I delidded 6700K. Allow me to share my experience. (It didn't work out that well--until it did.)

[Guest]

hey, I have that case - love it! 

[knightslugger]

14 hours ago, done12many2 said:

You have some thermal overhead to go a little further.

I wonder how much voltage overhead i actually have... i hit 4.7GHz with adaptive set to 1.370v +.015v seems stable.

 

...seems... I haven't had much time to sit and let it burn. bought a different PSU since my SeaSonic is in for RMA. Bought a Corsair unit instead. CXM series (T3).

[Business Corgi]

7 hours ago, done12many2 said:

 

I hate to break this to you, but your AIDA64 actually crashed.  I say this for two reasons.  Firstly, AIDA64 won't let you close during a stress test without stopping the test first.  Secondly, the red warning area with the information stating "stress test session terminated unexpectedly" only appears when it crashes.

 

You might need more Vcore to stabilize that overclock.  Also, try hitting it with a variety of tests just to make sure that you're good.  It only needs to be as stable as you need it to be, but I've found that AIDA64 is just way too easy to "pass". 

 

Good luck man.

I went and tested this because of your post. Removed the OC including XMP. I get the same results if I start AIDA64 and interrupt the stress test after just a few seconds. 

 

It only happens if I "X" out of AIDA64 without stopping the test. That's what comes up after I reopen Aida64 after stopping it unexpectedly. 

 

I've validated for almost a half a day in AIDA64 and real bench at 4.7 and ~1.415V per the above picture and the only way I can reproduce the above error is by closing Aida64 while it's running. I can make a YouTube video to corroborate

[done12many2]

17 minutes ago, Ducks_McGoo said:

I went and tested this because of your post. Removed the OC including XMP. I get the same results if I start AIDA64 and interrupt the stress test after just a few seconds. 

 

It only happens if I "X" out of AIDA64 without stopping the test. That's what comes up after I reopen Aida64 after stopping it unexpectedly. 

 

I've validated for almost a half a day in AIDA64 and real bench at 4.7 and ~1.415V per the above picture and the only way I can reproduce the above error is by closing Aida64 while it's running. I can make a YouTube video to corroborate

 

That's interesting because I can't even close the test without hitting stop first. It's actually a feature that stops you from accidentally closing it.

[Business Corgi]

Welp, maybe it's a setting or something. Check it out. 

 

 

[done12many2]

3 minutes ago, Ducks_McGoo said:

Welp, maybe it's a setting or something. Check it out. 

 

 

 

Thanks for sharing.  It's because you're closing the entire program.  The stress test window itself shouldn't let you accidentally close out though.  Either way, nice overclock and keep pushing it!!

 

Also, you should run your test without the memory test selected.  Just run CPU, FPU and cache.  This causes the CPU to get battered a little more than if you test memory at the same time.  Memory should be tested on it's own with something a lot better than AIDA64.

[Business Corgi]

Yeah, it was an accident before. Womp Womp. That's really good to know. Do you think running without the ram test will lower the temps in the results by a significant amount.

 

You mentioned something better than AIDA64 for testing memory. Did you have any suggestions?

 

I really couldn't be happier with the results. I still feel like I got gonged on the CPU. Should it need 1.415V to reach 4.7Ghz? That seems excessive. I could be wrong but man, I'm not used to pushing past 1.4V

 

[iamdarkyoshi]

19 minutes ago, done12many2 said:

 

Thanks for sharing.  It's because you're closing the entire program.  The stress test window itself shouldn't let you accidentally close out though.  Either way, nice overclock and keep pushing it!!

 

Also, you should run your test without the memory test selected.  Just run CPU, FPU and cache.  This causes the CPU to get battered a little more than if you test memory at the same time.  Memory should be tested on it's own with something a lot better than AIDA64.

4.7 wasn't quite stable for me at 1.375, but I'm scared to go any further. Temps are good though

[done12many2]

12 minutes ago, Ducks_McGoo said:

Yeah, it was an accident before. Womp Womp. That's really good to know. Do you think running without the ram test will lower the temps in the results by a significant amount.

 

You mentioned something better than AIDA64 for testing memory. Did you have any suggestions?

 

I really couldn't be happier with the results. I still feel like I got gonged on the CPU. Should it need 1.415V to reach 4.7Ghz? That seems excessive. I could be wrong but man, I'm not used to pushing past 1.4V

 

 

Running AIDA64 with CPU, FPU and cache testing only will actually raise the temperature as it's hammering the CPU more than if you include memory testing.  That's why I recommend that you do it.  Even with CPU, FPU and cache only testing, AIDA64 is way too easy to pass so enabling memory only makes it easier which is not a good thing.  

 

As far as the overclock itself, it seems pretty fair.  I've provided you a link to an overclocking statistics thread over on OCN.  Darkwizzie manages this thread as well as similar threads for other Intel mainstream CPUs.  I just wish he hadn't made the stress test requirements harder for Kaby Lake.  Doing so surely skewed direct comparisons between Skylake and Kaby Lake stable OCs. 

 

Submit your results after testing.  It only helps the community.

 

http://www.overclock.net/t/1570313/skylake-overclocking-guide-with-statistics/0_100

[done12many2]

7 minutes ago, iamdarkyoshi said:

4.7 wasn't quite stable for me at 1.375, but I'm scared to go any further. Temps are good though

 

Don't be scared of voltage.  

 

 

 

Seriously though, it's more than just voltage.  It's about how much current you're running through the CPU.  If you aren't doing a lot of heavy AVX loads or similar power draws, higher voltage is not an issue.  So much misinformation about voltage in this forum.  Voltage is nothing more than potential.   

 

No, I don't run mine at this voltage daily, but I'm at 1.4v + all day long. 

[iamdarkyoshi]

1 minute ago, done12many2 said:

 

Don't be scared of voltage.  

 

 

 

Seriously though, it's more than just voltage.  It's about how much current you're running through the CPU.  If you aren't doing a lot of heavy AVX loads or similar power draws, higher voltage is not an issue.  So much misinformation about voltage in this forum.  Voltage is nothing more than potential.   

I believe my bios allows me to set a current limit

[done12many2]

9 minutes ago, iamdarkyoshi said:

I believe my bios allows me to set a current limit

 

This is a good thing.  

 

Intel / Asus has taken things a step further by enabling a few extra perks with Kaby Lake.  Intel brought AVX offset over from the server chips, which is great for folks that want to run super high clocks/voltage, while reducing their concerns for accidentally running an AVX load through the chip with too much voltage.  All you do is set the offset you want and it's automatically triggered once the CPU detect AVX instructions.  So if you are running a 5.1 GHz overclock with an AVX offset of say 3, once the CPU detects AVX, the overclock will automatically drop to 4.8 GHz and the voltage will drop as well based on the VID curve.

 

The other thing that Asus did was they added CPU temperature control to the BIOS.  They had it previously in a software version, but most of us didn't like software over firmware.  You simply set a cap on CPU temperature.  Then you set a multiplier and voltage for it to drop to once that cap is reached.   The last thing you set is the minimum temp at which the CPU will scale back up to maximum overclock.  It's essentially a user adjustable version of throttling and works great.

 

With all of that said, I don't use either of these features as my chip runs nice and cool at my daily clocks, but they are definitely nice to have.

[Business Corgi]

Advice taken on the AIDA64 stress test. Coming soon: screenshot of revised AIDA64 stress test. 

 

Screenshot of an x264 stress test from the provided link (following the directions from the thread/read me file from the modified x264 test). 

 

And a screenshot of Realbench. I want to see what happens to the CPU temps when the GPU dumps heat into the case. 

 

All tests done for an hour. The Realbench one is going right now. If temps don't go too crazy and it doesn't crash I'm going to... Well.... What should I do? 

 

Do you think I should push further? 4.8Ghz with similar vcore crashed immediately. 

 

If you think this is the best I can do I'll leave the x264 stress test running overnight and use these settings as my daily OC. 

[done12many2]

3 minutes ago, Ducks_McGoo said:

Advice taken on the AIDA64 stress test. Coming soon: screenshot of revised AIDA64 stress test. 

 

Screenshot of an x264 stress test from the provided link (following the directions from the thread/read me file from the modified x264 test). 

 

And a screenshot of Realbench. I want to see what happens to the CPU temps when the GPU dumps heat into the case. 

 

All tests done for an hour. The Realbench one is going right now. If temps don't go too crazy and it doesn't crash I'm going to... Well.... What should I do? 

 

Do you think I should push further? 4.8Ghz with similar vcore crashed immediately. 

 

If you think this is the best I can do I'll leave the x264 stress test running overnight and use these settings as my daily OC. 

 

Good initiative on taking those tests on!  Your stability will thank you.  

 

An hour for each is fine, but what really matters is what makes you feel comfortable with your stability?  You've already done well more than most people do so enjoy your OC!

[Business Corgi]

I have to remember not to bury the lead here, that I couldn't get a reliable 4.5Ghz OC at 1.335V without hitting 80°C even in the most forgiving tests. 

 

It really makes me wish Linus and Co. would revisit the delidding question and this time do it right.

[done12many2]

Just now, Ducks_McGoo said:

I have to remember not to bury the lead here, that I couldn't get a reliable 4.5Ghz OC at 1.335V without hitting 80°C even in the most forgiving tests. 

 

It really makes Linus and Co. would revisit the delidding question and this time do it right.

 

 

Too many people have said this, but I heard that he stated that he would not.  Oh well.  You can lead a horse to water, but you can't...

 

@MageTank

[Business Corgi]

I can see maybe on one hand he doesn't want to be seen as being responsible for a bunch of people ruining their processors. On the other hand I have a huge problem with him closing the book on a topic with the final word being a huge bit of misinformation. 

 

It makes me question other things from LTT and frankly it isn't the first time I've had contradictory results to their findings. E.G. My Evolv ITX. Linus: "This case has amazing airflow." I love the case but it does NOT have amazing airflow. The front and top are choked off with tiny slits/holes for air to come thru or blow out. 

[Ensho]

Looking at that voltage...I would say done12many2 may have drunk12many2

 

Good to see your results man.  I am putting of de-liding as a last resort..though I know sooner or later I am going to do it as well.

 

For all the hate that Kabylake gets at times...it is still amazing to me I am stable at 4.5gzh at 1.175v and it takes Skylake cpus 1.35v or so to get there.  Even though ipc is the same...Intel did manage to optimize the core well.  Just too bad they gimped the tim.

[stconquest]

31 minutes ago, Ensho said:

Looking at that voltage...I would say done12many2 may have drunk12many2

 

Good to see your results man.  I am putting of de-liding as a last resort..though I know sooner or later I am going to do it as well.

 

For all the hate that Kabylake gets at times...it is still amazing to me I am stable at 4.5gzh at 1.175v and it takes Skylake cpus 1.35v or so to get there.  Even though ipc is the same...Intel did manage to optimize the core well.  Just too bad they gimped the tim.

If the temperature variation between cores is similar (+/- 2 degrees, or so), you might not need to delid... unless you decide you want to compete for bench scores or something.

 

You seem to have a good chip. 

[Business Corgi]

Nothing had a problem passing after an hour until Realbench. Realbench bluescreened on me.

 

I added a hair to the vcore in the UEFI (.005) and it passed the benchmark for an hour. I don't actually see a difference in the vcore in HWmonitor though. I wanted to tax the GPU too because I wanted to see what would happen to the CPU temps once a bunch of heat got dumped into the case.

[Ensho]

21 minutes ago, stconquest said:

If the temperature variation between cores is similar (+/- 2 degrees, or so), you might not need to delid... unless you decide you want to compete for bench scores or something.

 

You seem to have a good chip. 

Yeah...it is ok I think.  It can run Cinebench at 5.1ghz with 1.4v or so.  I have one core that likes to run 4c-6c cooler than the others.  No matter how I adjust voltage settings it always ends up the same.  Every once in a while it may end up 2c-3c cooler than the others..but just that one is like that.  The other three are 1c-2c separated at worse.

 

I go back and forth between running stock with low volts and overclocking because Gigabytes LLC drives me nuts.  It is terrible.  In the next month or so if they don't fix it with their next bios release I am going to get the Maximus IX Hero.  I prefer adaptive voltage/clocks overclocking because I don't see the need to run the cpu at max speed and max volts all the time.  Speed Shift 2.0 has a 10ms response time so performance doesn't suffer.  Overall I think it is a better way.  Gigabyte just does not have a firm grasp on this method as of yet though.

[stconquest]

1 minute ago, Ensho said:

Yeah...it is ok I think.  It can run Cinebench at 5.1ghz with 1.4v or so.  I have one core that likes to run 4c-6c cooler than the others.  No matter how I adjust voltage settings it always ends up the same.  Every once in a while it may end up 2c-3c cooler than the others..but just that one is like that.  The other three are 1c-2c separated at worse.

 

I go back and forth between running stock with low volts and overclocking because Gigabytes LLC drives me nuts.  It is terrible.  In the next month or so if they don't fix it with their next bios release I am going to get the Maximus IX Hero.  I prefer adaptive voltage/clocks overclocking because I don't see the need to run the cpu at max speed and max volts all the time.  Speed Shift 2.0 has a 10ms response time so performance doesn't suffer.  Overall I think it is a better way.  Gigabyte just does not have a firm grasp on this method as of yet though.

Well, it is better than my 13+ degree variance on my 3570K. 

[Business Corgi]

It's been two weeks and we're still going strong! Before the liquid metal I used Noctua and MX4 TIM on my delidded die. Any thermal benefit I would have gotten from this would be looooooooooooooong gone by now.

 

Linus really ought to redo the video about this.

 

I also just made the switch to a blower style cooler which I think might improve my cpu temps since I'm using an MITX case.