Question 8700K delid temps

[toyo]

Hello guys,

First things first: my main curiosity is - are the temps OK and within expectations? Or can it be improved?

I took the jump and delidded my 8700K. It's quite a mediocre CPU compared to what I've seen on the web for people winning the silicon lottery. But my preoccupation is not that much with the OC itself, but with the temps and if I applied the stuff correctly. There are quite a few delid tests on the web, but few are actually properly documented, i.e. you don't know for sure what sort of exact test people use, if they disabled AVX, what settings did they use inside prime95, what exact voltage and which voltage method etc.

I'll start with the settings and components, and then will link 2 screenshots. I tried to use a common test for p95/Coffee Lake OCs, 1344K Inplace FFTs with AVX disabled at 1.35V manual voltage inside the BIOS, so it's easy to replicate.

Specs: 8700K delidded, Conductonaut on die and back of IHS, as thin a layer as I could. Clear nail varnish on the 4 contacts near the die. Light resealing with Reinzosil sealant for high temps (can't really recommend this, it kind of forms a very thin layer on the substrate that's very hard to clean, no scraping works, and isopropilic alcohol 99% or acetone won't dissolve it.). Kryonaut between IHS and cooler coldplate, small amount, tried to spread it as much as I could.

Ambient temp: 24C.

Cooling: Meshify C case, 4x140mm Noctua PWM fans (2 intake, 2 exhaust), 1x120mm exhaust at the back, Noctua D15S cooler with additional 120mm Noctua PWM fan on the front.

In the tests the fans run at around 60-75% (approx 1500rpm for the 120mm, 1000 rpm for the 140mm).

Motherboard is a ****ty 4 phase VRM MSI Gaming Plus. I haven't seen it throttle yet with up to 165W package power on the CPU. It's basically the same VRM as on the MSI Z370-A Pro "reviewed" by der8auer in this video, where he could OC to 5.1GHz without a fan on the VRM. The VRM temps for my use case are 80C-ish, peaking at 92C. I personally don't feel comfortable drawing more than 150W through the VRMs.

Because of the motherboard VRM and the insane rise in power drawn with AVX2 and Prime95, I can't really run SmallFFTs with AVX2 as a test. My CPU is not a great sample and it seems to require around 1.33V-1.34V for 4800MHz if I enable AVX2. 5GHz would likely require 1.4V to get "fully" stable.

Now, the tests. 1.35V in BIOS. 50x multi, no AVX offset, 44x Uncore.

Test 1: prime95, last version, AVX disabled, 1344K Inplace FFTs, 15 mins (seem to be enough for peak temps, in fact they seem to go down over the test).

Max core 75C, average core/package temp 71C, power draw peak at 144W.

Spoiler

Test 2: Realbench, last version (apparently this is an AVX test. It will also use the GPU, so this will contribute some to higher temps). 15 mins stresstest, 8GB RAM used.

Max core temp 77C, average core/package 73C, peak power draw 134W.

Spoiler

If you want me to try and replicate your own conditions, hopefully it's not too low of a voltage, my 8700K is unstable at below 1.32V@5GHz, so I can't really test that. Also I can't test with P95 AVX2 at higher than 4.8GHz or so because of the ****ty VRMs and me needing quite high voltage to reach AVX2 stability.

Sorry for the long post. Just trying to see if I did apply the Conductonaut properly basically. Already redid the delid once, but not much has changed, if anything.

Bonus pics (with liquid metal seen affecting the IHS after just 2 days, and the very thin silicon pellicle that remains after the Reinzosil sealant and I can't remove). Cleaning the IHS is pretty much impossible too. Acetone, isopropilic alcohol couldn't undo the reaction. LM seems quite reactive even with the nickel plated IHS.

Spoiler

 

[zx128k]

Great job with the delid, looks perfect. Your 8700k is not so bad if you need 1.35V in BIOS. 50x multi, no AVX offset, 44x Uncore. Only 17% make a 5GHz AVX @ 1.425 volts in the silicon lottery statistics. The prime95 temps show a lot of head room temperature wise. VRM temps of 80C-ish, peaking at 92C. Yes they are high but I ran 98-104c vrm temps for 5 years+ with no issues with lots of air flow. They dropped to 98c in prime 95 small fft's. This is with a rampage 4 black edition and the 4930k.

You could get a monoblock that covers both cpu and vrm area. That should hopefully bring drop VRM temps, if you can find one for your board. Random example https://www.overclockers.co.uk/ek-w...ix-z390i-d-rgb-monoblock-plexi-wc-9sy-ek.html

8700K	4.80GHz	4.60GHz	1.375V	100%
8700K	4.90GHz	4.70GHz	1.387V	Top 99%
8700K	5.00GHz	4.80GHz	1.400V	Top 83%
8700K	5.10GHz	4.90GHz	1.412V	Top 49%
8700K	5.20GHz	5.00GHz	1.425V	Top 17%
8700K	5.30GHz	5.10GHz	1.437V	Top 4%

 

[toyo]

Great job with the delid, looks perfect. Your 8700k is not so bad if you need 1.35V in BIOS. 50x multi, no AVX offset, 44x Uncore. Only 17% make a 5GHz AVX @ 1.425 volts in the silicon lottery statistics. The prime95 temps show a lot of head room temperature wise. VRM temps of 80C-ish, peaking at 92C. Yes they are high but I ran 98-104c vrm temps for 5 years+ with no issues with lots of air flow. They dropped to 98c in prime 95 small fft's. This is with a rampage 4 black edition and the 4930k.

You could get a monoblock that covers both cpu and vrm area. That should hopefully bring drop VRM temps, if you can find one for your board. Random example https://www.overclockers.co.uk/ek-w...ix-z390i-d-rgb-monoblock-plexi-wc-9sy-ek.html

8700K	4.80GHz	4.60GHz	1.375V	100%
8700K	4.90GHz	4.70GHz	1.387V	Top 99%
8700K	5.00GHz	4.80GHz	1.400V	Top 83%
8700K	5.10GHz	4.90GHz	1.412V	Top 49%
8700K	5.20GHz	5.00GHz	1.425V	Top 17%
8700K	5.30GHz	5.10GHz	1.437V	Top 4%

SL probably tests with AVX2/SmallFFTs, other than that I can't explain their voltages. It's understandable to be honest. I played around quite a bit with this chip, within the limits of my motherboard, as at the time it was a large purchase of many components and I had to save on stuff I wouldn't otherwise. With undervolting, the chip was never fully stable until I got SmallFFTs with AVX2 stable for 12hrs or so, there was always that one app or game that would crash 2-3 time/month, and even the very rare BSOD. Once I passed those 12 hrs of AVX2 hardcore stress, there were no other game crashes or BSODs at all, for months.

Water cooling will probably wait a lot for my case. I've had quite a few friends that had weird (biological?) disgusting growth inside their AIOs or loops, and a few even had leaks that killed parts of their PC. I'm way too risk averse to go into it.

As for the chip itself, I can't really stresstest the 5GHz OC properly, there's no way I can draw the 250W I've seen for others with SmallFFTs/AVX2. At around 170-180W the VRMs go above 100-105 and while there's no throttling of any sort, it's also not comfortable to watch things get so toasty. I used 1.35V fixed because I know many just dial that in and run their PCs like that with the 8700K. The CPU is fine with 1.33V for p95/1344k inplace/no AVX, at least fine for 3hrs. Realbench also passes fine. But LinX/LinpackX, OCCT with AVX or P95 with AVX fail at 1.33 (and 1.35) almost on the spot.

But yeah, worth noticing that even with the 1.4V for 5GHz, SL still used the AVX offset, so that's basically a 4.8GHz OC when you run AVX. Even at 4.8GHz, with 1.4V I'm sure they hit 200W power draw figures, the Coffee Lakes are insanely power hungry once you OC them.

 

[zx128k]

SL probably tests with AVX2/SmallFFTs, other than that I can't explain their voltages. It's understandable to be honest. I played around quite a bit with this chip, within the limits of my motherboard, as at the time it was a large purchase of many components and I had to save on stuff I wouldn't otherwise. With undervolting, the chip was never fully stable until I got SmallFFTs with AVX2 stable for 12hrs or so, there was always that one app or game that would crash 2-3 time/month, and even the very rare BSOD. Once I passed those 12 hrs of AVX2 hardcore stress, there were no other game crashes or BSODs at all, for months.

Water cooling will probably wait a lot for my case. I've had quite a few friends that had weird (biological?) disgusting growth inside their AIOs or loops, and a few even had leaks that killed parts of their PC. I'm way too risk averse to go into it.

As for the chip itself, I can't really stresstest the 5GHz OC properly, there's no way I can draw the 250W I've seen for others with SmallFFTs/AVX2. At around 170-180W the VRMs go above 100-105 and while there's no throttling of any sort, it's also not comfortable to watch things get so toasty. I used 1.35V fixed because I know many just dial that in and run their PCs like that with the 8700K. The CPU is fine with 1.33V for p95/1344k inplace/no AVX, at least fine for 3hrs. Realbench also passes fine. But LinX/LinpackX, OCCT with AVX or P95 with AVX fail at 1.33 (and 1.35) almost on the spot.

But yeah, worth noticing that even with the 1.4V for 5GHz, SL still used the AVX offset, so that's basically a 4.8GHz OC when you run AVX. Even at 4.8GHz, with 1.4V I'm sure they hit 200W power draw figures, the Coffee Lakes are insanely power hungry once you OC them.

I have had leaks that have killed parts. Everybody who uses water cooling understands the risks of leaks. If you believe the vrm is too hot then you are doing the right thing. You might be able to limit the power draw in bios, so applications like prime95 get limited power but games don't hit maximum.