Talk:Kaby Lake

Come On

Not significant? This is a microarchitecture change, as significant as any other. Well, maybe not as earth-shattering as Skylake, but still deserves mention. This article needs to be fleshed out as facts become known, not deleted. It's also significant as it is a symptom of the breakdown of Moore's Law.Davidlark (talk) 18:32, 3 October 2015 (UTC)

Windows support

This should be clarified if possible with currently available information. Old windows versions normally always work on newer CPUs because CPUs are backwards compatibel (in this case with AMD64). Does Intel plan to use a different principal CPU architecture than AMD64 for Kaby Lake (in this case, the info box in the article would be wrong)? Or is this just about new instruction set extensions and maybe Microsoft driver components for the GPU part? --MrBurns (talk) 11:43, 19 January 2016 (UTC)

175.140.122.244 is vandalizing the article

A guy from malaysia consistently spread false and unsourced info in this article, we must block him from this article.

Then take it up at WP:AIV or request protection at WP:RPP - do not add a semi-protected request to an unprotected page - Arjayay (talk) 12:52, 27 January 2016 (UTC)

Socket 2066, not 1151

According to some rumors, Intel's Kaby Lake will sport a new socket, coded 2066. I'm currently looking for some english sources other than WCCFTech. 190.211.111.173 (talk) 03:57, 19 July 2016 (UTC)

That's Kaby Lake-X, the HEDT variant. Those always use a different socket than the mainstream CPUs.--24.230.151.194 (talk) 20:16, 21 July 2016 (UTC)

What is a HEDT variant? i never heared of this. --MrBurns (talk) 00:32, 27 August 2016 (UTC)

coreboot support

It has been announced that coreboot will support Kaby Lake: https://www.coreboot.org/pipermail/coreboot/2016-August/081857.html --87.79.170.158 (talk) 13:46, 7 August 2016 (UTC)

Encode/Decode support

There have been an number of edits and reverts relating to encode/decode support. All sources cited and that I can find list for encode: H.264, H.265 and VP9 8-bit only, NOT VP9 10-bit. For decode: H.264, H.265 and VP9 8-bit and 10-bit are supported. Perhaps the language was confusing, I have rephrased it twice now. If there is another citable/reputable source which shows VP9 10-bit encode support, this would be in direct contradiction to other sources, and we should talk about it here. Dbsseven (talk) 21:49, 1 September 2016 (UTC)

Skylake merge proposal

Aside from minor changes in GPU and a speed bump in CPU, Kaby Lake doesn't offer new substantial features which could warrant a separate article. I propose merging this article with Skylake. Artem-S-Tashkinov (talk) 15:16, 9 September 2016 (UTC)

While they do seem very similar in features, they are marketed distinctly separately. I think that people who search for "Kaby Lake" would be better served with a separate Kaby Lake article, where all Kaby Lake information is collected. Also, for completeness I would prefer to have a separate article for each architecture code name. Thue (talk) 19:20, 9 September 2016 (UTC)

Do not merge. While the are similar in CPU architecture, there are significant differences in the GPU and encode/decode. As these are Intel defined product families, it is probably best if we do not externally try to decide 'ad hoc' what is similar enough. Dbsseven (talk) 20:49, 9 September 2016 (UTC)

Improved 14nm process?

I'm just curious why the edit on Aug. 30 at 17:01 was made to sound like Kaby Lake has a new improved manufacturing process. In reality, it has the same process as both Skylake and Broadwell. In the architecture section, it points out the "improved process" as an architecture feature. Improved compared to what? Can we clarify this? It seems like the process should be mentioned only once since it's nothing new. — Preceding unsigned comment added by 99.197.234.70 (talk) 12:38, 18 September 2016 (UTC)

I don't know if it's fair to say nothing is new. While the process size may not change, details of the process (and it's maturity) can allow for improved performance on the new process. But you are right this should be clarified. These technical details are less commonly discussed in the references, so finding a citable source may be difficult. Dbsseven (talk) 03:07, 5 October 2016 (UTC)

Desktop processors

I recently put up the (currently leaked) desktop processors of the Kaby Lake family and the table has since been removed, due to lack of credible sources.

http://wccftech.com/intel-kaby-lake-desktop-lineup-leak/

WCCFTech is not THE most credible source, but they're up there.

https://www.techpowerup.com/225217/intel-core-kaby-lake-desktop-processors-first-wave-detailed

I won't talk for TechPowerup, but they agreed with WCCTech and added Turbo Boost.

http://www.tweaktown.com/news/53662/intel-core-i7-7700k-kaby-lake-performance-teased/index.html

TweakTown confirms the 7700K, which lends credence to the other reports.

http://www.kitguru.net/components/cpu/matthew-wilson/retail-leak-shows-specs-of-intel-kaby-lake-7700k-and-7600k-cpus/

KitGuru confirms the 7700K as well as the 7600K.

...or is Intel themselves the only source that truly matters? Well, yes, but... 85.227.222.109 (talk) 17:33, 5 October 2016 (UTC)

Determining credible sources for leaked information is very difficult, but critical. If a product is confirmed then any source will do, but how do you trust a poor-credibility source's specs of an unconfirmed product? In these cases, I believe it is reasonable to require a higher standard in the sources. But this is a perfect place to try and find consensus on this issue. In this case the TweakTown cites a Estonian site that has been taken down, and WCCFtech. WCCFtech has had some problems with getting the Kaby Lake capabilities right recently, and other sources are skeptical of the 7700K claims. Dbsseven (talk) 15:46, 8 October 2016 (UTC)

Added a signature for you, just so it wouldn't look as though I was talking to myself...anyway, what WCCFTech was having troubles with was specific benchmark results, since it seems something is going on with clock speeds (or possibly cores used) in the benchmarked CPUs. Benchmarks are (in general) irrelevant to WP articles on CPU architectures, whereas the "standard" (not overclocked, not underclocked) clock speeds (and names) of CPUs is. These seem fairly certain. Regarding your second link, I'm unfortunately unable to listen to podcasts or view videos - is the source sceptical of claimed nomenclature, clock speeds or benchmark results? 85.227.222.109 (talk) 19:58, 10 October 2016 (UTC)

"Dual Core" CPU Clock Rate

The specs for the mobile processors list "single core" and "dual core" clock rates for each. Comparing the numbers to Intel's specs pages, the Wikipedia page lists "Processor Base Frequency" as the "single core" speed and "Max Turbo Frequency" as the "dual core" speed. This not my area of expertise, so I'm not going to make the change, but - isn't Max Turbo Frequency not the same thing as "dual core" speed? The Intel website defines Max Turbo Frequency as "Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology." (e.g. [1]) I'm unclear as to what "dual core" speed is supposed to mean. Thanks in advance to whomever comes along and knows more about this than me. Feel free to make changes if necessary. LakeHMM (talk) 00:30, 10 November 2016 (UTC)

I restored the correct table headers from an old version of the page, now they make sense. 190.163.199.153 (talk) 22:46, 20 November 2016 (UTC)

Can pls somebody add the rest of the portfolio?

Source: http://www.3dcenter.org/news/weitere-modell-daten-zum-kaby-lake-portfolio-inkl-des-core-i3-7350k-bekannt

Kerne = Cores; Basetakt = base clock; Skylake-Vorgänger = Skylake precessor;

Thanks! -- 09:24, 16 November 2016 (UTC) — Preceding unsigned comment added by 194.0.94.12 (talk)

This need to be from a reputable and citable source prior to inclusion. There have been a number of rumors of what the mainstream desktop parts will be, but none from citable sources. I don't believe there is consensus on this source being citable. See Identifying_reliable_sources Dbsseven (talk) 16:11, 21 November 2016 (UTC)

What... is it?

This article just starts off by talking about all the new features of "Kaby Lake". It's 14nm! What is it, though? A chicken pot pie? A new form of currency? An IUD? Come on, this article has a middle but no beginning! moeburn (talk) 14:33, 24 November 2016 (UTC)

TDP

The line about TDP has been edited by myself and another editor repeatedly. I would like to discuss it here to find a consensus on how to include it.

The two versions are:

1. Thermal design power or (TDP) is the maximum heat generated by the chip. As power increases with voltage and frequency, TDP is directly proportional to computational power of a chip.

2. Thermal design power or (TDP) is the maximum heat generated by the chip. As power increases with voltage and frequency, TDP is directly proportional to thermal dissipated power of a chip.

I would argue that #2 is redundant, TDP is the (designed) thermal dissipated power of a chip.

My point in the first definition it that a microarchitecture's computation power or instructions per second (IPS) ${\text{IPS}}$ ∝ ${\text{clock}}\times {\frac {\text{IPs}}{\text{cycle}}}$ is proportional to frequency. And TDP is proportional to frequency $P=CV^{2}f$. This leads to ${\text{IPS}}$ ∝ ${\frac {P}{CV^{2}}}\times {\frac {\text{IPs}}{\text{cycle}}}$. Therefore, the maximum computational power is ultimately proportional by the TDP. Dbsseven (talk) 16:26, 8 December 2016 (UTC)

TDP is a unit of heat, not power dissipation. A chip with TDP of 96 W might dissipate around 185 W. The entire TDP classification section needs to do away anyway, it's out of scope. --CyberXRef^☎ 09:07, 7 January 2017 (UTC)

Why no memory column?

Kaby Lake has now support for DDR4 dual channel 2400 mhz and DD3L dual channel 1600 mhz. — Preceding unsigned comment added by 62.57.198.153 (talk) 20:11, 5 January 2017 (UTC)

It's not sensible to add a feature to the table that is common in almost all of the processors. You can add this in the "Features" section. Mehshiraz (talk) 14:33, 7 January 2017 (UTC)

It will asap Skylake-X apears on 2017. We shoul do the table with the same format than the previous architecture generations since Sandy Bridge, and not change the format now. — Preceding unsigned comment added by 62.57.198.153 (talk) 21:39, 8 January 2017 (UTC)

Move discussion in progress

There is a move discussion in progress on Talk:Haswell (CPU) which affects this page. Please participate on that page and not in this talk page section. Thank you. —RMCD bot 05:31, 11 February 2017 (UTC)

Significant digits of clock speed on Intel CPU articles

I'd suggest that Wikipedia begin recording the frequency of Intel chips to an extra digit. For example, Intel records the frequency of their flagship Kaby Lake processor, the i7-7700k, as 4.20 GHz base. Note that they include a "0" after the tenths place in order to signify to the reader that it is not actually anything like 4.49 GHz or 4.53 GHz, but in fact exactly 4.50 GHz. While there will always be some uncertainty in these numbers, Intel has chosen to record three digits for clock speeds on ark.intel.com. As such, I believe that it would be sensible to record the clock speeds to this amount of precision, as Intel has done. —Atvelonis (talk) 18:36, 27 February 2017 (UTC)

Bumping. Anyone have thoughts on this? —Atvelonis (talk) 15:38, 9 March 2017 (UTC)

That's actually not true and incorrect. The use of '0' on ark is mainly to be consistent and not an indication of sigfigs. The base frequency (core clock; CCLK) of an Intel chip is the product of your bus speed (BCLK) and your core multiplier. Note that the bus speed is a virtual value nowadays. Take for example the Xeon X5660 - 2.80 GHz/3.20 GHz or this Xeon L5638 - 2.00 GHz/2.40 GHz. In both examples Intel specified a '0' however it's not a sigfig because in both cases the BCLK is 133.33 MHz, therefore their "real" on-paper speeds, if we want to be pedantic, would be 2799.99 GHz/3199.99 GHz and 1999.99 GHz/2399.99 GHz respectively. --CyberXRef^☎ 19:25, 25 April 2017 (UTC)

Ah, thanks for the correction then. —Atvelonis (talk) 02:19, 1 May 2017 (UTC)

to be truly pedantic though. 0.99 is equal to 1.0. We probably have that somewhere on wikipedia. Multiply by 10 substract 1 and divide by 9.Carewolf (talk) 11:47, 2 May 2017 (UTC)

On paper, not in practice. It's never really 99.99... Oscillators may be 99.99 or 99.94 or 99.74 etc.. Google images "cpuz". Anyway my point was that Ark doesn't really use sigfigs, it's just a formatting thing. --CyberXRef^☎ 08:02, 11 May 2017 (UTC)

There are always variations which is another reason to stick with official numbrs. But the clock being analog it could a "perfect" third of something. Carewolf (talk) 10:53, 12 May 2017 (UTC)

Name /origin?

I can't find anything here about why it's named "Kaby Lake". Does anybody know anything about the origin of or reasoning behind this name? Albrecht Conz (talk) 21:37, 5 June 2017 (UTC)