6700k Overclocking and what additional heat to expect. (And more questions.)

[Mina Darsh]

Heya, I got a small system with beefy stuff in it and I'm kind of curious about overclocking, mostly, what increments of heat to expect.

 

Let me first lay down the situation; I got a small form factor build where the cooler can be no higher than 58mm, my signature has the details on the stuff used. (And there's MX-4 cooling paste on the CPU.)

At stock clocks, the system becomes 78°C in CPU heavy games without dust filters attached, and 86°C at most in AIDA64's stress test. So yeah, not much headroom. (In fact, with dust filters it becomes as high as 92°C in AIDA64.)

 

What surprised me is that Salazar Studios made a mini-ITX build in a Fractal Design Node 202, with an i5 7600K at 4.5 GHz and a Cryorig C7 on top. His temps were 75°C at most... So I'm wondering if the C7 is better or is the Kaby Lake i5 just less hot despite being 91 Watt TDP as well? Or maybe both?

 

I'm planning on going all out and get a Dynatron K129 and mount a fan on it, kind of like what Linus did to keep his 140 Watt TDP CPU cool in the Dan A4-SFX, but I can't seem to find any fan wire clamps for 92mm fans... (Also no idea if the backplate fits on my motherboard.)

 

Anyway, back to the subject, what I wonder for now, as my CPU somehow runs at 1.328v when flexing is if an overclock without adding extra voltage could be done (of course depending on chip quality, so would have to find out myself), and, how many additional degrees at average can be expected per 100 MHz added?

[TubsAlwaysWins]

I wouldn't even OC if your CPU is hitting 92C in games at stock

[Mina Darsh]

2 minutes ago, TubsAlwaysWins said:

I wouldn't even OC if your CPU is hitting 92C in games at stock

In games it gets 86°C at most.

[needlessly_weekday]

2 minutes ago, Mina Darsh said:

In games it gets 86°C at most.

Even then, I wouldn't OC it unless you get a better cooler.

[WoodenMarker]

Are you running voltage at Auto? --that would use more voltage than necessary and cause higher temps. 

[Mina Darsh]

1 minute ago, spacemen13579 said:

Even then, I wouldn't OC it unless you get a better cooler.

Hmm, a better cooler it is then, been Googling for comparisons, but it's hard to get results for where the Cryorig C7 is compared to the Argon AR06, or others like the Thermalright AXP-100.

[Mina Darsh]

1 minute ago, WoodenMarker said:

Are you running voltage at Auto? --that would use more voltage than necessary and cause higher temps. 

I think I am, I believe the recommended voltage for the 6700K shouldn't even be above 1.3 Volts... What would be the recommended setting on an Asus board?

[needlessly_weekday]

2 minutes ago, Mina Darsh said:

Hmm, a better cooler it is then, been Googling for comparisons, but it's hard to get results for where the Cryorig C7 is compared to the Argon AR06, or others like the Thermalright AXP-100.

But keep in mind that this case has restricted airflow; so a better cooler may not benefit much. The heat is not being removed from the system.

[WoodenMarker]

1 minute ago, Mina Darsh said:

I think I am, I believe the recommended voltage for the 6700K shouldn't even be above 1.3 Volts... What would be the recommended setting on an Asus board?

There's no recommended setting. Try manually setting it to 1.3v and keep lowering it from there in small increments while testing for stability. 

[Mina Darsh]

3 minutes ago, spacemen13579 said:

But keep in mind that this case has restricted airflow; so a better cooler may not benefit much. The heat is not being removed from the system.

Yeah, the only way in this tiny case is to have a fan as close to the intake as possible and hope positive pressure takes care of the heat.

 

Edit: Actually an idea popped up, I have my case standing upright with the GPU on top... The chambers are separate and the GPU spews its heat out the case already... So I wonder if the temps would become better if I flip it over so the CPU area is on top so heat can exit due to rising up out of the case, going to test this as well.

 

2 minutes ago, WoodenMarker said:

There's no recommended setting. Try manually setting it to 1.3v and keep lowering it from there in small increments while testing for stability. 

Alright, will try that then. Would offset be recommended to prevent it from sticking to one voltage?

[PCGuy_5960]

25 minutes ago, Mina Darsh said:

What surprised me is that Salazar Studios made a mini-ITX build in a Fractal Design Node 202, with an i5 7600K at 4.5 GHz and a Cryorig C7 on top. His temps were 75°C at most... So I'm wondering if the C7 is better or is the Kaby Lake i5 just less hot despite being 91 Watt TDP as well? Or maybe both?

Kabylake is pretty dang efficient, and at 1.3V, it isn't drawing much power....

[WoodenMarker]

8 minutes ago, Mina Darsh said:

Alright, will try that then. Would offset be recommended to prevent it from sticking to one voltage?

It's easier to use a manual voltage first to see how well good your chip is.

3 minutes ago, PCGuy_5960 said:

Kabylake is pretty dang efficient, and at 1.3V, it isn't drawing much power....

The temps are due to the OP using an i7 and 1.328v > 1.3v

[PCGuy_5960]

2 minutes ago, WoodenMarker said:

The temps are due to the OP using an i7 and 1.328v > 1.3v

I know, I just pointed out the reason why Salazar was getting these results with the C7....

[WoodenMarker]

9 minutes ago, PCGuy_5960 said:

I know, I just pointed out the reason why Salazar was getting these results with the C7....

It isn't because of Kabylake as opposed to Skylake.

[PCGuy_5960]

1 minute ago, WoodenMarker said:

It isn't because of Kabylake as opposed to Skylake.

Yes, but Kabylake can achieve higher clocks than Skylake with a specific voltage... In that sense...

[WoodenMarker]

1 minute ago, PCGuy_5960 said:

Yes, but Kabylake can achieve higher clocks than Skylake with a specific voltage... In that sense...

The temps would still be the same at the same voltage.

[done12many2]

Kaby Lake chips will almost always draw less power than Skylake chips under the same conditions.  

 

As an example, a 6700k at 4.7 GHz runs more current through the chip than a 7700k at 5.0 GHz.  Kaby Lake is more efficient with regards to power used under the same conditions.

 

This has translated into the additional overclocking headroom that we are seeing from Kaby Lake.

 

 

4 minutes ago, WoodenMarker said:

The temps would still be the same at the same voltage.

 

That is not completely true.

 

 

[Mina Darsh]

Well, found another setting that seems interesting, the VCore is set to 1.280v but somehow the system goes over it, but in a video, I hear that 'CPU Load-line Calibration might do this, so going to drop it in level to see if the end voltage drops, though, need to find out what this VDroop is...

 

Edit: Okay, not messing further with that and leaving it at Auto, lowering it too much makes Windows almost not responsive. (Can still enter Task Manager to reboot, but that's about it.)

 

[WoodenMarker]

9 minutes ago, done12many2 said:

Kaby Lake chips will almost always draw less power than Skylake chips under the same conditions.  

It's mentioned that Skylake pulls more current at 4.7ghz.

By the same conditions, are you referring to the same voltages being used ?

[done12many2]

17 minutes ago, WoodenMarker said:

It's mentioned that Skylake pulls more current at 4.7ghz.

By the same conditions, are you referring to the same voltages being used ?

 

Voltage isn't relative to heat output from two different chips regardless of if they are identical or not (6700k compared to another 6700k for example).  Tighter, low voltage chips can and often do output more heat at lower voltages than some looser chips at higher voltages.  At the end, the output (work produced) / actual power running through the chip is the governing factor in heat.  

 

Example.  One 6700k can OC to 4.7 @ 1.4v and another can OC to 4.7 @ 1.44v.  They can an often will produce the same thermal load.  A low voltage chip doesn't mean a cooler chip, nor does it directly translate into a "silicon lottery" winner as repetitive threads withing LTT would indicate.  

 

Voltage is simply the carrier if you will.  The amount of current passing through the chip is what translates into heat.  

 

How voltage impacts one chip is not comparable between two chips.  I wouldn't compare voltages between two identical SKL chips let alone a SKL and KBL chip.  

[WoodenMarker]

8 minutes ago, done12many2 said:

How voltage impacts one chip is not comparable between two chips.  I wouldn't compared voltages between two identical SKL chips let alone a SKL and KBL chip.  

 

So.. does Kabylake/Skylake in itself make any significant impact in temps?

[done12many2]

24 minutes ago, WoodenMarker said:

So.. does Kabylake/Skylake in itself make any significant impact in temps?

 

At the same power output, they are both going to roughly produce the same temp.  Very close cousins. 

 

The difference is in the voltage required to output the same power between both.   It's going to be much higher on a SKL chip when compared to a KBL chip.  For the same work load, a KBL chip at 1.25v might output 130w for the package, where an equivalent SKL chip might do the same 130w at 1.33v.  Both should return roughly the same thermal footprint.  

 

The difference in work output from that same thermal load is different though.  130w on a KBL chip will take you quite a bit further than 130w on a SKL chip.  That's to say that one may be able to run 4.7 GHz stable at 130w on the package while the other can do 5.0 GHz with the same power and stability.  Thermal load is the same, but the amount of work produce is completely different.  

 

You've probably seen a lot of thread popping up about KBL chips running hot.  This would explain it.  When those folks approach voltages in the same way as they would have for SKL, they are going to be upset with temps simply because KBL chips are an example of those low voltage chips I was referring to in my first post.

[Mina Darsh]

I think I found a solution! I'm benchmarking now and it stays at the specified voltage and seems to run a bit cooler.

 

All I did was update the BIOS to 3016 (I don't think this did it, might test later) and enable EPU, which seems to focus on running as efficiently as possible.

 

Voltage doesn't seem to go over 1.280v now and the system seems stable. (So far.)

[done12many2]

18 minutes ago, Mina Darsh said:

I think I found a solution! I'm benchmarking now and it stays at the specified voltage and seems to run a bit cooler.

 

All I did was update the BIOS to 3016 (I don't think this did it, might test later) and enable EPU, which seems to focus on running as efficiently as possible.

 

Voltage doesn't seem to go over 1.280v now and the system seems stable. (So far.)

 

Great to hear.

 

That chip will scale up differently when running EPU so I'd suspect that you just took a slight hit in performance, but if that's got you stable/cool, then stable/cool always wins.

 

Congrats!

[Mina Darsh]

4 hours ago, done12many2 said:

 

Great to hear.

 

That chip will scale up differently when running EPU so I'd suspect that you just took a slight hit in performance, but if that's got you stable/cool, then stable/cool always wins.

 

Congrats!

Actually did some further reading and was shocked that a lot of 6700K's could easily run a 4.5 Ghz overclock at 1.264v... So I tried it myself, though 4.4 Ghz...

 

After running a few rounds of AIDA64 and IntelBurnTest (20 times, High setting) it remained stable and non of the cores went past 86°C.

 

Seems the board's auto feature is indeed way too greedy on the power.