First time overclock

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
Hi all,

I just finished my first build, and I would like to do a light and stable overclock to about 4.0-4.2 ghz. I'm a noob to overclocking, so I'm sorry for any ignorant questions. I tried just using the ai suite 3 and bumped it up to 4.2 and it crashed, but I didn't change the voltages or anything so I'm sure that's the reason for the crash.

I5-4670k
Maximus VI formula
8gb gskill ram 1866
Asus GeForce gtx 780 directcu ii oc
Cm 212 evo

So there is probably already a thread somewhere, and I would appreciate any help you can offer. So anything like a link to a thread or video on how to overclock using these components would be great!
 
Solution
O.k so your temps are a little high and you need a higher than normal amount of voltage which means that your CPU is probably doing all it's going to do for you as far as overclocking (I'd recommend going down to 4.0GHz and lowering your voltages slightly and check temps and stability again)

You definitely need to test your RAM. You should lower your clocks and volts back to default settings and test your RAM before doing anything else.

Your CPU cooler touching one of your ram sticks is not bad as long as it isn't putting pressure on it. If you can slide a piece of paper between the cooler and the stick you're fine. If not you may want to get some lower profile ram or a different CPU cooler. Do the memtest first!

Dom_79

Distinguished
To be honest I'm not going to do exactly what you can do which is type overclocking i5-4670K in the search bar at the top of the page right here. Because even if you can't do overclocking on your own yet (no one knows everything, we all learn) you CAN do that search on your own. There are many, many threads and even one specifically about this CPU. It's called something like "So and so's (can remember his name) quick and dirty i5-4670K overclock" or something like that.

I'm new to OC'ing as well and I found that the videos (Youtube for example) were not as helpful as the threads and articles I looked up myself here on Tom's Hardware. Myself and others would be happy to help you with your overclock when you have specific questions about it and not posts asking people to do simple thread/article searches for you and post you links.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
Ok so I read a few articles and was able to simply adjust the multiplier to 40 for all cores and the system solid during the prime 95 torture test for 15 mins. However I did observe 2 things that changed. First, on the motherboard where it shows numbers or letters (not sure what that's called) before my overclock it showed 40,and after it showed AO. And second, when booting up the computer, red led's would light up on the 24 pin power connector to the motherboard. But then would turn off once the computer was booted. Even with the 2 changes everything worked fine, and my CPU never went over 58 degrees under full load.

Should I worry about either of those two things I noticed?

Also when I tried to let ai suite 3 optimize the overclock at 4.2, I get a problem showing "uncorrectable error" in my windows 8 screen. I'm sure that has something to do with instability, but if it's possible I would like to have my CPU at 4.2 ghz nice and stable. Suggestions?
 

Dom_79

Distinguished
I would recommend you run prime95 for a couple hours at least on Large FFT (don't know what "high" setting is) if you want to determine stability and heat management.
If you want faster results, try Intel Burn Test (this will push your CPU past any real world conditions)

Don't use apps to overclock. Going into bios and setting the core multipliers to your desired level and raising vcore just enough (be very careful when adjusting voltages in bios, you can set things high enough to immediately and permanently damage things) to make it stable is one of the best noobie ways to OC (that's how I did it)

I have no idea what the lights on your motherboard mean. Check your motherboard manual (and keep it handy as you will need to reference it often while you learn to OC)
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
Thanks for the suggestion. So i tried putting the multiplier to 40 and the vcore to 1.25 and after running prime95 for half an hour and everything was stable. i put it through intel burn test for 20 runs and it passed. however, when i ran BF4 it took less than 20 minutes for the computer to crash. should i continue to raise the vcore voltage or do i just have a shitty chip? because just about every chip should be able to get up to 4.2 ghz from what ive read?
during intel burn test at those settings the temperatures would peak at about 72 degrees.
 

Dom_79

Distinguished
Half an hour is still not enough in prime95 with large or small FFT's to determine system stability. You really will need to run it for quite some time (at the very least 2 hours - people will often run it for 24) to get a good idea of stability.
What setting was IBT? Standard or higher?

If your were using at least standard level in IBT and you got up to 72C that is fine (the i5-4670K should be good up to 80C on full load)

BF4 crashed but IBT didn't? That is very odd. Have you overclocked your RAM or GPU? What was your GPU temp when BF4 crashed?
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
yea i just thought the chip would be able to handle 4 GHz without a problem, even without a raise in vcore voltage. at least thats what ive read on some other threads. however, the IBT was at standard setting. i will have to try and run it for at least 3-4 hours tonight, but i need to make sure i have everything set up properly.

its very weird that it crashed in-game. my gpu is at stock levels, i havent tried to overclock it yet and in game it is steady at 78 degrees. as for my ram i just have it in xmp prifile #1 which is 1866 Mhz at CL9, timings at 9-10-9-28-2N

Edit: i never wanted to do any kind of extreme overclocking so if i could get it to run perfectly stable at 4.0-4.1 GHz i would be extremely happy
 

Dom_79

Distinguished
I've been doing some research and BF4 very likely crashed because it is poorly optimized for NVidia GPU ("right now" is what people keep saying so I guess a patch is being developed to fix that).

I was able to get 4Ghz at stock volts but I've heard some have trouble getting the "Turbo" (3.8GHz) to work stabile at stock volts. With Vcore at 1.25v you should be able to get between 4.3 and 4.5GHz (I'm at 4.4GHz with 1.23v Vcore)

Most likely you were stable but BF4 had a driver issue. IBT is pretty damn rough and if you passed at least 10 passes at standard setting it's 90% sure you're system is stable. BF4 can't push your CPU and RAM as hard as IBT guaranteed.

BTW your setup and temps seem pretty solid. You should try for 4.4 - 4.5 GHz. I could actually see a difference in load and process times between 4.0 and 4.4. I would recommend seeing if you can run IBT on High for 15 passes without going over 75C at 4.4 or 4.5. You'll be happy you did :)

One last thing. You mentioned running what seemed like IBT for 3-4 hours tonight? You meant prime95, right? and if you're going to run over night, why not run for close to 8 hours on Small FFTs while you're sleeping. If you pass that you are almost (97-98%) guaranteed that your OC is stable.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
hmm i hope it is a driver issue, and that it could be fixed soon, its not fun having a huge game going and then boom crash. Yeah im sorry i meant prime 95 for a few hours, not IBT. im going to try having vcore voltage at 1.25 and multiplier at 4.4 and see how far it gets. also, should i be changing anything like cache voltage or imput voltage, or should i just stick to regular vcore? ill do that test now and ill notify what results i get.

also gpu temp while playing for 2 hours was 72 degrees constant, not 78.
 

Dom_79

Distinguished
That temp while playing BF4 seems fine it wasn't the reason it crashed.

Usually when someone is going to OC they don't do much but change settings for the OC, monitor temps and run test programs to make sure it's stable. After doing that until they are comfortable with the results, THEN they game. Because you don't use a game to check if you are stable and you don't wanna wreck your gaming experience playing on an unstable OC and crashing out.

Before you game again, make sure your OC is stable then run memtest (google how to download and run that prog). Make sure there is nothing wrong with your ram (I really doubt there is but that is a small possibility why BF4 crashed)

As for changing anything else in bios for OC, I would strongly recommend against it. (I mean REALLY strongly)

Right now you are just learning about what all these things are and what their function is. More in depth customizing is for when you are knowledgeable about what these settings will actually do for you when (and IF) you are "tweaking" to obtain the highest clock you can get or getting the absolute best results (which may not be highest clock speed) tailored for your specific applications and games
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
ok so i have my CPU at 4.2 GHZ at 1.232 vcore on the manual setting, which is what ive read it should be on for stress testing. it is said that once the system has been fully tested i can put that setting to adaptive so that the voltage can be brought down enough while the pc is idle (got that info from both Linus Techtips and the Asus guide on overclocking). also the temp of my cpu never went above 67 degrees during IBT. I just got through 15 runs of IBT on the high setting and it was successful. now I will run prime 95 until the morning on small FFTs, and like you said, if it gets through that I can be 97% sure that my system is stable.
 

Dom_79

Distinguished


That all sounds correct, except I would ask around a bit about the adaptive voltages thing. Since your temps are REALLY good (what program are you using to monitor temps by the way... I should have asked you that sooner because correctly monitoring your temps during stress tests is VERY important) and your voltage is below 1.3 you won't really see much benefit from adaptive voltages in my opinion.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
Ok that's good to hear that it's looking good so far. At idle under these voltages the CPU is at 32-40 degrees. I'm using the airport suite 3, because my Asus motherboard came with its own thermal probe for more accurate temperature reads.

Ok, so in the morning when I'm done stress testing, and (I'm hoping) everything is OK I'll do the members test. But I'll look more into the adaptive setting and see if it's really necessary, because I don't really want my system to become unstable by using that setting.
 

Dom_79

Distinguished


Nah, nah, nah. The motherboard sensor is not as accurate as the sensors on the CPU itself. Use Core Temp (only displays the temps of the CPU) or HW Monitor (displays the temps of anything in your entire system that reports a temp - can be a little too much/ distracting) to read the CPU temp (you will see the individual temperatures of each of the four cores of your CPU)
This is the only way to properly see your CPU temps.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
Oh no... I used core temp and now it's showing an average of 84 degrees between the 4 cores. That's a little too high for comfort. So maybe I should down clock to about 4.0 and lower the voltage to about 1.1?

Edit:
However, I've have been running prime 95 for over 2 hours on small ffts and everything was running normally.
 

Dom_79

Distinguished


Crap. I was worried about that. If your ambient temps (the temperature in the room) are 25C or less you may want to make sure your thermal paste was correctly applied (because your CPU cooler can do a better job than that for sure). Before you do that though, what thermal paste did you use?

So at least processing wise your OC is stable. We can work on getting your temps down.

I'm going to reply to any other questions you have tomorrow (bed time ;) ) Don't do anymore stress tests until we can get your temps down o.k?
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
i definitely wont be doing any stress tests from fear of frying my cpu. ok so my ambient temps are never over 23 degrees. i used Gelid GC extreme thermal paste, and i spread a layer over the top of the chip ( i know you can put a dab in the middle the size of a cooked grain of rice, or spread it out) before attaching the cooler. the cooler itself has a horrible mounting bracket IMO and even though i am sure i tightened everything just right and it is stable above the cpu. i made sure it is covering the cpu perfectly, im OCD that way.

just to give you an idea of what the cooler is doing temperature wise, at idle (using core temp) my cpu is at 29-32 degrees, during bf4 gameplay it is at 50-54.
 

Dom_79

Distinguished
From personal experience I would recommend cleaning off the thermal paste and using the "pea sized/ grain of rice" squish spread method. Search a bunch of different examples of doing that to get a proper idea of the size of "little glob" that you want and where exactly you want to put it on the Haswell family of CPUs. One thing to always keep in mind with thermal paste is that bare metal to bare metal is the best heat transfer method available. So when you use thermal paste you really are just trying to fill in the very small (microscopic) surface imperfections and any slightly un-even machining in the overall surface of the CPU lid and/or the cooling block.

I did the spread method the first time myself and I had Horrible temps (also my vcore was set too high with an overclocking app that I used instead of doing the OC myself)
Here's a link to the thread (it's only a couple weeks ago): http://www.tomshardware.com/answers/id-1957442/4670k-2ghz-max-temps-question.html

Your idle temp seems high (should be a few degrees lower) but under 60C during BF4 is good (it's not a game that pushes CPUs like ours very hard. It's harder on GPUs than anything else)

BTW don't worry about "frying" your CPU until you are getting close to 100C. Running close under 80C almost all the time is what similar laptop CPUs do.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
i looked through your thread and it is helpful. I will remove the old thermal paste and re-apply now. I possibly put a little too much paste the first time round. how long should i wait to try a new overclock? because ive read (not sure where) that you need to let it sit for a couple days? Also, if i dont see cooler temps after re-applying, could the problem be with my cooler?

Edit: i am going to go out and get another 120 mm fan to attatch to the cooler to create a proper push pull configuration to help with the cooling. any suggestion on fan?
 

Dom_79

Distinguished


I looked up your specific thermal paste (gelid gc extreme) and there is no curing time. So I would say leave the computer running idle (just at the desktop) for an hour before starting the OC stress test again.

If the temps don't drop significantly after re-applying the thermal paste we can do a few things to trouble shoot the cooler, such as making sure the fan is spinning up (spinning faster) when the CPU is getting hot. If not check the settings in bios to make sure the fan can spin faster when the CPU gets hotter.

If all that doesn't help, it is possible that you just got a bad chip. Some batches of the last two generations of Intel chips which have thermal paste under the lid that is so bad that they can't be overclocked past the built in "Turbo" mode that they come with without having these really high temperatures.
Putting thermal paste between the CPU die and the lid of the chip is Intel's new way of doing things as opposed to the normal and "the way it's been done forever" way of soldering the CPU die directly to the lid.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
ok so ive now cleaned off the old thermal paste, applied a pea size amount in the middle of the chip and have also added a second fan on the cooler itself making a push pull setup which should help at load temps. as you suggested i will let my pc run for about an hour at idle and then try stress testing again at the previous settings of 4.2 GHz and ill start the voltage at 1.2V to see if it can be stable with lower temps.

Edit:

The idle temps taken by core temp is 27-29 degrees and has stayed there for the past 20 minutes.
 

Dom_79

Distinguished


You might be able to go a bit lower then 1.2V for 4.2GHz. Try 1.6V (the lower your vcore the less heat generated in the CPU)
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
I have noticed a few irregularities once I started overclocking today. The first being my CPU getting through IBT at the standard setting for 20 runs but would crash within 2 minutes of prime95. So after raising my voltage up to 1.219 so far so good as prime 95 is still working after 18 tests. The second irregularity is when the computer would crash it would always show "whea_uncorrectable_error" whereas today I started getting "clock_watchdog_timeout" and the computer would be in the blue screen and say it's collecting data and will then restart but it stops at 84% and then I manually have to restart my computer (with the other error the computer would restart itself)
Temps during IBT did not pass 78 degrees and in prime95 all cores are around 69 degrees.
Any thoughts? I've searched for the error and it has something to do with a core failing, from what u understood. Hopefully I didn't damage my CPU.
 

Dom_79

Distinguished
No you didn't damage your CPU. When systems crash due to instability from OC it doesn't do damage to the hardware (the hardware will only be damaged by setting things -like voltage- too high or letting the heat get too high for too long)

Did you ever run the memtest program I mentioned. IBT running without crashing indicates that your system was stable at the settings you had. Depending on what test you were doing in prime it may be the RAM has an issue (which is what made me think the issue was for BF4 crashing)
BTW to push your system more (I guess to test it better) you don't really want to do more passes in IBT before you raise the settings to high or very high. That will give you a better idea of stability and temperature management than doing more runs

At least your temps are down. Keeping them below 80C under full load is a good temperature.
 

dempy19

Distinguished
Dec 28, 2013
152
8
18,715
As for prime I was using the large ffts and it would crash after about 50 min. Yea I never got to do the mem test because I kept getting some pop up saying not all my memory can be used and that I should open up multiple mem tests to test all the memory. But it never started. Just thinking about it now and my CPU cooler fan is touching one of my ram sticks. But it's showing up properly as 8 gb in Windows.

As for stress testing, I ran IBT at high level for 20 runs with no problems and I have been running AIDA 64 for over 2 hours now with temps averaging 60 degrees and not going over 73 ( only one core got that high). During the IBT the temps were around 80 degrees. Not going higher than 86 in core #0.

My system right now is set at 4.2 and 1.23v during testing.