Delid or second fan.

Associate
Joined
19 Sep 2019
Posts
78
Hello everyone

I have hit my temp wall overclocking my 8600k. I can run 5Ghz stable @1.34v but only with an AVX offset of 1.

Prime95 AVX & Intel Burn Test very high both push my hottest core to 89C. If I remove the 1 AVX outset I fail the stress tests. Any more volts I push 90C. So really a no go!

I have 2 options, install a 120mm fan over my high profile ram onto my Noctua NH-D15S or delid to apply some liquid metal.

So my question will a further 120mm Noctua fan make any real difference or is delidding my only option left if I want to be able to run stable 5Ghz with no AVX offset?

Thank you
 
Soldato
Joined
1 Jul 2011
Posts
8,632
Adding another fan to NH-D15S will only gain you 2-3c lower temps.

Another possibility is case fans are supplying air to cooler that is 5-15c above room air temp .. every degree lower the air temp into cooler is translates into same number of degrees cooler CPU will be. Link below is to a simple guide of how airflow works, how to optimize case airflow and how to monitor airflow temp into coolers:
https://forums.overclockers.co.uk/t...-i-put-my-temp-sensor.18564223/#post-26159770

I suggest you monitor air temp into cooler before adding another cooler fan.
 
Associate
OP
Joined
19 Sep 2019
Posts
78
I have a NZXT Full Phantom that I custom modded removing the old HDD tray bays for better airflow and open space. On top of this case are 2 200mm horizontal mounted Phanteks. One sits above and just in front of the CPU cooler pushing 110.1CFM/1.04mm static pressure threw no dust filter only a metal mesh directly into the Noctua cpu block. The rear top mounted 200mm is exhaust above and behind the cpu block. Directly behind the cpu block is a 140mm Fractal Dynamic X2 exhausting as well. At the bottom front I have a 120mm Fractal X2 pushing air directly into the GPU behind mesh no dust filter as well. I control these using a manual control set to max rpm when overclocking.

Working with what I have I have tried my best to create good airflow in my case. Its this reason alone I have not upgraded my case with the current trend of closed case fronts etc.. Coolmaster Mesh was an option but personally dislike the look of it.

Am I doing something wrong here or are we looking at the limitations of the cpu itself? Might delidding be the way forward for me?

P.s I can place my hole hand in front of my cpu cooler, I have done this many times to check the airflow and temperature. Cold air goes into the front of the cpu block and all hot is exhausted at the rear.
 
Last edited:
Soldato
Joined
5 Nov 2011
Posts
5,356
Location
Derbyshire
De-lid.
An extra fan will give maybe a couple of degrees less. A delid done properly will pull worst case 10 degrees, best case 20 off.

I have all the kit to delid in my cave at the moment and am waiting to see how much the 9900KS comes in at. If it's reasonable I'll sell my 8700K, if it's not then I'm delidding.
 
Associate
OP
Joined
19 Sep 2019
Posts
78
De-lid.
An extra fan will give maybe a couple of degrees less. A delid done properly will pull worst case 10 degrees, best case 20 off.

So it looks like delidding! Must admit a tad nervous doing it myself. Do you know of any reputable company that does it at a reasonable price?

Thank you Doyll & Steveocee for your help!
 
Last edited:
Soldato
Joined
1 Jul 2011
Posts
8,632
I have a NZXT Full Phantom that I custom modded removing the old HDD tray bays for better airflow and open space. On top of this case are 2 200mm horizontal mounted Phanteks. One sits above and just in front of the CPU cooler pushing 110.1CFM/1.04mm static pressure threw no dust filter only a metal mesh directly into the Noctua cpu block. The rear top mounted 200mm is exhaust above and behind the cpu block. Directly behind the cpu block is a 140mm Fractal Dynamic X2 exhausting as well. At the bottom front I have a 120mm Fractal X2 pushing air directly into the GPU behind mesh no dust filter as well. I control these using a manual control set to max rpm when overclocking.

Working with what I have I have tried my best to create good airflow in my case. Its this reason alone I have not upgraded my case with the current trend of closed case fronts etc.. Coolmaster Mesh was an option but personally dislike the look of it.

Am I doing something wrong here or are we looking at the limitations of the cpu itself? Might delidding be the way forward for me?

P.s I can place my hole hand in front of my cpu cooler, I have done this many times to check the airflow and temperature. Cold air goes into the front of the cpu block and all hot is exhausted at the rear.
I had Enthoo Luxe with it's 200mm fan and found a single 140mm fan (with all openings in fan mounting panel not covered by fan blocked) move more air into mobo compartment than the 200mm fan did. Based on that and what others found I think they are almost worthless to me, but in your case (no pun) they might work. Having one pulling out and other pushing in means the air coming out is being push/pulled by fan pushing air into case.

05064c1b_Intake-Exhauxtairlfowloop.jpeg


Case can only flow as much air as the lower of the two; amount of intake air or amount of exhasut air. Airflow is simple displacement .. what flow in must flow out / what flows out must flow in. It's explained better in link in my first reply to you.

Personally I don't like Phantom. It's bastardized case airflow into front bottom to top rear while the motherboard is oriented for just front to back airflow to keep GPU heat from moving up into CPU intake airflow.

Have you read the info in link I gave you? In it I tried to explain the principles of airflow hand how to apply them to a normal tower case.

Our hands in front of cooler don't give you much if any idea how much airflow ther is or what temp it is .. and if case is open the air temp will be much differnt then if case is closed.

So it looks like delidding! Must admit a tad nervous doing it myself. Do you know of any reputable company that does it at a reasonable price?

Thank you Doyll & Steveocee for your help!
Delidding mght help some, but to really get temps down we need your system airflow into cooler close to room air temp (2-3c warmer is okay). If air temp into cooler when gaming is 12-13c above room temp then CPU will be about 10c hotter than if air into cooler is only 2-3c warmer than room
 
Soldato
Joined
5 Nov 2011
Posts
5,356
Location
Derbyshire
Delidding mght help some, but to really get temps down we need your system airflow into cooler close to room air temp (2-3c warmer is okay). If air temp into cooler when gaming is 12-13c above room temp then CPU will be about 10c hotter than if air into cooler is only 2-3c warmer than room

It is well documented the really poor TIM used between the die and the IHS on CPU's from Ivybridge upwards to Coffeelake. A delid and use of liquid metal (conductonaught for product recommendation) is an easy way of dramatically reducing the temperature and relatively easy given the tools available.

You can have all the airflow in the world but it's useless if you can't get the heat off the die in the first place which is where the mentioned CPU's fall short.
 
Soldato
Joined
1 Jul 2011
Posts
8,632
It is well documented the really poor TIM used between the die and the IHS on CPU's from Ivybridge upwards to Coffeelake. A delid and use of liquid metal (conductonaught for product recommendation) is an easy way of dramatically reducing the temperature and relatively easy given the tools available.

You can have all the airflow in the world but it's useless if you can't get the heat off the die in the first place which is where the mentioned CPU's fall short.
All true, but everything you say about dramatically lowing temps can be said about improving case airflow.

Both might improve cooling one or the other might, or neither might.

But it's easy monitor air temp into cooler with a simple digital indoor / outdoor digital thermometer or remote sensor thermometer like terarium / aquarium thermometer costing less than 5 quid than it is to delid CPU or change fans.

If when gaming air temp into cooler is 12-13c higher than room case airflow, lowering air temp to 2-3c above room lowers CPU temp 10c. Then if it's still too hot a delid would be my next move.
 
Associate
OP
Joined
19 Sep 2019
Posts
78
Have you read the info in link I gave you? In it I tried to explain the principles of airflow hand how to apply them to a normal tower case.

Our hands in front of cooler don't give you much if any idea how much airflow ther is or what temp it is .. and if case is open the air temp will be much differnt then if case is closed.

Yes I did read it. Found it very interesting. Bookmarked for future reference. All I can add is that I have had this case for many years and tried a verity of different fan mountings. From 200mm side panel fan to the 2 120mm fans at bottom( Useless really as exhaust straight to other side of case) to no top exhaust etc.. I was worried about the top rear exhaust push/pulling heat back in and ran benchmarks using Core Temp, Hwinfo64 with it disabled to see. No real difference.

Remember also that I have a fan controller allowing me to adjust my fan speeds to tweak pressures. Also it takes 2 seconds to the remove the front and top panel, another 3 seconds to remove the side panels. Running prime with AVX instructions but with all of the panels removed, effectively very close to an open air test bench, creates a difference in temperatures of 1-2 degrees max. Really, I have tried every possible mounting, configuration, open aired testing, etc.. and can pretty much say imho that airflow is not the issue for me. My tower is huge with plenty of space and running benches with all panels of with no real change in temps is a testament that its all working as intended. Non overclocked idles temps are in the 20C, overclocked at 5Ghz in the 30C.

I think Steveocee has hit the nail on the head and its coming down to the poor TIM. Other activities like gaming, Frostpunk for instance, running on ultra using nvidia 4k synthetic (DSR) resolution never brings my hottest core above 61C clocked at 5Ghz. Heat is only a problem when benchmarking the toughest stress tests out there. A lot of people don't even bother with Prime95 AVX anymore because of the unrealistic workload and heat production.

If as steveocee says, 10 degrees of worse case scenario after delidding, will mean I am running prime95 AVX at 5Ghz in the 70C range! At best 60's!

That will do very nicely indeed!
 
Last edited:
Soldato
Joined
5 Nov 2011
Posts
5,356
Location
Derbyshire
All true, but everything you say about dramatically lowing temps can be said about improving case airflow.

Both might improve cooling one or the other might, or neither might.

But it's easy monitor air temp into cooler with a simple digital indoor / outdoor digital thermometer or remote sensor thermometer like terarium / aquarium thermometer costing less than 5 quid than it is to delid CPU or change fans.

If when gaming air temp into cooler is 12-13c higher than room case airflow, lowering air temp to 2-3c above room lowers CPU temp 10c. Then if it's still too hot a delid would be my next move.

Again though, having fantastic airflow is no use if that air flow isn't getting to the heated component it needs to cool down. Stick the PC in a fridge and bring ambient down, if the heat isn't dissipated into the fins on the cooler then cold air won't help, the cores will run hot as the die cannot transfer it's heat out to the heatsink efficiently enough.

This is a null argument as it has been proven time and again that the thermal interface paste Intel used from die to IHS is **** and replacing it with something decent reduces temps massively. Even those who don't overclock yield solid reductions in temperature due to poor build in the first instance.

***You're lucky in that your 980 was a soldered heatsink so you have no such issues as was the case up until Sandybridge.
 
Soldato
Joined
1 Jul 2011
Posts
8,632
Yes I did read it. Found it very interesting. Bookmarked for future reference.

Really, I have tried every possible mounting, configuration, open aired testing, etc.. and can pretty much say imho that airflow is not the issue for me. My tower is huge with plenty of space and running benches with all panels of with no real change in temps is a testament that its all working as intended. Non overclocked idles temps are in the 20C, overclocked at 5Ghz in the 30C.

I think Steveocee has hit the nail on the head and its coming down to the poor TIM. Other activities like gaming, Frostpunk for instance, running on ultra using nvidia 4k synthetic (DSR) resolution never brings my hottest core above 61C clocked at 5Ghz. Heat is only a problem when benchmarking the toughest stress tests out there. A lot of people don't even bother with Prime95 AVX anymore because of the unrealistic workload and heat production.

If as steveocee says, 10 degrees of worse case scenario after delidding, will mean I am running prime95 AVX at 5Ghz in the 70C range! At best 60's!

That will do very nicely indeed!
Delidding will almost assuredly lower temps. How much remains to be seen .. could be a couple degree, could be more. I suspect 10c is too optimistic to be real, but I could be wrong.

What I'm getting at is without actually monitoring airflow temp into cooler it's almost impossible to know what that air temp is. As for running wiht case open, about half of my builds run cooler with panels on then off. It is not a given an open bench test station will be supplying room temp air to CPU or GPU coolers.

At the end of the day a basic remote sensor digital thermometers (indoor/outdoor or tarrarium thermometers) are less 2-5 quid on ebay. Some tape, stiff piece of wire and clothspin added and you have a thermometer that sensor can be positioned an inch or so in front of cooler intake fan and we cab see actual temp of air going into cooler while gaming / stress testing / idling. Then you will know without a doubt what the air temp is.
 
Associate
OP
Joined
19 Sep 2019
Posts
78
I can see what your getting at Doyll. I think you approach your builds very scientifically and coach & tinker everything down to the tinniest detail.

I'll grab a remote sensor and try to see my temps going into the cpu cooler, what could it hurt? Something like the Exo Terra Digital Thermometer, it has a suction cup on for easy mounting for £9 on Amazon. If it does not help now does not mean it might not be helpful in the future especially when I finally upgrade my case to one I like, save hours of tinkering.

Even so, if there is an airflow issue naff all I can do about it other than replacing the case. That probably means upward of £100. However, delidding will cost me less than £40 with more than likely similar if not better results.

But ultimately, delidding is the way to go for me. I watch Gamersnexus, jayz2cents etc & know the difference delidding has but there is no video showing results of say adding a second cpu fan, hence my original question of fan vs delidding.

Thank you again Doyll and Steveocee, I think my question has very much been answered now!
 
Soldato
Joined
1 Jul 2011
Posts
8,632
The Exo Terra one is not as good as a cheap indoor outdoor digital from E-bay. Something like image below
LL

Then bend the wire so sensoris centered on fan about 1 inch in front of it. That suction cup one won't position sensor where it is best.

Keep in mind the science of airflow is named fluid dynampics because even though air is a gas it moves the same way liquids do. Easy way to think of it is our cases are ponds with intake vents being streams flowing into pond and exhaust vents being streams leaving pond. The water moves across the pond same way air moves though our cases. It following paths of least resistance and if smooth flow is disturbed it swirls around creating turbulence mixing wiht other water in the pond. This means the smoother the air flows / less turbulence there is the less warm air mixes with cool airflow and the closer to room temp the air will be going into coolers (CPU & GPU). Almost always slower airflow will supply more air to component than faster air because faster movement causes more mixing. So having case fans speed up and slow down in sync with cooler fans works best. Key is setting case fan speed to flow same or slightly more cool air to component than component is using an then moves on to exhaust vents (vents that most often do not need exhaust fans). Good intake fans move air through the case similar to push fan on air cooler. Intake / exhaust case fans are like push / pull fans on a cooler.
 
Associate
Joined
5 Mar 2017
Posts
2,216
Location
Cambridge
what temp do you get doing normal activities? why not reduce frequency and voltage
This. Quite often the trade-off for more 50 or 100Mhz isn't worth considering the "jump" in voltage needed to keep the overclock stable and the temperature under control.
Also during benchmarks, usually you'll see the worst scenario, which for 99% of the users/use of pc, simply won't happen for longer than few spikes.
 
Associate
OP
Joined
19 Sep 2019
Posts
78
what temp do you get doing normal activities? why not reduce frequency and voltage

With adaptive overclock 20C range Idle. Max is in the 50-60C range gaming. Just to keep things easy I have set an AVX offset of 2 for a 4.8Ghz when benchmarking. So now I have a fully stable adaptive 1.34v (1.165v with adaptive offset of 0.175 for 1.34v) 5Ghz overclock for non AVX & 4.8Ghz for AVX.

Temps hit 91C on hottest core if I add more voltage to lower AVX offset on Prime AVX or IntelBurnTest Very High.

Good new is that if I do delid pretty much guaranteed to run 5Ghz without any AVX offset.
 
Soldato
Joined
1 Jul 2011
Posts
8,632
With adaptive overclock 20C range Idle. Max is in the 50-60C range gaming. Just to keep things easy I have set an AVX offset of 2 for a 4.8Ghz when benchmarking. So now I have a fully stable adaptive 1.34v (1.165v with adaptive offset of 0.175 for 1.34v) 5Ghz overclock for non AVX & 4.8Ghz for AVX.

Temps hit 91C on hottest core if I add more voltage to lower AVX offset on Prime AVX or IntelBurnTest Very High.

Good new is that if I do delid pretty much guaranteed to run 5Ghz without any AVX offset.
Those temps tend to indicate heat transfer from chip is pretty good. How long does it take for temps to reach 90c once you push CPU to high load?
 
Associate
OP
Joined
19 Sep 2019
Posts
78
Those temps tend to indicate heat transfer from chip is pretty good. How long does it take for temps to reach 90c once you push CPU to high load?

Not long tbh. Within a few Prime cycles. Maybe as little as a few minutes its settles.

Actually dropped this overclock now as noticed that a lot of programs & games use some form of AVX as I was pretty much running at 4.8Ghz. Only non AVX Prime seemed to really utilise the 5Ghz. Currently testing at 4.8Ghz 1.295v. Well bellow 90C now and amazed at how much more voltage, lets not forget the heat, you need just for the extra 200 mhz.

Ill let you know what my final voltage & heat is as still testing and dropping.
 
Last edited:
Associate
OP
Joined
19 Sep 2019
Posts
78
4.8Ghz 1.29v. After running a few hours max temp recorded 87C.

When I first bought the cooler was getting similar heat results at higher speeds. Think I might strip the cooler down for a good clean. I have an air pressure tool but reckon its time for a wash to get all of that secreted dust out.
 
Back
Top Bottom