fails at 12 hours. Ok, but are you really stressing your CPU for 12 hours under normal circumstance?

Edit: I think this is the video but I'm at work and it will take half a day to load.

https://www.youtube.com/watch?v=Z7zPu9255ZI

[Gravewyrm]

Some nice guides in this thread, Kudos guys, and thanks for sharing them. I've actually bookmarked a cpl of them for future reference

On the topic of stress testing, I generally do mine in 3 diff stages. First is just getting into windows, and running the WEI "benchmark" (lolz). If it fails, or even bluescreens, I adjust voltages until I can get it to pass, with a final result. Next is AIDA64's "Stability Test". Again, I go until it fails, or I hit the 2 hour mark. Finally, Prime 95, until I can get it to give zero errors for 1 hour. The AIDA64 test is most likely adequate for the average builder whose doing a mild OC. P95, for anything longer than say, 2 hours, seems like overkill to me.

[chazus]

Your having an issue 4 hours into stress testing that you would never encounter on a day to day basis, yet your are most likely scaling back your OC to account for this error. Sacrificing performance in what you actually use the computer for to pass an unnecessary test. Heck I even heard a guy once say that his test fails at 12 hours. Ok, but are you really stressing your CPU for 12 hours under normal circumstance?

That indicates that there IS some issue, at that rate. I might not normally stress it that much, but sooner or later I'll hit that 'whatever it is'. So yes, it's worth dialing it back 100mhz and not have to ever worry about it.

Stress testing is about guaranteeing stable operation. Not hoping for it, or assuming. I know, for a fact, that it's stable at the lower clock. I do not know, at the higher clock. A lot of people don't stress test, and say "it works fine at XXX speed" and 4 months later, it crashes once a week. Well, guess why.

[Tomec]

TBH I can see both sides. While Swampmoose has a point that you are sacrificing performance, chazus is right in that there IS an issue there somewhere. Even if you don't normally put the same strain on your CPU normally, what happens that one day you do? Personally, I do a lot of graphic rendering with my system. Sometimes this takes DAYS to render everything out (yay not having access to a render farm when you need it..). If my system always fails @ the 4 hour mark, not only am I vastly increasing time it takes to do things, but I am risking losing a lot of data too.

[Swampmoose]

Some nice guides in this thread, Kudos guys, and thanks for sharing them. I've actually bookmarked a cpl of them for future reference

On the topic of stress testing, I generally do mine in 3 diff stages. First is just getting into windows, and running the WEI "benchmark" (lolz). If it fails, or even bluescreens, I adjust voltages until I can get it to pass, with a final result. Next is AIDA64's "Stability Test". Again, I go until it fails, or I hit the 2 hour mark. Finally, Prime 95, until I can get it to give zero errors for 1 hour. The AIDA64 test is most likely adequate for the average builder whose doing a mild OC. P95, for anything longer than say, 2 hours, seems like overkill to me.

AIDA64 can even be a better test than P95. It's recommended for Haswell at least as Intel supports it's ability to fully stress the architecture.

That indicates that there IS some issue, at that rate. I might not normally stress it that much, but sooner or later I'll hit that 'whatever it is'. So yes, it's worth dialing it back 100mhz and not have to ever worry about it.

But where do you draw the line? Using the 12 hour guy as an example. He inevitably hit's his "whatever it is" issue as well, but how applicable is that to regular use? Of course regular use varies person to person, hence the suggestion of testing to your particular uses.

Stress testing is about guaranteeing stable operation. Not hoping for it, or assuming. I know, for a fact, that it's stable at the lower clock. I do not know, at the higher clock. A lot of people don't stress test, and say "it works fine at XXX speed" and 4 months later, it crashes once a week. Well, guess why.

You aren't really guaranteeing stable operation if you're just picking an arbitrary number like 8 hours to stop at unless whatever program you are running can legitimately say it has completely stressed the CPU under all possible circumstances in that 8 hour window, which I don't think is the case or we'd know it. All you can say confidently is that it's stable for X hours under P95.

Not testing at all is a bit silly but so is testing for 8 hours when you're casually gaming 2 hours a night or browsing the web for a few mins here and there and never putting the CPU under any real stress. I'm with Gravewyrn. A hour or so with Aida and maybe another hour with P95 should be sufficient for this type of user. Now if you're using this is a work PC and doing some actual stressing...a lot longer could be necessary.

TBH I can see both sides. While Swampmoose has a point that you are sacrificing performance, chazus is right in that there IS an issue there somewhere. Even if you don't normally put the same strain on your CPU normally, what happens that one day you do? Personally, I do a lot of graphic rendering with my system. Sometimes this takes DAYS to render everything out (yay not having access to a render farm when you need it..). If my system always fails @ the 4 hour mark, not only am I vastly increasing time it takes to do things, but I am risking losing a lot of data too.

Edit: Perfect example of someone who should test a heck of a lot longer than a casual gamer.

[chazus]

If it happened only once, I might not worry, but there is something that P95 'does' at the 4 hour mark that the system does not like. It's rare, I don't even know what that 'thing' is exactly.. But what other programs do that 'thing'? The 8 hour burn in test effectively means that it's passed every different thing it does. Keep in mind, stress tests are not the program simply going "BLAAAAAAAAA" at the hardware until you turn it off. It rotates through a number of different tests with different variables, and going through all of them takes at least a few hours.

- - - Updated - - -

but how applicable is that to regular use?

That's the point. You DON'T Know how it's applicable to regular use. The longer it runs, the more sure you are.

You aren't really guaranteeing stable operation if you're just picking an arbitrary number like 8 hours to stop at unless whatever program you are running can legitimately say it has completely stressed the CPU under all possible circumstances in that 8 hour window, which I don't think is the case or we'd know it.

We DO know it though. Because hundreds of thousands of people have run benchmarks, and it's been determined through trial and error that ~6-8 hours is a reliably accurate stress test time. Any more is likely just burning time on tests already run. Any less is not stressing it adequately.

[montanaman]

Thanks guys! Even though a lot of what you are saying is above me, I am learning stuff. Some of it is starting to come together and I might have to leave out the beer while overclocking! I am glad I posted instead of trying to just wing it. Thanks again!

[Swampmoose]

If it happened only once, I might not worry, but there is something that P95 'does' at the 4 hour mark that the system does not like. It's rare, I don't even know what that 'thing' is exactly.. But what other programs do that 'thing'? The 8 hour burn in test effectively means that it's passed every different thing it does. Keep in mind, stress tests are not the program simply going "BLAAAAAAAAA" at the hardware until you turn it off. It rotates through a number of different tests with different variables, and going through all of them takes at least a few hours.

- - - Updated - - -


That's the point. You DON'T Know how it's applicable to regular use. The longer it runs, the more sure you are.

We DO know it though. Because hundreds of thousands of people have run benchmarks, and it's been determined through trial and error that ~6-8 hours is a reliably accurate stress test time. Any more is likely just burning time on tests already run. Any less is not stressing it adequately.

What is the source of 6-8 hours? You can search the internet endlessly and find that the "standard recommendation" is 6-8, 12, or 24 hours. No consistency at all. I think it would be a bit more obvious if hundreds of thousands agreed that 6-8 was reliable. I simply haven't seen that. The only real consistency is those recommending 24 hours, are those stressing their CPUs day in and day out, and those suggesting lower, simply are not.

And if people are hitting errors at 12, 16, 20 hours, 6-8 hours is still not a guarantee and it goes back to being an arbitrary number and you're simply assuming it's stable, not guaranteeing it. If a Blend loop takes 8 hours, and people are failing at 16 or 20, they have essentially passed a test the first time around and failed it the second time around. By that method/logic, you must test an infinite amount of time as it could fail the same test at any time. There is no difference between what you are doing and what I am doing other than the length of time run.

I won't disagree that there could be instability issues if you have a problem at 4 hours every time, but how you use your PC should determine if it is an issue to correct, or if it is not a concern.

Thanks guys! Even though a lot of what you are saying is above me, I am learning stuff. Some of it is starting to come together and I might have to leave out the beer while overclocking! I am glad I posted instead of trying to just wing it. Thanks again!

NEVER leave out the beer. EVER!

[chazus]

What is the source of 6-8 hours? You can search the internet endlessly and find that the "standard recommendation" is 6-8, 12, or 24 hours. No consistency at all. I think it would be a bit more obvious if hundreds of thousands agreed that 6-8 was reliable.

Many, many hours of research. 6-8 hours is what people usually recommend, or simply end up doing, either overnight or at work. Keep in mind, there is no place where hundreds of thousands all congregate and agree on, but on most of the forums I've looked into this, that's generally considered the most acceptable times.

And if people are hitting errors at 12, 16, 20 hours, 6-8 hours is still not a guarantee and it goes back to being an arbitrary number and you're simply assuming it's stable, not guaranteeing it.

I guess I shouldn't have used the word 'gaurantee', implying 100.0000000%. The fact of the matter is, no computer is 100% stable. Just like no computer is 100% secure, either. However, 6-8 hours is generally accepted in verifying that the highest a chance of failure has been covered without spending an inordinate amount of time. Meanwhile, any less time, and your chances of verifying stability goes down rapidly. I know for a fact that P95 takes at least 2 hours to go through it's full battery of tests, so certainly any less than 2 is literally 'not completing the test', and effectively implies zero accuracy of stability since it wasn't complete. Between 2-4 hours likely runs the full battery at top heat (early P95 tests might run at lower heat). 4-6 hours just double checks.

It seems silly not to do that to get a 'good assurance' especially if you just do it overnight. I assure you, any system set to stress test indefinitely will eventually fail.

AIDA64 is good too, for heat purposes, and there's no harm in running both. Heat isn't everything. They run different algorithms, and I've seen lots of people where one or the other can run a day straight flawlessly, while the other crashes quickly.

In the end, it depends how OCD you are about system stability. Can you OC to a certain degree, stress test it for 2 hours, and call it good? Yeah, probably. Is it entirely possible that you do that, and months down the road your system crashes on rare occasion and you don't know why? Also possible. Again, it all depends on what you want to do. However I don't see why you wouldn't run it overnight to be super sure, and not have to worry about it ever again.

[Tomec]

If it happened only once, I might not worry, but there is something that P95 'does' at the 4 hour mark that the system does not like. It's rare, I don't even know what that 'thing' is exactly.. But what other programs do that 'thing'?

I think that it's important to note that. While, right now, it could be that your average gamer/game isn't going to do that "thing" is that true in the future? It's not like games are getting EASIER for systems to run. You also have to think about the new consoles and what they can do, the better they look the better PC games will look. Which also means more power needed.

As was also mentioned, you don't know what that thing is. I had a WEIRD issue where I could render images and stuff just fine. But whenever I went to play a game I would get fairly frequent crashes. Why did the game playing cause a crash when I could render images/other things like that for days?

Also in regards to the amount of time tested, I personally find the whole "well I'm only going to play a game for 2 hours so wont test longer than that" thing to be a bit iffy. Even ignoring how tests work, are you telling me you NEVER play longer than 2 hours? What about raids? How about games like League of Legends where matches can last a long time? Or the weekend when you can just veg out on games?

[chazus]

Sort of a curiousity on that whole 'thing' concept.. Here's sort of an example of oddities people might run across.

4.4ghz everything runs fine. 4.6ghz everything runs fine. YET despite that, at 4.5ghz, Skype and WoW (and ONLY those two programs, nothing else) crash intermittently every couple days. That's the kind of 'thing' where it's difficult to weed out.