Intel Skylake / Kaby Lake

[HilbertSpace]

Does anyone have a good picture of a de-lidded Kaby Lake CPU? I was thinking of ordering this - https://shop.aquacomputer.de/product_info.php?products_id=3378 - if it can match up with Kaby Lake too. I'm slowly getting all the parts together for a water-cooled Kaby Lake mini-ITX build in an NCASE M1, and thought it would be fun to try de-lidding the CPU too for cooler temps/higher overclocks.

 

[RichUK]

Does anyone have a good picture of a de-lidded Kaby Lake CPU? I was thinking of ordering this - https://shop.aquacomputer.de/product_info.php?products_id=3378 - if it can match up with Kaby Lake too. I'm slowly getting all the parts together for a water-cooled Kaby Lake mini-ITX build in an NCASE M1, and thought it would be fun to try de-lidding the CPU too for cooler temps/higher overclocks.

I'll have one on Thursday, all being well.

 

[Ranulf]

Saw this in the other thread, synthetic benchmark sneak peek from HOCP:

http://www.hardocp.com/article/2016/12/05/intel_core_i77700k_cpu_synthetic_benchmark_sneak_peek/1

 

[Nothingness]

Saw this in the other thread, synthetic benchmark sneak peek from HOCP:

http://www.hardocp.com/article/2016/12/05/intel_core_i77700k_cpu_synthetic_benchmark_sneak_peek/1

Looks like it's been taken down: it gives me an "Error 520" and it's not linked from main page.

EDIT: it now works

 

[Ranulf]

Yeah, I saw that earlier. It seems like it is back up now though. Anyway, looks like KL and SL are well, the same. At least in Passmark, Timespy and SiSoft.

"Again we see a tiny difference in scores with the Kaby Lake 7700K showing only a 0.291% performance increase." (in SiSoft).

 

[coercitiv]

Yeah, I saw that earlier. It seems like it is back up now though. Anyway, looks like KL and SL are well, the same. At least in Passmark, Timespy and SiSoft.

"Again we see a tiny difference in scores with the Kaby Lake 7700K showing only a 0.291% performance increase." (in SiSoft).

Folks at tomshardware must be using a different 7700K.

Using that average of both Core i7-6700K configurations as a baseline, you can see that the Core i7-7700K performs about 8% better in applications that are primarily CPU limited.

 

[mikk]

Not really. Testing at the same clock and default clock for the SKU are two different things. You should now that Kabylake clocks higher. Of course IPC is the same, not sure why some people are going crazy when they see a 1% difference. Even with everything the same, in some benches there are small fluctuations from run to run. And the BCLK clock from the board could differ slighty and so on. The biggest joke is this thread: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...ck-worse-than-i7-6700k.2493728/#post-38614823

Less than 1% difference and some people came to the conclusion that Kabylake IPC has been decreased. Just lol.

 

[coercitiv]

Not really. Testing at the same clock and default clock for the SKU are two different things.

Wait, people are still looking for arch improvements?!

We need a sticky on this forum ASAP -> Kaby Lake = Skylake + 300Mhz + improved media engine. End of sticky.

 

[TheELF]

Wait, people are still looking for arch improvements?!

We need a sticky on this forum ASAP -> Kaby Lake = Skylake + 300Mhz + improved media engine. End of sticky.

Plus probable HTT on pentiums plus probable quad i3,which would be first steps for next gens mainstream hexa i7.

 

[coercitiv]

plus probable quad i3,which would be first steps for next gens mainstream hexa i7.

Could you please point to a source for that? I'd love to see the reasoning behind a quad i3 with clocks under i5 7400, meaning around 3Ghz.

 

[LTC8K6]

http://hexus.net/tech/reviews/cpu/99820-intel-core-i7-7700k-14nm-kaby-lake/

Didn't see this Hexus review posted.

 

[TheELF]

Could you please point to a source for that? I'd love to see the reasoning behind a quad i3 with clocks under i5 7400, meaning around 3Ghz.

#8357mikk, Today at 2:05 PM
Look at OP link to DSOgaming ,they have expressview as the source,last page of the review they comment on having a look at a future time at the pentium with HT and an quad i3,might just be bad translating but it says so.
http://www.expreview.com/51056-6.html

However, the seventh generation Core Duo processor in the low-end product changes will be more obvious, later we will see the dual-core four-threaded Pentium processor and even quad-core Core i3, and high-end Core i5 / i7 is only prompted frequency Compared with the system, the low-end product changes will be more interesting.

 

[LTC8K6]

I have not heard of 4C i3 before, except for that one place.

Intel needs to move everything up.

2C/2T Celeron or dropped
2C/4T Pentium
4C/4T i3
6C/6T i5
6C/12T i7

 

[TheELF]

I have not heard of 4C i3 before, except for that one place.

Yup,could be they meant 4 threads but google missheard

 

[Arachnotronic]

Wait, people are still looking for arch improvements?!

We need a sticky on this forum ASAP -> Kaby Lake = Skylake + 300Mhz + improved media engine. End of sticky.

I am surprised that people didn't sit there looking for IPC improvements in Bristol Ridge, either. AMD did the exact same thing that Intel did with Kaby.

 

[witeken]

We have a little more information about TSMC's 7m process.

https://twitter.com/Siliconicsdick/status/806153075622711297

With a metal pitch of 40nm (using SAQP), TSMC 7nm metal pitch is a "stunning" 0.77x smaller than Intel's 14nm (52nm).

For comparison, at 14nm, Intel shrunk the metal pitch by 0.65x. So even if they now only shrink by 0.7x (normal Moore's Law), Intel's 10nm pitch will be smaller than TSMC's 7nm.

This is how misleading the node names have become. Intel's 10nm will not be close to TSMC 7nm (like how their 14nm is pretty close to their 10nm), but it will in fact be smaller.

 

[TraumaRN]

I mean looking at these initial numbers I'm just going to get a Sky Lake on the cheap in Jan/Feb...no real need to get Kaby Lake for my new system.

 

[Sweepr]

First Kaby Lake-S Core i7-7700K Previews/Reviews


- bit-tech: Intel Core i7-7700K Preview - Gaming/3D

Now, any results should be taken with a pinch of salt. No two CPUs overclock the same so anyone saying Kaby Lake is a great overclocker just because their one sample is a daemon of an overclocker simply doesn't grasp the overclocking landscape particularly well. It's a lottery, pure and simple.

However, we managed to get our sample to 5.1GHz - the best result we've seen for an age and even better than we managed with Sandy Bridge.

Sure, this did need a hefty 1.44V and temperatures were edging into the 90s using Prime95's small FFTs test for stress testing with a Corsair H105 CPU cooler, but there is a glimmer of hope here that we may be dealing with a rather decent overclocker when it comes to Kaby Lake.

www.bit-tech.net/hardware/cpus/2016/12/05/core-i7-7700k-performance-overclock-preview/1


- Hexus: Review: Intel Core i7-7700K (14nm+ Kaby Lake)

http://hexus.net/tech/reviews/cpu/99820-intel-core-i7-7700k-14nm-kaby-lake

 

[LTC8K6]

I mean looking at these initial numbers I'm just going to get a Sky Lake on the cheap in Jan/Feb...no real need to get Kaby Lake for my new system.

By then we should have full reviews of KL, so we will know.

I'm not sure SL chips will be that much less expensive in 1Q, though.

 

[raghu78]

We have a little more information about TSMC's 7m process.

https://twitter.com/Siliconicsdick/status/806153075622711297

With a metal pitch of 40nm (using SAQP), TSMC 7nm metal pitch is a "stunning" 0.77x smaller than Intel's 14nm (52nm).

For comparison, at 14nm, Intel shrunk the metal pitch by 0.65x. So even if they now only shrink by 0.7x (normal Moore's Law), Intel's 10nm pitch will be smaller than TSMC's 7nm.

This is how misleading the node names have become. Intel's 10nm will not be close to TSMC 7nm (like how their 14nm is pretty close to their 10nm), but it will in fact be smaller.

dude first of all Intel has not revealed their 10nm minimum metal pitch. So stop guessing what it could be. The fact that Intel increased gate pitch with 14nm Plus proves that they were a bit too aggressive with 14nm and thus you might want to hold off on making any scaling assumptions. Its quite logical that Intel might play it safe at 10nm and chose less aggressive density scaling targets (especially wrt original Intel 14nm).

http://www.anandtech.com/show/10610...six-notebook-skus-desktop-coming-in-january/4

Secondly TSMC is a pureplay foundry who supports dozens of fabless customers spanning multiple markets like mobile, HPC, GPU, FPGA etc. and and their design rules are nowhere near as restrictive as Intel who is a IDM and pretty much designs the process for their CPUs. Thirdly with such huge difference in the process metrics like transistor contacted gate pitch and M1 metal pitch Intel was never able to translate it into actual product advantage especially in mobile where low power Atom rarely beat chips like Apple Ax series on actual product transistors per sq mm. Apple A10 with 3.3 billion transistors comes in at 125 sqmm using TSMC 16FFC. Now show me an Atom chip which is significantly denser than A10 given that both are low power and run at clock speeds of around 2.4 Ghz.

btw I do agree that foundry node naming is a joke. I have always thought that Intel node naming was better.

 

[Dave2150]

I have not heard of 4C i3 before, except for that one place.

Intel needs to move everything up.

2C/2T Celeron or dropped
2C/4T Pentium
4C/4T i3
6C/6T i5
6C/12T i7

I suspect something similar:

i7 = 6C/12T Unlocked
i5 = 4C/8T Unlocked
i3 = 4C Locked
Pentium = 2C/4T
Celeron = removed.

 

[RichUK]

First Kaby Lake-S Core i7-7700K Previews/Reviews


- bit-tech: Intel Core i7-7700K Preview - Gaming/3D

www.bit-tech.net/hardware/cpus/2016/12/05/core-i7-7700k-performance-overclock-preview/1


- Hexus: Review: Intel Core i7-7700K (14nm+ Kaby Lake)

http://hexus.net/tech/reviews/cpu/99820-intel-core-i7-7700k-14nm-kaby-lake

5.1Ghz, 1.44v, P95 Small FFTs and only edging into the 90s on a H105 - I doubt it from what I've seen!!!

 

[TheELF]

I suspect something similar:

i7 = 6C/12T Unlocked
i5 = 4C/8T Unlocked
i3 = 4C Locked
Pentium = 2C/4T
Celeron = removed.

No,celeron still 2c/4t but with less cache just like now.
Celeron will never go away.

 

[Dave2150]

No,celeron still 2c/4t but with less cache just like now.
Celeron will never go away.

Eventually it will, everything comes to an end. We all die, all companies disolve, etc etc It might take a 100 years, it might take a thousand. Who knows

 

[Dave2150]

5.1Ghz, 1.44v, P95 Small FFTs and only edging into the 90s on a H105 - I doubt it from what I've seen!!!

Golden samples might have been sent, considering all the negative publicity Kabylake and the 14nm+ process has attracted so far. This coupled with Zen's imminent release could well have opened the golden coffers