My i7 7700k Has Arrived - Insights, Benchies & Overclocks Inside (Now With Delid!)

[RichUK]

Thanks for the pictures. I believe those are the first pictures of a Kaby Lake de-lidded anywhere. I spent a lot of time trying to find some. Looks 'almost' identical to Skylake, but hard to tell about die size... my quick image analysis with counting pixels results in ~124-125 mm^2 versus 122 mm^2 for Skylake. Location seems a bit different too, die looks a little closer in one direction and further in the other to the gold marking triangle than for Skylake, so it looks like this product (https://shop.aquacomputer.de/product_info.php?products_id=3378) might not fit Kaby Lake too.

The die is not dead centre, it's slightly closer to the top of the pcb.

 

[HilbertSpace]

My rough pixel counting measurements put Kaby at 9.2 mm from top side (end with notches), versus 10.8 mm for Skylake. And ~13.8 mm from triangle side versus 14.3 mm for Skylake.

 

[RichUK]

My rough pixel counting measurements put Kaby at 9.2 mm from top side (end with notches), versus 10.8 mm for Skylake. And ~13.8 mm from triangle side versus 14.3 mm for Skylake.

Here are some more, if it helps / you're interested:

https://s30.postimg.org/97zrj5fgh/image.jpg

https://s23.postimg.org/c91oadnnf/image.jpg

https://s29.postimg.org/50d40oth3/image.jpg

 

[RichUK]

I'll test properly with Prime95 tomorrow, but passing IBT at 5.1Ghz @ 1.376v and only 72 degrees max temps is encouraging.

5.1Ghz seems to be the ceiling, as even with 1.54v, 5.2Ghz still failed IBT.

 

[TheELF]

Shut up and run the ryzen graphics file in blender...

(Just kidding with the shut up, but pretty pretty please? )
http://www.amd.com/en-us/innovations/new-horizon
directly under the video

 

[RichUK]

Shut up and run the ryzen graphics file in blender...

(Just kidding with the shut up, but pretty pretty please? )
http://www.amd.com/en-us/innovations/new-horizon
directly under the video

I'll run it tomorrow.

 

[UncleCrusty]

Nice results! 5100 MHz is a particularly satisfying number for me.

 

[Phynaz]

It is a crying shame that Intel ships these things with the poor TIM. Great results!

I thought it was confirmed that the TIM is fine, it's the distance between the chip and heatspreader that causes the higher temps?

 

[FlanK3r]

5100 Mhz IBT is awesome man! How much you can run Cinebenchs? At 5300 Mhz is possible (1.45V or so)?

 

[TheManWithTheGoldenGun]

what case is this?? looks clean asf

searched powered by primo, all i found is the phanteks? thanks

 

[FMPVince]

Hi,
Fantastic findings RichUK, could you run Geekbench 4 on your latest 5.1 delided achievement ?
The 5Ghz (is it ?) GB you posted at 5879 single core seems disapointing

 

[imported_bman]

RichUK you are famous.

http://wccftech.com/intel-core-i7-7700k-delid-performance-tests/

Hello wccftech scavengers!

 

[RichUK]

RichUK you are famous.

http://wccftech.com/intel-core-i7-7700k-delid-performance-tests/

Hello wccftech scavengers!

Nice

 

[RichUK]

The journey to rock solid 5Ghz continues.

Some more interesting finds. Before delid the temps in Prime 95 v28.7 Small FFTs were insane and would almost instantly fail - i believe the increased temps destabilises overclocks (from what I've read, due to noise caused by heat).

After delid, I was able to run Prime 95 v28.10 small FFTs (latest version!) at 5Ghz 1.392v for over 7 hours before one logical core failed.

At 5Ghz using Prime 95 v28.10 Small FFTs, I observed:

• 1.36v fails in 9 mins (one logical core)
• 1.376v fails in 17 mins (one logical core)
• 1.392v fails on one logical core in 7 hours 18 minutes.

The pic shows 1.408v as the load dropped after one logical core failed (this is vdroop in play):

 

[RichUK]

Hi,
Fantastic findings RichUK, could you run Geekbench 4 on your latest 5.1 delided achievement ?
The 5Ghz (is it ?) GB you posted at 5879 single core seems disapointing

I should be able to do a 5.2Ghz run in Geekbench - I'll do it later.

I may try 5.3Ghz

 

[coercitiv]

At 5Ghz using Prime 95 v28.10 Small FFTs, I observed:

• 1.36v fails in 9 mins (one logical core)
• 1.376v fails in 17 mins (one logical core)
• 1.392v fails on one logical core in 7 hours 18 minutes.

Did you happen to write down the max temps for each of the sessions above?

 

[vissarix]

The journey to rock solid 5Ghz continues.

Some more interesting finds. Before delid the temps in Prime 95 v28.7 Small FFTs were insane and would almost instantly fail - i believe the increased temps destabilises overclocks (from what I've read, due to noise caused by heat).

After delid, I was able to run Prime 95 v28.10 small FFTs (latest version!) at 5Ghz 1.392v for over 7 hours before one logical core failed.

At 5Ghz using Prime 95 v28.10 Small FFTs, I observed:

• 1.36v fails in 9 mins (one logical core)
• 1.376v fails in 17 mins (one logical core)
• 1.392v fails on one logical core in 7 hours 18 minutes.

The pic shows 1.408v as the load dropped after one logical core failed (this is vdroop in play):

Prime is absolute crap, ive had an overclock for long time, using it for everything, benchmarking, gaming, and any other kind of stuff...tried prime and it failed instantly...

But i dont care about prime stability since ive never had a blue screen before using this prime crap..

So dont worry about prime stability..

 

[plopke]

Very nice thread to read , all tough I am a bit shocked how much improvement you got after delid. The gap is ridicules before and after. So if am seeing this correct you got a ±30C difference out of it ,...?

 

[.vodka]

¿The gap is ridicules before and after. So if am seeing this correct you got a ±30C difference out of it ,...?

Yup, that staggering number is more or less the usual difference you get by removing the black adhesive and using a liquid metal TIM at the extremes of what a CPU can be pushed to. You'll see results like these on Ivy, Haswell and now Skylake and Kabylake.

Intel used to solder their high end CPUs to the IHS... now you have to do this equivalent process if you intend to make the most out of your purchase.


---------------


Nice to see 5GHz stable! That's quite the powerhouse.

 

[dark zero]

So.. the deciding this time screwed up big time on the TIM? The hell Intel!

30 C of temperature difference is more than noticeable. Is the vital difference between a 4.9 OC to 5+ ones.

Also SB clocks at last! Only if Intel should have done right the TIM or just used the soldering....

 

[realibrad]

So.. the deciding this time screwed up big time on the TIM? The hell Intel!

30 C of temperature difference is more than noticeable. Is the vital difference between a 4.9 OC to 5+ ones.

Also SB clocks at last! Only if Intel should have done right the TIM or just used the soldering....

This has been going on for a while now actually. Its not all that new.

 

[LTC8K6]

So.. the deciding this time screwed up big time on the TIM? The hell Intel!

30 C of temperature difference is more than noticeable. Is the vital difference between a 4.9 OC to 5+ ones.

Also SB clocks at last! Only if Intel should have done right the TIM or just used the soldering....

Should you really judge KL overall by this one example?

 

[RichUK]

Out of curiosity I googled 7700k delid and saw a number of websites reporting on my findings of the 30 degree temp drop, but were unsure on cooling method in both tests so thought I should clarify: The before and after tests were 100% identical (with exception of delid) to isolate the delid, both tests used:

• The same cooler at the same setting: Kraken X62 - Fan @ 50% and pump at @ 65% (by the way, this cooler is only about 3 degrees better than the H110i, but has better config available)
• Exactly the same BIOS settings - which was a x50 multiplier and voltage offset with LLC level 2 resulting in a stable 1.344v under prime 95 load
• Memory stock (C15) and Uncore stock.
• Prime 95 v27.9 run for 15 minutes in both tests.
• The same case / config / install (tested with case side panel installed - i.e. how i would actually use it for 24/7 use so the test would be more meaningful to me): https://s30.postimg.org/gzluzd7w1/Case.jpg

 

[[DHT]Osiris]

Should you really judge KL overall by this one example?

Unless something went really terribly wrong with this one sample (which would be indicative of bad QA), it's indicative of the entire line, yes. Every single one may not be exactly like this, but you cannot ignore a 30c delta as a result of the TIM/spreader design.

 

[thesmokingman]

I'm gonna have to delid my skylake so it doesn't feel so inadequate now! Thanks a lot Rich, it's all your fault.