i7 8700k overclocking question(s)

[Elthrael]

I haven't OCd a CPU in literally more than a decade, so I just want to make sure I'm up to speed.

First off, I just want to check that I'm properly equipped. I have an ASUS Prime Z370-A board and a CoolerMaster MasterLiquid 240mm AIO watercooling system. I have a Fractal Design Define R6 case with two front intake fans and one back exhaust fan. My chip isn't delidded, but I get good thermals as-is: around 30°C idle and after doing a six-core 100% load torture test for 2 hours, the temp was 65°C. I think I shouldn't have issues with temperatures with this setup since I used to OC chips even with stock coolers and cheap air coolers, but if you see anything questionable, let me know.

I'm not a stranger to overclocking, I know how it works, multipliers, voltages, the lot, there's just one thing I'm puzzled about when it comes to overclocking these days, and that's boost clocks.

I know modern processors usually run at a low frequency when idle and then boost up to a high frequency when under load. My question is: when you OC a CPU, I assume you turn off all the frequency boost stuff and just use a constant frequency for all six cores at the overclocked frequency? So my chip (after OC) will no longer jump from 800MHz to 4.3GHz on all six cores, but will stay on 4.8 or 5.0 or whatever I can push it to all the time?

By the way, I'm fully aware that for gaming, the performance difference is very minimal and hardly noticeable. I'm a tweaker and I like fiddling with stuff I have to make it better or different. So help a tweaker out, willya? :D

[FeilDOW]

Overclocking sure can be noticable depending on your GPU and monitor.

https://youtu.be/CoUtA7DKXhU

Here is a good video to start you off. Switching your power plan to extreme in windows is what will stop your CPU from down clocking.

[Elthrael]

FeilDOW の投稿を引用：

Overclocking sure can be noticable depending on your GPU and monitor.

https://youtu.be/CoUtA7DKXhU

Here is a good video to start you off. Switching your power plan to extreme in windows is what will stop your CPU from down clocking.

Hahah that's the exact video I was saving to watch, thanks anyway!

I think I already did that WAY back when I installed Windows, but I'll check to be double sure.

[Elthrael]

So I followed the guide and I got it at 5.0GHz @ 1.35V with no problems, ran Prime95 for an hour and Cinebench R15 benchmark and no issues.

However, my thermals are abysmal, while running Prime95, all cores peaked at way over 90°C, Core#2 even hit 100°C which naturally throttled down the CPU back to 3.7GHz.

I guess my only options are trying to lower the voltage some or delid the bastard?

[FeilDOW]

eS の投稿を引用：

So I followed the guide and I got it at 5.0GHz @ 1.35V with no problems, ran Prime95 for an hour and Cinebench R15 benchmark and no issues.

However, my thermals are abysmal, while running Prime95, all cores peaked at way over 90°C, Core#2 even hit 100°C which naturally throttled down the CPU back to 3.7GHz.

I guess my only options are trying to lower the voltage some or delid the bastard?

Both my 8700k's are delided and thermals are much better, I very rarely see above 75°c at 5.0. Like you said either delid or drop your clocks a couple hundred MHz along with voltage.

[Elthrael]

FeilDOW の投稿を引用：

Both my 8700k's are delided and thermals are much better, I very rarely see above 75°c at 5.0. Like you said either delid or drop your clocks a couple hundred MHz along with voltage.

I tried fiddling around with it some more, and it's kind of pointless, I went to 4.8GHz, then 4.6GHz and tried dropping the voltage but anything under 1.325V produces a lot of errors in prime95 while still running hot as hell (85-90°C or more).

I'll maybe try the Rockit Cool delidding kit, seems like a very solid and safe way.

[FeilDOW]

eS の投稿を引用：

FeilDOW の投稿を引用：

Both my 8700k's are delided and thermals are much better, I very rarely see above 75°c at 5.0. Like you said either delid or drop your clocks a couple hundred MHz along with voltage.

I tried fiddling around with it some more, and it's kind of pointless, I went to 4.8GHz, then 4.6GHz and tried dropping the voltage but anything under 1.325V produces a lot of errors in prime95 while still running hot as hell (85-90°C or more).

I'll maybe try the Rockit Cool delidding kit, seems like a very solid and safe way.

That's what I bought to delid mine and I recommend it, it's very straight forward and easy to do.

[tacoshy]

eS の投稿を引用：

FeilDOW の投稿を引用：

Both my 8700k's are delided and thermals are much better, I very rarely see above 75°c at 5.0. Like you said either delid or drop your clocks a couple hundred MHz along with voltage.

I tried fiddling around with it some more, and it's kind of pointless, I went to 4.8GHz, then 4.6GHz and tried dropping the voltage but anything under 1.325V produces a lot of errors in prime95 while still running hot as hell (85-90°C or more).

I'll maybe try the Rockit Cool delidding kit, seems like a very solid and safe way.

if you already delidd, look for a custom pure copper IHS to replace the stock one. they cost like 10-20$ max while the have a larger surface and are not nickle plated.
Also gives a decent temperature decrease when overclocking.

On the other hand CM doesnt build good AiO IMO. Currently there is pretty much only one choice with the Corsair Pro lineup that dont go with Asetek pumps anymore. And the ML-Series fans that come with it are much better.

[SeriousCCIE]

Also.. consider purchasing a liquid metal TIM (thermal interface material) for your CPU lid and cooler block, if you are not using aluminum for a heatsink.

It can make a significant difference in your ability to draw heat away from a chip that is overclocked/really hot for any reason.

[tacoshy]

SeriousCCIE の投稿を引用：

Also.. consider purchasing a liquid metal TIM (thermal interface material) for your CPU lid and cooler block, if you are not using aluminum for a heatsink.

It can make a significant difference in your ability to draw heat away from a chip that is overclocked/really hot for any reason.

highly depends on the cases. For most it doesnt make a difference to use LM between IHS and Coldplate. But if he delidds he has to get LM anyways. The whole reason to delidd is to sue LM between Die and IHS.

[SeriousCCIE]

It's odd to just emit the heat to the lid of the CPU and then use regular grease on top of the surface, isn't it? No comment was mentioned as to the materials used, so I thought I'd mention it.

Anyway, the prospect of delidding was raised, without also mentioning the type of TIM. It is helpful as a compound--whether or not the cpu is delidded.

Just make sure that the contact material/heat sink/surface plating is not aluminum.

[tacoshy]

SeriousCCIE の投稿を引用：

It's odd to just emit the heat to the lid of the CPU and then use regular grease on top of the surface, isn't it? No comment was mentioned as to the materials used, so I thought I'd mention it.

Anyway, the prospect of delidding was raised, without also mentioning the type of TIM. It is helpful as a compound--whether or not the cpu is delidded.

Just make sure that the contact material/heat sink/surface plating is not aluminum.

no the difference between LM and non conductive TIM is less then the margin of error. If you ask Omega who tried it out it gave no improvement at all same for others. If you do high overlocks (I did one with 1.465V) then the difference for me was 4-5C compared to the NZXT Kraken Stock TIM. You can get the about the same with other non conductive TIM's

[Elon Mosque]

tacoshy の投稿を引用：

eS の投稿を引用：

I tried fiddling around with it some more, and it's kind of pointless, I went to 4.8GHz, then 4.6GHz and tried dropping the voltage but anything under 1.325V produces a lot of errors in prime95 while still running hot as hell (85-90°C or more).

I'll maybe try the Rockit Cool delidding kit, seems like a very solid and safe way.

if you already delidd, look for a custom pure copper IHS to replace the stock one. they cost like 10-20$ max while the have a larger surface and are not nickle plated.
Also gives a decent temperature decrease when overclocking.

On the other hand CM doesnt build good AiO IMO. Currently there is pretty much only one choice with the Corsair Pro lineup that dont go with Asetek pumps anymore. And the ML-Series fans that come with it are much better.

I would advise against copper IHS as gallium will make an alloy and eventually gallium will dry as it bonded itself with copper. Best to keep the original IHS or look for other metals.

[tacoshy]

You know that the stock IHS is copper too? Was never an issue and they are specifically made as replacements in case you decide. Even Der8auer sells such CPU's with warranty through Caseking. Can't be such an issue then.

[Elon Mosque]

tacoshy の投稿を引用：

You know that the stock IHS is copper too? Was never an issue and they are specifically made as replacements in case you decide. Even Der8auer sells such CPU's with warranty through Caseking. Can't be such an issue then.

Do you know that stock IHS is copper with nickel layer on top? That's what makes the electron migration so slow with gallium. Der8auer ofc sells those products in the end he has to make money. Buying the CPU directly from them would make a sense but even that do you also know this happens as well?
https://www.reddit.com/r/overclocking/comments/ant5v6/silver_ihs_reacted_badly_with_liquid_metal_help/?utm_source=amp&utm_medium=&utm_content=post_body
Edit: You can see how gallium formed alloy with silver in that example.
Copper gallium alloy doesnt happen overnight that's why in the short term you cannot observe any side effects of this but in the long run the gallium will diffuse with copper which will happen a lot shorter than a cpu IHS which has an nickel plate.
A lot of people just keep repeating "aluminum no no but copper is okay" but there's more than that in metals...

[UTFapolloMarine]

eS の投稿を引用：

So I followed the guide and I got it at 5.0GHz @ 1.35V with no problems, ran Prime95 for an hour and Cinebench R15 benchmark and no issues.

However, my thermals are abysmal, while running Prime95, all cores peaked at way over 90°C, Core#2 even hit 100°C which naturally throttled down the CPU back to 3.7GHz.

I guess my only options are trying to lower the voltage some or delid the bastard?

depends on your cooler and thermal quality,