Overclocking 6600K

[-Ice]

Getting bored and thinking of re-OC'ing my 6600K to run some game benchmarks on non-OC and OC and compare. I've come across this guide on TweakTown from 2015 but I've also watched a few YT videos. My main question is what programs would be best to use today in testing the OC stability? The guide says IBT, Handbrake, and Prime95 but a video by Paul's Hardware uses wPrime, Cinebench R15, Passmark, and POV-Ray and a video by Linus suggests Aida 64 and ROG Realbench and he even recommends against using Prime95.

So what's the best program(s) to use nowadays for stress-testing a 6600K overclock? Is there a better OC guide than the one I've linked?

Thanks!

 

[Mr.Scott]

The best stress testers are the programs you use on a daily basis.
If you can do everything you normally do without BSOD or problems of some kind, then it is stable for you.

 

[-Ice]

I probably should've expected that response and I agree with you on that note. However, I'm also sure you know what I meant with my question

 

[trents]

Try the Realbench stress test. It does a good job of checking for stability in a real world way and doesn't drive temps up as high as Prime95 or OCCT. If you can pass it for 2-4 hrs. you're good to go. Set the RAM usage to 8gb. It will use the CPU and the GPU to render.

 

[-Ice]

Thanks! I just realized this question has probably been asked a thousand times so I guess it's best if I rephrase it.

I'm planning to use Aida64, Cinebench R15, IBT, and Realbench to stress-test, aiming for maybe 4.7GHz under 1.4v. Is there any reason not to use any of these stress programs? ie, Linus recommended against using Prime95 which I used in the past on my i5 750.

Thanks!

 

[trents]

IBT is a burner! Watch those temps when using it. It is also very dated, not having been abandoned by the author in 2012. Anymore, I only use it for very short tests to see if I am grossly unstable when overclocking. It offers the advantage of cutting right to the chase in that scenario, driving up temps very quickly. I use the "Very High" setting. It is also good for letting you know pretty quick if your cooling is inadequate.

 

[-Ice]

Just ran a bunch of tests and maybe I lost in the silicone lottery? The CPU will not hold 4.6GHz even up to 1.42v, it will boot at 4.5GHz at 1.35v but will fail Aida/Cinebench/IBT. At 1.41v, it will pass Aida and Cinebench but will BSOD with IBT almost straight away. I feel like going up to 1.4+ volts to hold a 4.5GHz overclock is too much, am I wrong here?

So far, I'm at 4.4GHz and IBT crashes until I get to 1.36v which according to the TweakTown guide should be 1.30v for 4.5-4.6GHz so again, too much power for lower GHz and reinforcing the idea that I lost the silicone lottery on this chip.

 

[trents]

You are correct. You did not win the silicon lottery. 1.4 vcore would be the highest I would consider for 24/7. But have you played with Load Line Calibration yet?

And you did not mention temps under load during stress testing. Are you checking that?

 

[-Ice]

No. I remember having to enter values for LLC and other power modes back when I was overclocking the i5 750 but the guides for the 6600K rarely if ever mentions any of these. I found this odd and that's why I was asking about other guides I might follow?

Temps are fine, monitored via MSI AB and HWiNFO and with a Noctua NH-D15, everything's running well cool.

 

[trents]

One reason for having to use high vcore to be stable at lower than expected overclocked frequencies is "vdroop" under load. Vdroop=Voltage droop. The purpose of LLC is to counteract that. Most people find that the next highest LLC setting is the one that comes closest to stabilizing the vcore under load. So for example, if you set the vcore at 1.4 in bios but under load it droops to 1.3 it may cause the system to be unstable. It my actually need close to 1.4 or 1.39 to be stable at that frequency. So without a sufficient LLC offset you will just have to raise the vcore so that when droop happens it won't take the vcore below 1.39. Make sense? I'm guessing you left LLC on "Auto." You need to take it off Auto and experiment with manual settings, regardless if "the guides" talk about it or not.

 

[-Ice]

I just didn't want to be messing with stuff if it wasn't mentioned for fear of breaking things. I learned a little about LLC and vdroop and other things back in the i5 750 OC days which was almost a decade ago so I wouldn't trust my own skills regarding that aspect.

Everything is at Auto. It looks like the system is currently stable at 4.4GHz OC with 1.36v. Do you think putting LLC on the next step higher or even the highest setting, I'll be able to get 4.5GHz as <1.4v? Will I need to change LLC if I decide to stick to 4.4GHz OC?

I'm thinking I just might try the auto-overclock from Gigabyte and see how far that OC's the system, if nothing but to give me an idea of how the software can OC the system.

 

[trents]

Most of the time, auto overclock genies in bios assign much more vcore than is necessary. They sometimes also lower RAM frequency.

I guess the place to start would be to watch your monitoring software and see how much vdroop you are actually getting when stress testing.

But it's your system and if you are not comfortable with adjusting certain settings, what can I say?

 

[-Ice]

That's why I'm asking for OC guides; I'm not comfortable adjusting it by myself but I can follow an OC guide.

The idea of the auto OC software was just to see how much the software can push it and then I'll see if I can get that same number manually OC'ing the system. How do I check for vdroop via the monitoring system?

 

[-Ice]

Looks like LLC goes from Auto to Standard to High. I set it on High and managed to get to 4.5GHz stable on 1.36v with Aida 64 (20 min), Cinebench R15, and IBT Standard 10 and Very High 5. Highest temps are 78C on IBT VH 5, so very manageable. It absolutely doesn't like 4.6GHz OC though, even going up to 1.40v and LLC on High, it would crash as soon as IBT Standard 10 is run.

Is there anything else I can try for 4.6GHz?

I guess I'll go back to 4.5GHz OC for now and run some longer stress test. 1hr Aida 64 and 20 min on IBT Standard then another 20 min on IBS Very High should do, yes?

 

[trents]

Install HWINfo64 and have it open while you stress test. Compare the minimum amount of vcore under load with what you have it set to in bios. Another important number is the VMOS temp. VMOS is the temp of VRM (mosfets) or the power producing section of the motherboard.

Guides are helpful but unless you can find one that exactly matches your system components (same CPU, same motherboard, same RAM) they are just general purpose helps. I get the impression you are kind of up tight about this overclocking thing. Relax dude!

AIDA64 is a good test but for it to be indicative of stability I would run it for 5 hours or so. It's not particularly intense. When I'm not quite stable with an overclock it will often fail at about the 5 hr. mark.

So you only have three LLC levels to choose from?

 

[-Ice]

What do you mean by uptight about OC'ing? Not sure what you mean, but just don't want to make a boo-boo that results in burnt electronics Unless you mean something else?

Is the data/entry on HWiNFO dependent on system components? I don't have that CPU Core Voltage line or the VR MOS line, I'm running HWiNFO v6.08-3830. I would assume these are the entries you are referring to?




Only three LLC levels, Auto, Standard, and High.

Based on temps, Aida is not very taxing on the CPU as IBT on Very High (5 runs) pushes it to 65-78C max among the four cores but 20 minutes on Aida only goes 43-58C among the four cores. On previous tests, I could have the settings pass Aida and Cinebench but fail IBT so I started testing on IBT Standard and IBT Very High first to quickly see if it was stable before running Aida and Cinebench.

 

[Mr.Scott]

Leave your LLC set to high. You still droop slightly. Actual vcore is showing roughly 1.35v under load. You have a little vcore headroom left if you can keep the temp under control.
Pretty sh1tty CPU you have there. I thought my 6700K sucked.

 

[trents]

Looks like your CPU is just not a good overclocker and you will need to be content with 4.4 ghz or maybe even a truly stable 4.3 ghz.

Yes, the actual line items that show up in HWINfo are component dependent. It turns around where the motherboard manufacturer decided to put sensors.

You are not even close to voltage levels that would result in "burnt electronics." We would not suggest taking measures that would put you at that risk. You came to us for advice in addition to the information you had gleaned from guides and I was just trying to help.

 

[-Ice]

You are not even close to voltage levels that would result in "burnt electronics." We would not suggest taking measures that would put you at that risk. You came to us for advice in addition to the information you had gleaned from guides and I was just trying to help.

Thanks for the help and that is exactly why I came here and I appreciate the suggestions. Keep 'em coming! I think we got kind of mixed up somewhere in this conversation?

I was simply saying that I found it odd how older OC guides (i5 750) were more detailed with regards to LLC and other power settings whereas most of the guides I could find for the 6600K was "just up the multiplier, give it a little juice, test for stability, and done!" I was only explaining why I have not played with LLC at that point and why I'm not playing with other settings not mentioned in the guide and not suggested to be fiddled with by anyone on here. Since you brought up LLC, I've adjusted it. Now I'm testing for stability at 4.5GHz instead of 4.4GHz. With LLC on Auto, it wouldn't even run Aida at 1.36v and would BSOD with IBT even at 1.41v. With LLC on High, it's passing all those tests on the "light" setting and 1.36v, but looks like it needs more juice for the heavier/longer tests, so thank you for that suggestion!

- - - Auto-Merged Double Post - - -

Hahaha!! Thanks for the kind words Mr.Scott!

I'm now on 1.37v vcore and vdroop is 1.356 according to HWiNFO but it was also 1.356 when I had my vcore on 1.36v. It's failing after about 10 minutes of IBT Standard, so I wonder if 1.38v or 1.39v would do? I'm disappointed in the CPU especially with you guys' reinforcement that it seems to be a major fail in the silicon lottery but hey, I was running this on stock settings until now so a gain of 600MHz shouldn't be too bad. I just need to find out where it's stable for my testing then I'll move on to testing with my usual games and see how much improvement I get.

 

[trents]

No harm no foul.

Sounds like you have made some progress by taking LLC off of Auto. Your temps are cool so if you need to increase vcore to be stable at 4.5ghz go for it . . . up to a max of 1.4. 4.5ghz is a pretty common overclock ceiling for that CPU.

Most motherboards in the class of the one you have will offer more granularity in LLC control, like 5-7 levels. I'm surprised yours doesn't.