Cooling Issues i7 6700K

s1xtyn1n3 · 1 Jul 2017 at 15:54

I've recently noticed that my CPU was hitting pretty extreme temperatures under full load.

Idle: 28.25°C (it's pretty warm today)
100% Load: 98.75°C

I stressed the CPU using Prime95 and monitoring temps using CAM.

My PC is a 6700K (running at stock speeds) being cooled with a Corsair H80i v2 with a single fan in pull configuration. It's in a BitFenix Prodigy chassis with plenty of fans circulating air inside the case. I have both the pump and the fan curves set to performance but it seems to make little difference. I have cleaned the whole PC paying particular attention to the rad and that may have made 1 or 2 degs of difference, but really, not much.

Short of putting another fan on the rad (which i really don't want to do) and reapplying the thermal paste, I'm not sure what else to do.

Thoughts?

Doogles · 1 Jul 2017 at 17:17

Try reseating the heatsink with fresh TIM, another fan isn't going to drop temps much and it's obvious something is wrong cos it should be much cooler than that. Can you hear the pump working?

freddie64 · 1 Jul 2017 at 18:35

You say you have single fan in pull config on the rad, could you not try it in push config?
Would have thought it is better to force air through the radiator rather than draw it through.

s1xtyn1n3 · 1 Jul 2017 at 21:49

So I cleaned up the CPU and the cold plate, reapplied the TIM (Coolermaster E1 is what I have to hand) and still hitting around 100°C under full load. I checked the pump's connection to the CPU header on my Asus Z170i Pro Gaming while I was in there and that looked fine. I assume the pump is working fine as Corsair Link sees it and is reporting the pump speed of around 2910-2940 rpm. I am pretty sure I can hear the pump too, but its a small case so its difficult to isolate the sound from any one component.

Unfortunately I can't easily run the AIO in push as I would have to remove a case fan first. I may yet try this, but i think something more fundamental is wrong here, I'm just not sure what. Idle temps seem fine, but load is way higher than it should be. Is it likely to be the cooler that's got a problem or the CPU?

Doug2507 · 2 Jul 2017 at 08:27

Latest prime or earlier? Go run 26.6 or r15 etc and see what temps are.

Get a wee tube of kryonaught also.

s1xtyn1n3 · 2 Jul 2017 at 13:07

So with v26.6 (x64) I see temps around 88 running Small FFTs and 80 running Blend (CPU still hitting 100% each time).

Obviously these are significantly lower than when I was running v28.1 (x64). They are higher than I would expect on water at stock + XMP, but still much closer to acceptable. Any reason why the different versions cause such a massive change in load temps?

I will definitely check out kryonaught too - Was contemplating a chassis change from my Prodigy to the Enthoo Evolv M-ATX case and swapping out the H80i v2 for a Kraken x62 so will definitely have some better TIM for that.

Doug2507 · 2 Jul 2017 at 13:22

Latest prime uses latest AVX. No need for regular use in most cases.

If you've not touched bios then you've got auto vcore (not great) and auto LLC (not great) working against you. Tweak the bios a bit and temps will drop for given clock speed.

Use a multimeter or check CPU-Z vcore under load.

If you don't want to raise clock speed, just set vcore manually somewhere around 1.225v and should be solid with better temps. If you want to OC set vcore at 1.3v, load to check temp then raise core multi till you loose stability. That'll give you a nice boost in clock speed and still keep temps very acceptable.

What board? (If asus set LLC 5/6). Other boards just go middle of the road, put cpu under 100% load and check vcore in CPU-Z etc. Adjust LLC (Load line calibration) either way to have load vcore as close as possible to bios vcore. (llc5 on asus will slightly undervolt on load, ll6 will slightly overvolt).

s1xtyn1n3 · 2 Jul 2017 at 14:04

CPU-Z is saying 1.392 V (fluctuating up to 1.408 V) under full load.

I have just tried tweaking some of the settings in the bios for my ASUS Z170 and the results weren't great:

Starting values:
BCLK: 103
CPU Core Ratio: 45
LLC: Auto
CPU Core/Cache Voltage: Auto

1st change:
LLC: level 5
CPU Core/Cache Voltage: 1.300
Result: boots ok, bsod immediately when stressed

2nd change:
LLC: level 6
CPU Core/Cache Voltage: 1.300
Result: boots ok, bsod immediately when stressed

3rd change:
LLC: level 5
CPU Core/Cache Voltage: Auto
Result: boots ok, stable, load: 96.75°C/1.472V

s1xtyn1n3 · 2 Jul 2017 at 14:13

I have also tried dropping the base clock down to 100 (so a manual oc essentially of 4.5GHz compared to the 4.6GHz oc that XMP set) which takes maybe 1 or 2°C of difference under load

HoneyBadger · 2 Jul 2017 at 15:53

What are the amibient temps of the room the PC is in?

s1xtyn1n3 · 2 Jul 2017 at 18:01

It's about 23/24°C here today

freddie64 · 3 Jul 2017 at 07:05

s1xtyn1n3 said:
CPU-Z is saying 1.392 V (fluctuating up to 1.408 V) under full load.

Was that with no overclock?

You may just be unfortunate and lost the silicon lottery and have a bad CPU.

Are you on latest bios version?

s1xtyn1n3 · 3 Jul 2017 at 09:43

Those voltages were with the XMP pushing it up to 4.6GHz

My bios isn't up to date, can't remember how up to date it is. Will update that tonight and see if that makes a difference.

doyll · 3 Jul 2017 at 10:31

How old is your H80? I ask because CLCs are known to sometimes have pump problems after a few years, and maybe your H80 is not cooling like it should. Do you have another cooler you could put on to see if it solve the cooling problem?

Doug2507 · 3 Jul 2017 at 15:03

s1xtyn1n3 said:
CPU-Z is saying 1.392 V (fluctuating up to 1.408 V) under full load.

I have just tried tweaking some of the settings in the bios for my ASUS Z170 and the results weren't great:

Starting values:
BCLK: 103
CPU Core Ratio: 45
LLC: Auto
CPU Core/Cache Voltage: Auto

1st change:
LLC: level 5
CPU Core/Cache Voltage: 1.300
Result: boots ok, bsod immediately when stressed

2nd change:
LLC: level 6
CPU Core/Cache Voltage: 1.300
Result: boots ok, bsod immediately when stressed

3rd change:
LLC: level 5
CPU Core/Cache Voltage: Auto
Result: boots ok, stable, load: 96.75°C/1.472V

Looks like your chip just isn't a great clocker. Set LLC5/1.3v/stock multi, load in win (r15 etc) and check temp. If all good at 1.3v then start raising core multi. You're kind of going about this the wrong way.

If temp is still lousy at 1.3v then i'd look into a better cooling solution. OCUK branded AIO are excellent btw.

s1xtyn1n3 · 3 Jul 2017 at 20:45

doyll said:
How old is your H80? I ask because CLCs are known to sometimes have pump problems after a few years, and maybe your H80 is not cooling like it should. Do you have another cooler you could put on to see if it solve the cooling problem?

It's probably about 12 months old so I wouldn't expect any issues from it really.

Doug2507 said:
Looks like your chip just isn't a great clocker. Set LLC5/1.3v/stock multi, load in win (r15 etc) and check temp. If all good at 1.3v then start raising core multi. You're kind of going about this the wrong way.

If temp is still lousy at 1.3v then i'd look into a better cooling solution. OCUK branded AIO are excellent btw.

More quality information that I will absolutely try as soon as I get the chance - I'm way too tired tonight.

I'm also going to try flipping the fan on the rad round.

It will be pushing cold air from outside the case through the radiator into the chassis, but as you can see, there's a lot of other fans that will disperse that pretty quick. It will be a temporary solution anyway just to see if it makes a difference.

Doug2507 · 3 Jul 2017 at 22:06

Looks like a 240mm asetek would slot in nicely.

s1xtyn1n3 · 4 Jul 2017 at 17:53

Update: I updated the bios and dropped the overclock down to 4.2GHz - now seeing max temps of about 70°C. So yes, it appears its a bit of a lemon when it comes to OC. Am still going to try reconfiging the rad/fan setup.

And yep! A 240mm would fit like a glove up there! I'm half contemplating a change of case though and would probably go to a 280mm AIO at the same time.

Doug2507 · 4 Jul 2017 at 21:14

And that's with 1.3v vcore/llc5? If so then you've still got a little headroom on vcore with regards to temp. Might be good around 1.325v give or take. Once checked start upping core multi till you loose stability then roll back 1 multi. Repeat with cache. Job done.

FYI, 120mm fans tend to work a little better than 140mm.

jabbz · 4 Jul 2017 at 21:37

How fast do the temps rise? If it is in a few seconds like 95c after 30seconds then I dont think extra fans will help at all.
The reason for this is aio's take a while to reach max temps unlike air coolers. They also take longer to cool down after, so this might give you
a clue as to what is going on. I would say it was a slow flow in the aio if it was maxing out quite soon after the stress test has started.