Bad NH-D15S temps; should I try buying different paste?

[Techhog]

Before I say anything else, no, I will not even consider lapping or deliding the CPU. Please don't bother suggesting it.

I've tried repasting it, but it didn't work. I'm going to try one more time this weekend, but if that fails I'll need to look into other options. It's only at 1.25V, but in IBT it hits 92C (package temp) within two passes with fixed voltage! Encoding in Adobe Premiere hits 72C on the hottest core, and in gaming it sometimes gets as high as the mid 60s. That seems way too high for this cooler, which should be able to handle this CPU at 1.3V. :/ If trying it that last time fails, I'm going to have to look into other options. A second fan is one of those options, but paste is another one I'm thinking about.

What do you guys think? Am I bad at mounting coolers, or did I get a bad chip, or is something else wrong? My graphics card is running at temps consistent with reviews, so I doubt that it's an airflow issue. Even is it is, I'd need to buy a new PSU to add another intake due to the cables blocking the bottom fan mount. I'm not sure if rewiring everything would help.

 

[Seba]

You used the included (with your cooler) Noctua NT-H1 thermal paste? I do not see a point in buying a different paste, if that is the case.

 

[Techhog]

You used the included (with your cooler) Noctua NT-H1 thermal paste? I do not see a point in buying a different paste, if that is the case.

Yes, that's what I used. It hard to apply though. :/ I guess that makes a second fan my next option...

 

[lehtv]

Of course IBT runs hot, it's IBT. And of course not-delidded 4790K runs hot, it's got HT ramping up power consumption. If reseating the cooler doesn't help, there's really nothing you can do apart from delidding - which I won't suggest since you asked not to

Your Adobe temps are more than fine. Stay under 80C in real world 100% load and you're fine.

 

[Techhog]

Of course IBT runs hot, it's IBT. And of course not-delidded 4790K runs hot, it's got HT ramping up power consumption. If reseating the cooler doesn't help, there's really nothing you can do apart from delidding - which I won't suggest since you asked not to

Your Adobe temps are more than fine. Stay under 80C in real world 100% load and you're fine.

I know the temps are fine with taken on their own but at this voltage? I can't push any farther with these temps which seems off compared to other people. Hopefully it's just that I used a bit too much paste.

 

[lehtv]

But 4.6GHz is definitely a good overclock for an air cooled, non-delidded Haswell i7. I would not expect any better results without delidding. It's to be expected with the poor quality TIM under the IHS that you'll sooner run into temperature limits than voltage limits.

Also, with 72C in Premiere, you still have some thermal headroom.

 

[Techhog]

Looking at more results, I guess my temps are correct. That said, I'm going to repaste it again still just because Premiere was only getting up to 68C when I first installed the cooler.

 

[ClockHound]

How much paste was applied? Less is more with TIM. Did you use the 'pea-sized' dot method or other?

 

[Techhog]

How much paste was applied? Less is more with TIM. Did you use the 'pea-sized' dot method or other?

Did pea-sized just now. No difference and in fact it's even much hotter (4C) initially (though I think the paste needs a little time to spread and the room is a bit hotter).

I'm blaming the TIM under the IHS at this point. I've sees the coolest core be up to 12C cooler than the hottest for short bursts. So, I may have a decent chip but the middle cores have a gap or something. That's just my luck! Oh well...

EDIT: and Premiere crashed... I really hope that doesn't mean that my overclock is unstable...

 

[lehtv]

By the way, overclocking a 4790K on Z87 Pro3 isn't the best idea I can think of... Z87 Pro4 apparently has 3 true phases with average quality MOSFETs (see vrmlist), and I'm thinking Pro3 is the same or worse. Hopefully you won't squeeze the board dry too quickly.

 

[Techhog]

By the way, overclocking a 4790K on Z87 Pro3 isn't the best idea I can think of... Z87 Pro4 apparently has 3 true phases with average quality MOSFETs (see vrmlist), and I'm thinking Pro3 is the same or worse. Hopefully you won't squeeze the board dry too quickly.

Yeah, I thought that might be the case. Do you think I should just leave it at stock and call it a day then?

I'm hoping that 1.25V isn't too stressful, though I was looking into lowering it a bit. I'm going to try again tomorrow. If you really think that I shouldn't OC at all on this board though, I'll drop it. It's already 2 years old and had an OC'd 4670K for most of that time.

 

[lehtv]

Given that 4790K is already 4.4Ghz max turbo by default, your 4.6GHz "overclock" is so minimal that it really doesn't make a difference. It's different with a 4670K where the stock speed is much lower- you have more overclock headroom thus more incentive to overclock.

I would probably just find the minimum voltage required for 4.4GHz on all 4 cores.

 

[Techhog]

Given that 4790K is already 4.4Ghz max turbo by default, your 4.6GHz "overclock" is so minimal that it really doesn't make a difference. It's different with a 4670K where the stock speed is much lower- you have more overclock headroom thus more incentive to overclock.

I would probably just find the minimum voltage required for 4.4GHz on all 4 cores.

Funny story about that... It's not letting me reduce the voltage to less than 1.25V! It's completely ignoring what I set! 1.25V also happens to be the stock single-core voltage for the 4.4GHz turbo, so it might be related...

And yes, I know that the OC doesn't make much difference. I was just doing this for fun and to match a stock 6700K. How dangerous do you think this is on a scale of 1-10?

 

[Techhog]

Wait wait wait

My ring voltage is set to auto, and it's at 1.331V. Doesn't that seem wrong?

 

[Techhog]

Okay, so I lowered the ring voltage, but it looks like the adaptive core voltage is set at 1.25V with spikes to 1.26V no matter what I do unless I set everything back to stock. If that's the case, I might as well leave it alone. FIVR should drastically reduce any risk anyway, and it's not like 1.25V is an extreme voltage anyway. Even at stock I'd be getting spikes to that, so I don't see a reason to worry.

 

[ClockHound]

If you still see a 12c swing between cores at stock, the chip is out of spec and is cause for a RMA.

 

[MongGrel]

Still just using an old X5680 with a NH-D14 myself, it didn't sound bad till you said one core was swinging 12C out of range.

That might be something to worry about.

I'm not familiar with the newer chips, but that is a lot of variance.

Of course it is newly installed, should require time to even seat right even.

might let it burn in a bit and check to see if continues.

 

[lehtv]

The 12C difference could also result from poorly applied paste, e.g. covering only 2/3 of the IHS.

 

[Techhog]

The 12C difference could also result from poorly applied paste, e.g. covering only 2/3 of the IHS.

That would imply that I messed it up in the exact same way 3 times in a row. I can also safely say that it was covered the first two times. There's a different reason. The 12C difference only happens as the censors are updating anyway. The difference in peaks is usually 9c, though in IBT all of the cores reached the same temp. That implies a faulty censor, doesn't it?

Also, I should mention that the coolest core is the 4th one (the only one that's not sandwiched between anything).

 

[lehtv]

I doubt there's anything wrong with the sensors. Where do you see the the 9C difference, if not in IBT? Maybe it's not actually stressing all cores equally

 

[Techhog]

I doubt there's anything wrong with the sensors. Where do you see the the 9C difference, if not in IBT? Maybe it's not actually stressing all cores equally

In... Well, pretty much everything except idle. RealBench, Premiere, Cinebench, gaming, OCCT, even Prime95 26.6 iirc (I haven't run that in a while). 7-9C. For example, here are my RealBench peak temps:

#0: 70
#1: 74
#2: 70
#3: 65

I know that there can be up to 10C difference within spec though, so calibration might just be slightly off. My guess is that the last core is on the lower end and is off by -3 or -4C (probably not -5, since that core being a bit cooler makes some sense on its own).

 

[guskline]

In the "for whatever it's worth department" I installed IBT v2.54 in my 4790k rig below and ran it measuring temps via AIDA64. My 4790k is running 4.7Ghz with fixed vcore of 1.305 on an Asus Maximus VII Hero mb.

After 20 runs the MAX core temps were 77C-80C-80C-77C.

IBT is known to run hot.

AIDA64 stress test never exceeded 69C

I have a custom water cooled rig with an EK Supremacy MX block.

 

[Techhog]

Here's another RealBench run, only this time I actually set my fan controller on high! (Parenthesis is comparison to how it was with fans on low)

69 (-1)
71 (-3)
71 (+1)
65 (0)

That's kinda interesting...

EDIT: This might be slightly interesting as well:

EDIT 2: Even more interesting is the fact that the HD 4600 is running at full boost even though it's not being used right now. That seems really weird... (I know that I can just disable it, but I have it enabled for a reason).

In the "for whatever it's worth department" I installed IBT v2.54 in my 4790k rig below and ran it measuring temps via AIDA64. My 4790k is running 4.7Ghz with fixed vcore of 1.305 on an Asus Maximus VII Hero mb.

After 20 runs the MAX core temps were 77C-80C-80C-77C.

IBT is known to run hot.

AIDA64 stress test never exceeded 69C

I have a custom water cooled rig with an EK Supremacy MX block.

Yeah, I guess my temps make sense in comparison to custom water in this case.

 

[Techhog]

If you still see a 12c swing between cores at stock, the chip is out of spec and is cause for a RMA.

At stock I think the highest I saw was 10 or 11, but they change so fast that I might have missed something.


With the overclock, I saw a difference of 15C at one point. How important is RMAing it in this case? Should I try to get a second fan for the cooler first to eliminate the possibility of uneven cooling due to one of the towers being "fanless"? (For those who don't know, the fan is in the middle and acts as a pull for one tower and a push for the other, Also of note is the fact that the tower getting the "push" also has my rear fan almost acting as a pull, and I can feel a lot of hot air coming out of the rear exhaust, and I mean A LOT.)

 

[MongGrel]

Wow, I was not even familiar with a 15S.

I was thinking you had one of these on there.

http://www.amazon.com/Noctua-NH-D15...e=UTF8&qid=1449942460&sr=8-2&keywords=nh+d15s

I guess if it is a twin tower just sticking another 140 would be the same.

I'd still not think that would be the problem if one core jumping out of range.

I'll just bow out as I'm not familiar with that CPU.