• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

6700k Core #1 around 30* cooler than the rest?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

surensm

Registered
Joined
Feb 10, 2017
This doesn't add up. OC'd to 4.8 ghz, DVID set to 0.015V. I was planning on stress testing for some time, then call it done 30 min into prime95 for the max setting (4.8GHz). It will be dropped down to 4.6 for daily driving, but I did notice something very odd:

7QS399W.png


I'll be doing this a few more times to see if I can see some patterns, but so far it's doing ok in OCCT. Maybe linpack will help the weaknesses surface?
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
Well, haven't been able to replicate this. Hopefully this was a glitch in HWMonitor, and I already dialed DVID down to 0.00 and core clock to 4.7 (so far it's happy).
 

trents

Senior Member
Joined
Dec 27, 2008
My suspicion is that during the original Prime95 stress test one of the working threads dropped out because it generated an error but the others kept going. Make sure all threads are visible in Prime95. Used to be that when one thread threw an error the little Prime95 icon at the top of that thread window would turn from green to read. In the newer versions of Prime they don't turn color and it's easy to miss.

I've started using OCCT instead of Prime95 as when there is an error you get a "quack" sound and the whole thing just terminates. The intensity of the OCCT stress test when using Large Data Set is pretty much the same as it is in Prime95 and there are temp and voltage graphs displayed. Plus, you can set it to run for a given amount of time which I really like. There is also a Linpack with AVX instruction set stress testing option in OCCT. I run the OCCT large data set for 3 hr. to confirm stability.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
My suspicion is that during the original Prime95 stress test one of the working threads dropped out because it generated an error but the others kept going. Make sure all threads are visible in Prime95. Used to be that when one thread threw an error the little Prime95 icon at the top of that thread window would turn from green to read. In the newer versions of Prime they don't turn color and it's easy to miss.

I've started using OCCT instead of Prime95 as when there is an error you get a "quack" sound and the whole thing just terminates. The intensity of the OCCT stress test when using Large Data Set is pretty much the same as it is in Prime95 and there are temp and voltage graphs displayed. Plus, you can set it to run for a given amount of time which I really like. There is also a Linpack with AVX instruction set stress testing option in OCCT. I run the OCCT large data set for 3 hr. to confirm stability.

Thanks, will do! I usually run OCCT, but was under the impression it wasn't as hard on the components as p95. Good to know that it's basically the same.
 

trents

Senior Member
Joined
Dec 27, 2008
I should clarify what I said. The OCCT Large Data set test gives approximately the same intensity of stress test as an older version of Prime95 with AVX1 like, for instance, v. 27.7. The newer versions of Prime95 with AVX2 instruction sets make way more heat.

I think you will find that if you have enough vcore to pass 3 hr. of OCCT large data set you will not have any problem with the system being unstable in real life apps.
 

Tgrable

Member
Joined
Feb 8, 2017
Location
Minneapolis, MN
I'm not sure no one else has pointed this out, but there is no way I would recommend running at those temps regardless You are over 90c and hitting 100c max on a few of the cores. While these chips don't throttle until 100c being in that area also has a high risk of degrading your chip. I would try to get temps under 80c if at all possible.

Not sure what Cooler you are using, but if you want to continue on this path I would definately upgrade it.

Also to the temp variance it could be several things. One being P95 errors on 1 or both of the threads for that core.. so temps would drop. Second it could be something as simple as a faulty sensor or having several temp programs running at once which are causing issues with one another. Lastly it could be the application of your TIM or a issue with the gap/TIM between die and IHS. I have personally encountered all 3 of these issues at one point or another and can attest to them causing this type of issue. The latter of the 3 though I only noticed about a 12c difference on though.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
I should clarify what I said. The OCCT Large Data set test gives approximately the same intensity of stress test as an older version of Prime95 with AVX1 like, for instance, v. 27.7. The newer versions of Prime95 with AVX2 instruction sets make way more heat.

I think you will find that if you have enough vcore to pass 3 hr. of OCCT large data set you will not have any problem with the system being unstable in real life apps.

Thank you, sir. I have to do some nore adjustments to the vcore since I keep getting BSODs in the middle of p95 (I thought p95 would crash before a bsod, nice to be in the ballpark for once), but I will do that.

I'm not sure no one else has pointed this out, but there is no way I would recommend running at those temps regardless You are over 90c and hitting 100c max on a few of the cores. While these chips don't throttle until 100c being in that area also has a high risk of degrading your chip. I would try to get temps under 80c if at all possible.

Not sure what Cooler you are using, but if you want to continue on this path I would definately upgrade it.

Also to the temp variance it could be several things. One being P95 errors on 1 or both of the threads for that core.. so temps would drop. Second it could be something as simple as a faulty sensor or having several temp programs running at once which are causing issues with one another. Lastly it could be the application of your TIM or a issue with the gap/TIM between die and IHS. I have personally encountered all 3 of these issues at one point or another and can attest to them causing this type of issue. The latter of the 3 though I only noticed about a 12c difference on though.

Yeah I was running stupid high voltages because I'm still figuring out DVID on this mobo. With the bios update, DVID doesn't work the same, so I'm back to using regular voltage adjustment.
 

trents

Senior Member
Joined
Dec 27, 2008
All of the stress testing programs can trigger BSOD sometimes. In myexperience, that is most likely to happen when vcore is significantly lower than needed for stability when running the stress test you are using. If the vcore is close to providing stability then you might get system lockup or just notification in some way from the testing program itself that you failed it.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
All of the stress testing programs can trigger BSOD sometimes. In myexperience, that is most likely to happen when vcore is significantly lower than needed for stability when running the stress test you are using. If the vcore is close to providing stability then you might get system lockup or just notification in some way from the testing program itself that you failed it.

Thanks for the info! Some more news: 1.425V and 4.7ghz so far, OCCT failed at 30 min mark due to error in core 3. Another thing I noticed between this CPU and the one that was replaced is that the stress testing apps can actually communicate with it to figure out where the error was (which core, sometimes even the cause of the error). That last CPU would just error out, and that's it. P95 would not even bother running for more than 5 seconds.
 

wingman99

Member
Joined
Dec 10, 2003
You are fine at 90c it will throttle the clock speed at 100c to keep it in the safe limit and Intel will shut down the processor 120c to prevent damage.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
You are fine at 90c it will throttle the clock speed at 100c to keep it in the safe limit and Intel will shut down the processor 120c to prevent damage.

Thank you! Just had this happen again, but at 4.6 ghz. Same difference, just lower temps.
 

trents

Senior Member
Joined
Dec 27, 2008
I would not run the CPU 24/7 at anything greater than 1.4 vcore. I think you need to back off on your overclock. It doesn't look like you won the silicon lottery and you need to find the sweet spot for that CPU. You might need to settle for 4.4-4.5 ghz for daily use.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
I would not run the CPU 24/7 at anything greater than 1.4 vcore. I think you need to back off on your overclock. It doesn't look like you won the silicon lottery and you need to find the sweet spot for that CPU. You might need to settle for 4.4-4.5 ghz for daily use.

Yeah, it's not looking like I'll be able to run anything more than that. I also kept getting an error at core #3 in OCCT, and after updating to 4.5.0, it stopped testing altogether. Prime95 still works, though, but it keeps finding errors as well. One bad CPU is just bad luck. Two bad CPUs can still be attributed to bad luck, but three? Either I'm cursed, or the quality control at Intel has dropped off...
 

Tgrable

Member
Joined
Feb 8, 2017
Location
Minneapolis, MN
Yeah, it's not looking like I'll be able to run anything more than that. I also kept getting an error at core #3 in OCCT, and after updating to 4.5.0, it stopped testing altogether. Prime95 still works, though, but it keeps finding errors as well. One bad CPU is just bad luck. Two bad CPUs can still be attributed to bad luck, but three? Either I'm cursed, or the quality control at Intel has dropped off...

I would go with option 3 on that. Intel hasn't had real competition for a while and haven't really been pushed to be better. Ryzen is close enough it should help with that though.

The problem with these chips is the cheap TIM and Uneven silicon bead they run to secure the IHS to the chip. That is why people who delid are seeing such massive gains. I know on my own delid the silicon looked excellent but I had huge air pockets in the TIM they used.

So while I would never recommend that anyone delids because of the inherent risks involved it does give you a little more thermal headroom and that headroom can make a previous unstable OC stable just because of how temps can affect leakage etc.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
I would go with option 3 on that. Intel hasn't had real competition for a while and haven't really been pushed to be better. Ryzen is close enough it should help with that though.

The problem with these chips is the cheap TIM and Uneven silicon bead they run to secure the IHS to the chip. That is why people who delid are seeing such massive gains. I know on my own delid the silicon looked excellent but I had huge air pockets in the TIM they used.

So while I would never recommend that anyone delids because of the inherent risks involved it does give you a little more thermal headroom and that headroom can make a previous unstable OC stable just because of how temps can affect leakage etc.

Right, delidding is a whole another ball game that I'm about as familiar with as I am with underwater welding (I can weld, but definitely not underwater lol). Is the silicone something that can be reapplied?
 

Tgrable

Member
Joined
Feb 8, 2017
Location
Minneapolis, MN
Right, delidding is a whole another ball game that I'm about as familiar with as I am with underwater welding (I can weld, but definitely not underwater lol). Is the silicone something that can be reapplied?

Yes, I used RTV Gasket maker for example, but others have just used the locking mechanism to hold things in place or even resorted to super glue (Would not advise the use of super glue though).

Then again delidding is not for the faint of heart and if done incorrectly can render a chip useless. So if you do decide to go that route make sure to get the right tools and do plenty of research.
 

wingman99

Member
Joined
Dec 10, 2003
Yeah, it's not looking like I'll be able to run anything more than that. I also kept getting an error at core #3 in OCCT, and after updating to 4.5.0, it stopped testing altogether. Prime95 still works, though, but it keeps finding errors as well. One bad CPU is just bad luck. Two bad CPUs can still be attributed to bad luck, but three? Either I'm cursed, or the quality control at Intel has dropped off...

Overclocking is not guarantee, that is why it's called overclockign. Average overclock for skylake is 4.5GHz. overclocking quality control :rofl:, Intel does not give a s*** about overclocking.
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
Overclocking is not guarantee, that is why it's called overclockign. Average overclock for skylake is 4.5GHz. overclocking quality control :rofl:, Intel does not give a s*** about overclocking.

Never said "overclocking quality control", go back and read my original comment. I was talking about overall quality control.

Yes, I used RTV Gasket maker for example, but others have just used the locking mechanism to hold things in place or even resorted to super glue (Would not advise the use of super glue though).

Then again delidding is not for the faint of heart and if done incorrectly can render a chip useless. So if you do decide to go that route make sure to get the right tools and do plenty of research.

Yeah, going to pass on delidding, and just stick to 4.6 ghz for now. Thanks!
 

wingman99

Member
Joined
Dec 10, 2003
Never said "overclocking quality control", go back and read my original comment. I was talking about overall quality control.



Yeah, going to pass on delidding, and just stick to 4.6 ghz for now. Thanks!

What was bad about the 3 CPUs?
 
OP
surensm

surensm

Registered
Joined
Feb 10, 2017
What was bad about the 3 CPUs?

Not sure about the specific issues with this current one (assuming there is one, but I'm getting quite a few errors and one of the cores occasionally doesn't want to function), but the one before this was about as stable as an elephant trying to do a vertical on a mattress, and the first one was constantly overheating. Mind you, I never had issues overclocking with a 775 socket back in the day, and had a stable overclock all day on both my Pentium Q and my Core 2 Quad.