Intel Kaby Lake

[cobito]

Since a thread about Zen has been opened here, I'm opening one about Kaby Lake, a generation that seems pretty boring in terms of new features. You can read some details here. Basically they are Skylakes with some new features that aren't too exciting: support for USB 3.1 and hardware 4K video decoding. Apart from that, it will be the last generation at 14nm before moving to 10nm in about a year (if we're optimistic).

Well, some images of a Core i5-7600K have been published today, but they are just that, images without any tests or anything.

In principle, the i5-7600K should go to 3.8GHz with Turbo Boost of 4GHz, so it seems that the one in the image is a bit overclocked. You can find out more at Guru3D.

Let's see if this rehash gives any surprises, although I think the thread about Zen is going to be much more interesting.

[whoololon]

Well, that's a good idea, so they can't accuse us of only taking money from AMD. :rofl:
By the way, MSI has had to withdraw the BIOS that was supposed to support those processors; that said, only MSI has had problems with the code.

[Obione]

I won't change until they put an x6 on mainstream platform today.

[Yorus]

I like Zen because it fits more to what I am looking for on paper, but when the time comes, you know... search, compare...

[cobito]

Well, Anandtech has published the possible models and features. It seems that they go up by about 200MHz compared to Skylake.

Let's see how the tables work in Markdown...

Kaby Lake	Núcleos/Hilos	Frecuencia base	TDP	*	Skylake	Frecuencia base
i7-7700K	4/8	4.2GHz	95W	*	i7-6700K	4.0GHz
i7-7700	4/8	3.6GHz	65W	*	i7-6700	3.4GHz
i7-7700T	4/8	2.9GHz	35W	*	i7-6700T	2.8GHz
i5-7600K	4/4	3.8GHz	95W	*	i5-6600K	3.5GHz
i5-7600	4/4	3.5GHz	65W	*	i5-6600	3.3GHz
i5-7600T	4/4	2.8GHz	35W	*	i5-6600T	2.7GHz
i5-7500	4/4	3.4GHz	65W	*	i5-6500	3.2GHz
i5-7500T	4/4	2.7GHz	35W	*	i5-6500T	2.5GHz
i5-7400	4/4	3.0GHz	65W	*	i5-6400	2.7GHz
i5-7400T	4/4	2.4GHz	35W	*	i5-6400T	2.2GHz
E3-1205v6	?/?	3.0GHz	?	*	-	-

[cobito]

A Dutch store has already put the Kaby Lake on pre-sale and the prices at the moment seem 30-40% higher than the current 6000 series.

You can see the full list at Guru3D.

[whoololon]

[cobito]

More benchmarks of the 7700K are coming out this time on the Z270 platform and the difference compared to the 6700K (200-300MHz less) is about 8%. Considering that the frequency difference between both models is 5-7%, it represents a relative improvement of about 1%, which can be considered the same error as the tests.

In short, Kaby Lake does not represent absolutely any improvement in performance compared to Skylake.

These results have been published on Guru3D.

[cobito]

Reviews of the 7700K keep coming out and it seems that it performs better in OC than its predecessor. In one it was pushed 50MHz higher than the 6700K while keeping the voltage at 1.3V and air-cooled. In another it was pushed to 5.1GHz at 1.44V, this time water-cooled and reaching dangerous temperatures.

It's worth remembering that these reviews are done with samples provided by Intel, so they probably didn't send random processors.

[cobito]

It seems that Intel is putting a pretty bad thermal paste between the silicon die and the IHS. On Anandtech forums someone has changed the paste and the temperatures have improved quite a bit.

[Obione]

These guys want to cut our wings (or for us to cut their IHS xD) from the Ivy. Maybe it's a dark deal with CPU cooler manufacturers ¬¬

[whoololon]

I agree that the best thing is to do the "delidding" and put the spacer on it, and save the IHS (like AMD on the K7).
Besides, they already sell gadgets to unseal them safely; there are even projects for you to print said gadgets in 3D.

[kynes]

I'm going to be a conspiracy theorist. Isn't it very strange that Intel has come to a sudden stop at 14nm when the others have gone from 28 to 20 to 16/14 to 10 in no time? Maybe they are simply maximizing profits and if they see that Zen bothers them, they release the 8xxx in a few months and already in a new technology. The tim problem would confirm that they don't want the common user to do a beastly overclock, to have room for performance increase in case they need it. Squeeze AMD but don't drown them.

[cobito]

@kynes said in Intel Kaby Lake:

To put pressure on AMD but not to suffocate them.

AMD represents a thin layer between Intel and the antimonopoly laws and, being thin, it is fragile. Sometimes I think that the issue of Netburst and Kaby Lake is not really coincidental. On both occasions, it has coincided that Intel has released a particularly mediocre microarchitecture and AMD has released one that is particularly remarkable. But who knows...

[whoololon]

But let's face it: AMD is Intel's white label.
:rofl:

[kynes]

@whoololon for Intel is a "enemy" little bothersome. It's like cobito says, a life insurance to avoid being accused of dominant position, if they die perhaps they force intel to release x86, or worse for them, microsoft to develop a version for arm64, which would be the death for intel of almost all its market.

The alliance between qualcomm and microsoft to make a layer of emulation of x86-32 over the snapdragon processors is a first step...

[cobito]

A Russian has overclocked a 7700K to 7Ghz. The animal has set it to 2V. Now, disabling two cores.

Source.

[krampak]

The incredible thing is that it has passed long benchmarks, it hasn't just been for the photo.

[cobito]

It seems that an interesting change is coming for the low end: the Kaby Lake Pentiums bring HyperThreading, thus differentiating themselves a bit more from the Celerons.

Source.

[cobito]

Some Germans have overclocked the 7700K to 7.3GHz with liquid nitrogen: