i7-8700k delid results

[partymarty]

Hello guys and girls,

 

I've recently purchased an Intel Core i7-8700k, after installing it in my system I started by doing some thermal tests. For cooling I'm using an EKWB 360 SE + EKWB 240 SE (with 5x Corsair SP120 Quiet PWM). My GPU and CPU are in the same loop. 

Mobo = Asus ROG Maximus X Hero

Case  = Phanteks Enthoo Evolv ATX TG (this case is really, really, really bad for airflow)

 

Pre-delid I got these results, after 30 minutes, which I've tested with only FPU enabled in the AIDA64 stresstest (running on stock bios settings, only changed fan curve to PWM Standard)

Fan RPM = 1450

Vcore = 1.26-1.28

CPU = 84 degrees C

Maximum temperatures per core = 85-85-82-83-83-82-82

Average temperatures per core = 71-80-76-71-78-79-76

 

With FPU + GPU enabled (to dump as much heat into the loop as possible) I got these results:

Fan RPM = 1450

Vcore = 1.26-1.28

CPU = 93 degrees C (Stock bios settings... With a 360se+240se custom waterloop???)

Maximum temperatures per core = 95-95-90-93-91-91-90

Average temperatures per core = 77-84-82-88-85-86-83

---

After seeing those results I just had to take the gamble and delid this thing. Because I think it's ridiculous that the CPU get's this hot with that kind of cooling on top of it. So I delidded the 8700k with the Der 8auer Delid-Die-Mate v2, cleaned the thing, gave it some liquid metal love (thermal grizzly conductonaut) and resealed the CPU with some heat resistant glue.

---

After delidding this puppy I did the same tests as before. The results:

Stock settings. Only Fan curve to PWM Standard. 30 minutes. First only FPU, then FPU+GPU.

Only FPU:

Fan RPM = 940

Vcore = 1.26-1.28

CPU = 59-60 degrees C (it dropped 24-25 degrees!)

Maximum temperatures per core = 62-62-57-60-61-60-60 (22-23 degrees lower than before)

Average temperatures per core = 46-58-55-55-56-54-55 (again more than 20 degrees lower)

 

FPU + GPU:

Fan RPM = 1036

Vcore =  1.26-1.28

CPU = 72 degrees C (dropped 21 degrees)

Maximum temperatures per core = 76-76-71-74-73-72-72 (18-20 degrees lower)

Average temperatures per core = 59-70-67-68-68-67-67 (again around 20 degrees lower)

--

 

Conclusion:

Dafaq Intel. I knew it would help the chip but I never expected it to do this much work. 

At 5GHz it throtteled instantly, now I can run 1.35 volts without the temperatures going through the roof instantly. I think it's ridiculous that Intel uses that kind of crap on an unlocked 350 euro's chip.

[Enderman]

It's a consumer CPU.

 

99.9% of consumers do not overclock.

 

It may seem like a lot do, because you're on an enthusiast tech forum, but very few actually oc.

[partymarty]

Just now, Enderman said:

It's a consumer CPU.

 

99.9% of consumers do not overclock.

 

It may seem like a lot do, because you're on an enthusiast tech forum, but very few actually oc.

I know. But still... If I see those results; a standard CM Hyper 212EVO is going to have a hard time with this CPU. Even at stock speed.

[Enderman]

3 minutes ago, partymarty said:

I know. But still... If I see those results; a standard CM Hyper 212EVO is going to have a hard time with this CPU. Even at stock speed.

Well, it's definitely not ideal to pair a $30 cooler with a $300 CPU, but in those cases usually it's an 8700 and not an 8700k in the PC.

People who buy K CPUs know they need a good cooler if they want to OC.

At stock, it's just loud, maybe a bit of throttling, that's all. Not gonna kill the CPU.

[iamdarkyoshi]

17 minutes ago, Enderman said:

It's a consumer CPU.

 

99.9% of consumers do not overclock.

 

It may seem like a lot do, because you're on an enthusiast tech forum, but very few actually oc.

Those who spend the extra money and buy a K series chip generally do OC

 

13 minutes ago, Enderman said:

Well, it's definitely not ideal to pair a $30 cooler with a $300 CPU, but in those cases usually it's an 8700 and not an 8700k in the PC.

I'm very limited for cooler space in my PC. This is the heatsink I'm using now, its only 6cm tall:

 

I'm actually running into VRM limitations before hitting thermal limitations on my 6700K after a delid with this tiny little heatsink. 

 

Before a delid, anything above stock was extremely hot and the fans were loud. Now temps are in the 60-70 range with the fans at a much lower speed.

 

I also managed to hit 4500mhz without excessive temps on a stock intel cooler after a delid, albeit a bit loud.

[Enderman]

1 minute ago, iamdarkyoshi said:

Those who spend the extra money and buy a K series chip generally do OC

 

You would be surprised how many millions of prebuilt computers with K CPUs are sold to consumers who have no idea what overclocking is.

[iamdarkyoshi]

1 minute ago, Enderman said:

You would be surprised how many hundreds of thousands of prebuilt computers with K CPUs are sold to consumers who have no idea what overclocking is.

On prebuilts, I'll give you that. Though most of the prebuilds I've seen either have FX CPUs with the stock heatsink, or intel non K series chips...

 

But on a chip who's only real extra selling point is an unlocked multiplier, whats the point if you can't get any reasonable OC off it?

[Enderman]

4 minutes ago, iamdarkyoshi said:

On prebuilts, I'll give you that. Though most of the prebuilds I've seen either have FX CPUs with the stock heatsink, or intel non K series chips...

 

But on a chip who's only real extra selling point is an unlocked multiplier, whats the point if you can't get any reasonable OC off it?

Well intel doesn't make any claims about what frequencies you can overclock to, only that it's unlocked, so there's pretty much no reason for them to make it reach a bajillion GHz, they just don't care.

[partymarty]

I'm not delidding for overclocking. But for silence... When gaming my PC made so much noise.

Now after the delid + undervolting (-0.70v) my 8700k won't reach 50 degrees celcius. So I can let my fans run as low as possible.

 

My pc is is located in the living room and my girlfriend got really annoyed by the noise that thing produced when I threw some heavier workloads at the system.

 

Problem is solved now  

[iamdarkyoshi]

3 minutes ago, Enderman said:

Well intel doesn't make any claims about what frequencies you can overclock to, only that it's unlocked, so there's pretty much no reason for them to make it reach a bajillion GHz, they just don't care.

 

My 6700K was throttling at stock clocks before a delid. I reseated the heatsink and got exactly the same result.

 

Temps dropped drastically after a delid, even with an extra 500mhz:

 

 

Yes, it could be argued that a much better cooler would yield a better OC, but even delidded, this heatsink managed to do better than the nexus low 7000 I originally tested with at stock clocks:

 

 

[For Science!]

@partymarty

Sounds consistent with my 8700K (copy paste from another one of my replies below) but TL;DR: 16 degrees improvement by delidding.

 

CPU: 8700K @ All-core 4.8 GHz (due to Asus MCE enable) at "Auto Voltages"

GPU: 1080Ti Aero (MSI, blower-style)

Case: NCase M1

Radiators when custom: 240 + 120 mm HWLabs GTS

 

I was using RealBench as a stress test since it puts stress on the entire system (GPU. CPU. Memory)

 

When using "Air cooling" Which was with a 240 mm AIO on the CPU and the stock GPU cooler (so not really Air, but lets just assume it is "BEST case scenario"). Basically the CPU temperatures are likely to not get this good with any air cooler that fits inside the Ncase M1. The AIO fans were on exhaust to prevent re-heating the case internals.

 

CPU Temperature: 75 degrees

GPU Temperature: 67 degrees

 

Now under these conditions, the system was very loud. The high rpm of the GPU cooler was very noticeable and the H100i v2 was also noticeable.

 

Since we know that intel CPUs benefit a huge amount from delidding, the stock TIM between the IHS and CPU die was replaced with liquid metal (TG Conductonaut). In my case, I have also re-lidded the CPU (i.e. re-introduce the silicone sealant) many people do not like this and thinks it kills the point of delidding, but see the results below. The same test was conducted using the same configuration as above.

 

CPU Temperature: 59 degrees (-16 degrees)

GPU Temperature: 67 degrees

Total change vs previous: -16 degrees

 

Since the system is kicking off the same amount of heat, the system is still just as loud. But once again, we see the benefit of delidding the CPU and replacing the TIM with LM. 16 degrees improvement is fairly representative of what people see with a LM swap, and shows at least the way I reseal my CPUs with silicone does not adversely affect the gains. Or lets put it this way, perhaps without the silicone I may have gained more, but I am happy to take 16 degrees improvement without the worry of my IHS falling off when my GF opens up the CPU latch for some reason. 

 

So here comes the interesting part. We take out the AIO, slap on the custom loop parts. In my case that means I have a total of 360 mm radiator space which is 30 mm thick, meaning an addition of 120 mm total radiator space when comparing vs the H100i v2. We run the same test:

 

CPU Temperature: 75 degrees (+16 degrees)

GPU Temperature: 54 degrees (-13 degrees)

Total change vs previous: +3 degrees