Intel Kaby Lake i7-7700K CPU De-Lid & Re-Lid Results @ [H]

Zarathustra[H] said:
Ah. Maybe I have misunderstood how those shims worked. I thought the shim sat around the core of the CPU with a hole in the middle exposing it directly to the water block, essentially just offloading some of the cooler weight and making it less likely that you chip your core.

Turns out they DO have kits for mounting on "naked" CPU's but they are just a set of different mounting screws and washers to account for the height difference of running naked.
Click to expand...

Oh you must mean a die protector?
 
200 minutes..... I gotta go do school stuff. Hope it is still running when I get back.

5.1-2666.png
 
Since RAM doesn't really seem to add too much performance, is there any chance that we can dial back the RAM OC for more CPU OC?
 
Now I'm waiting for the follow-up article [H]ow to get 5GHz on almost any 7700K!
 
Has Kyle stated why he reattached the IHS? When I delidded my 4790K the videos said to just place the cleaned IHS onto the core and new TIM.
 
Qrash said:
Has Kyle stated why he reattached the IHS? When I delidded my 4790K the videos said to just place the cleaned IHS onto the core and new TIM.
Click to expand...
The substrate the chip is mounted on is thinner/weaker than previous generations.
The full pressure of the heatsink pushes down on the core unless the lid has a solid foundation.
This causes the substrate to bend and could cause damage.

Also socket 1151 is weaker than previous generations.
If the pressure of the chip is made less uniform, the socket will distort more.
This will aid further bending of the substrate.
 
Boo. Just got home and it had crashed. 1.47v, here we come!
 
Nenu said:
The substrate the chip is mounted on is thinner/weaker than previous generations.
The full pressure of the heatsink pushes down on the core unless the lid has a solid foundation.
This causes the substrate to bend and could cause damage.

Also socket 1151 is weaker than previous generations.
If the pressure of the chip is made less uniform, the socket will distort more.
This will aid further bending of the substrate.
Click to expand...

OK, so how do we know how much sealant to use if we don't have a device like Kyle is using?

Also, I guess we should leave a gap in the sealant like Intel did, right?
 
Qrash said:
OK, so how do we know how much sealant to use if we don't have a device like Kyle is using?

Also, I guess we should leave a gap in the sealant like Intel did, right?
Click to expand...

You dont need to guess, however much sealant fits under the lid when it is resting on the chip is the right amount.
Let it set, you are golden.

The lid is set firmly in place resting on the die.
Pressure is applied evenly to the die and the rim of the lid.
 
Nenu said:
You dont need to guess, however much fits under the lid when it is resting on the chip is the right amount.
Let it set, you are golden.
Click to expand...

Nah, you need to compress the IHS as gasket sealant is pretty thick. With a rockit tool and relidder its pretty darn easy, you can dial in precise amounts of compression force to squeeze out excess sealant. Personally I don't use much sealant at all especially considering it takes a day or two for it to cure. Instead, I use two small drops of crazy glue in opposite corners and then line the rest with black rtv (leaving an air gap), then using the relidder, squeeze the excess out. The crazy glue cures in 2-3 mins, so then I can take out of the relidder and wipe the excess sealant off. Voila, it looks like a stock chip.
 
Intel could have used a fraction of the $ they saved not including heat sinks to solve this problem. It's unfortunate Intel feels so secure with their place in the market that it was found acceptable. In the past I've sanded down CPUs and heatsinks on plates of glass to gain a few degrees. Paid several hundred dollars on water cooling to drop temps a handful of degrees. Really ticks me off to see Intel cheap out on thermal paste when thinking how much additional risk, money, and effort the enthusiast needs to put forth to make up for it.
 
It's a shame you didn't get the over clock you wanted, but that temperature drop is very, very nice.
 
23 degrees drop didn’t add more OC speed even after throwing more vcore :eek:

The temp difference between the best and worst cpu cooler in the latest [H] review is 5 degrees, so I think it’s time for a new article about the cheapest/quietest CPU coolers that would “do the job” with Kaby Lake regardless of the cooling performance, worst cooling performance might even be a plus in the winter :D
 
Chas said:
Great fine and wonderful.

You got it stable enough to read the CPU.

How's it do in actual benchmarks that stress the CPU?
Click to expand...
About as expected, I'd imagine?
 

Attachments

  • Cinebench 5Ghz.PNG
    Cinebench 5Ghz.PNG
    379.1 KB · Views: 88
For those asking about relidding - it's enough to put a little glue/silicone at the 4 corners of the the IHS. The only purpose is to keep the IHS in place nothing else.
 
motqalden said:
not sure why everyone is calling this a wash... i would do this just for the temps. with that big of a difference i would bet stock Intel heat sink on delided chip would come pretty close to non delided chip with top of the line air cooler. while this comparison isn't realistic in a real world sense, the point is people pay big bucks to get a couple degree lower temps between the best end air cooling and water cooled setups.
Click to expand...

And a lower core temp gets you what? Sure you can run fanless. But if you aren't overclocking, the newest intel chips undervolt very well and would get you the same for things like media servers.
 
Last edited by a moderator:
Folterknecht said:
For those asking about relidding - it's enough to put a little glue/silicone at the 4 corners of the the IHS. The only purpose is to keep the IHS in place nothing else.
Click to expand...
And to spread the heatsink + cpu clamp pressure so its not all concentrated on the tiny cpu die.
 
Anyone think it's possible to buy some Indium solder and conjoin the die and IHS in an oven at 160C?

Would likely give some amazing results.
 
DejaWiz said:
Would likely give some amazing results.
Click to expand...
I dunno. From what I've read, the liquid metal stuff (like what Kyle used) are awfully damned effective. Not sure if it's worth that effort. I mean, it's already a PITA (thanks Intel!), so simpler is better for most of us.

Gawd...I haven't bought an AMD chip in ages. What do they do nowadays? Like on the Phenom II chips and the FX stuff?
 
Just got back from Fort Worth, watching my daughter wrestle...lo and behold, it made the 8 hour RealBench stress test. This was at 5.1/266/1.47(set in BIOS). Going to load the RAM up to 32GB (instead of the default 4GB that I used for this test) we use 32GB for stock testing) and see how it does.

5.1-2666.png
 
nice wonder how fast it will take for kabys to degrade my 2600k took little over 3 years befor it degraded enough i cant run 5ghz and 4.8 needs 1.49 now.
 
nice my 2600k degraded on 1.45 vcore over 3 years or so was sad when it started needing more volts.
 
my sabertooth p67 should of be able to handle the vcore i was pushing to the cpu fairly easy.
 
Nenu said:
Because the motherboard isnt being pushed to its limit.
Click to expand...
The one constant is that electronics drift. That motherboard probably was not even calibrated at the factory. Who knows what the actual cpu voltage it supplies, or if it has changed in 3+ years.
Where I used to work we had to calibrate our Fluke meters, oscilloscopes and other metrology once a year.
And speaking of aging electronis. The [H] recently mentioned a 1000+ watt "gold awarded" power supply that no longer could pass the [H]ards power supply testing. I know apple and oranges, but i guarantee that power supply was not "pushed to it's limits" on a regular basis, if at all.

Bal3W0lf said:
my sabertooth p67 should of be able to handle the vcore i was pushing to the cpu fairly easy.
Click to expand...
Agreed, the voltage is not the issue. The issue is if the voltage measurement is accurate and has not drifted over time.

Just playing devils advocate here but this is something to be aware of when talking about degradation over time.
 
Last edited:
Jim Kim said:
The one constant is that electronics drift. That motherboard probably was not even calibrated at the factory. Who knows what the actual cpu voltage it supplies, or if it has changed in 3+ years.
Where I used to work we had to calibrate our Fluke meters, oscilloscopes and other metrology once a year.
And speaking of aging electronis. The [H] recently mentioned a 1000+ watt "gold awarded" power supply that no longer could pass the [H]ards power supply testing. I know apple and oranges, but i guarantee that power supply was not "pushed to it's limits" on a regular basis, if at all.
Click to expand...
That doesnt imply the motherboard failed.
If it supplies a slightly different voltage, its easy to adjust.
 
Nenu said:
That doesnt imply the motherboard failed.
If it supplies a slightly different voltage, its easy to adjust.
Click to expand...
WOAH.
I never said anything about failure. I am talking about drift and degradation.
For example:
If the motherboard displays 1.45 volts and delivers 1.45 volts in 2014.
BUT in 2017 it displays 1.45 volts but actually only outputs 1.43v.
Would you trust a 3 year old never calibrated motherboard and it's related voltage measuring systems, no, you would want it verified with another measurement device.
That's all I'm saying.
 
Jim Kim said:
WOAH.
I never said anything about failure. I am talking about drift and degradation.
For example:
If the motherboard displays 1.45 volts and delivers 1.45 volts in 2014.
BUT in 2017 it displays 1.45 volts but actually only outputs 1.43v.
Would you trust a 3 year old never calibrated motherboard and it's related voltage measuring systems, no, you would want it verified with another measurement device.
That's all I'm saying.
Click to expand...

yea i understand that if it has its driffted alot looks like its going to need 1.55vcore now to run 5ghz semi stable but cpu and motherboard could of degraded in some way who knows.
 
Jim Kim said:
WOAH.
I never said anything about failure. I am talking about drift and degradation.
For example:
If the motherboard displays 1.45 volts and delivers 1.45 volts in 2014.
BUT in 2017 it displays 1.45 volts but actually only outputs 1.43v.
Would you trust a 3 year old never calibrated motherboard and it's related voltage measuring systems, no, you would want it verified with another measurement device.
That's all I'm saying.
Click to expand...
The motherboard isnt being pushed to its limit, not close.
Even a 20% drift wont stop it maintaining an overclock.
 
Bal3W0lf said:
yea i understand that if it has its driffted alot looks like its going to need 1.55vcore now to run 5ghz semi stable but cpu and motherboard could of degraded in some way who knows.
Click to expand...
You're going to have to change you signature now, it still says 1.49. lol
I'll show myself out.
 
Nenu said:
The motherboard isnt being pushed to its limit, not close.
Even a 20% drift wont stop it maintaining an overclock.
Click to expand...
Once again I did not even infer nor did I say anything about the motherboard being pushed hard or to its limits.
You're not grasping what I'm saying. But since Bal3W0lf is the only person who needs to understand it and he does, I'm out. peace
 
Jim Kim said:
Once again I did not even infer nor did I say anything about the motherboard being pushed hard or to its limits.
You're not grasping what I'm saying. But since Bal3W0lf is the only person who needs to understand it and he does, I'm out. peace
Click to expand...
Its easy enough to tell if a motherboard is past its time re CPU voltage.
If the voltage set and voltage read by its onboard sensors show results that differ over time, it needs investigation.
Thats not the same as the CPU needing more voltage.

I havent encountered a motherboard that suffers from this.
Yet I've seen many an overclocked CPU that needs more voltage over time.
 
You must log in or register to reply here.
Back
Top