Intel Skylake / Kaby Lake

[Ajay]

Z370 chipset launch October.
Z390 chipset launch Next Year !

https://translate.google.com/translate?hl=en&sl=ja&tl=en&u=http://pc.watch.impress.co.jp/docs/column/ubiq/1076326.html

Nice find. So Cannon Lake PCH must be DMI 3.0 - again :sadface: Now, I'm actually hoping that the Ice Lake PCH is different - either integrated into the SoC, or using a PCIe 4.0 phy.

 

[pantsaregood]

Well, I am expecting the 8700K to be able to use memory speeds that the 7700K could not use.

Skylake could regularly hit 4133 MHz RAM frequency and Kaby Lake could regularly hit 4266 MHz on the right motherboards. You're unlikely to see Coffee Lake improve on that in any capacity just because DDR4 isn't going to go much further. As of now, only Samsung B-die appears to be able to consistently hit speeds that high.

Simply put, it doesn't really matter if they increase the memory controller multiplier range if RAM can't handle it. If you're that worried about memory bandwidth, go with Skylake-X. The memory controller appears to be just as excellent as Skylake/Kaby Lake, but it offers twice as many channels. Just be aware that once a CPU is no longer bandwidth-starved, increasing RAM frequency offers diminishing returns.

 

[LTC8K6]

Skylake could regularly hit 4133 MHz RAM frequency and Kaby Lake could regularly hit 4266 MHz on the right motherboards. You're unlikely to see Coffee Lake improve on that in any capacity just because DDR4 isn't going to go much further. As of now, only Samsung B-die appears to be able to consistently hit speeds that high.

Simply put, it doesn't really matter if they increase the memory controller multiplier range if RAM can't handle it. If you're that worried about memory bandwidth, go with Skylake-X. The memory controller appears to be just as excellent as Skylake/Kaby Lake, but it offers twice as many channels. Just be aware that once a CPU is no longer bandwidth-starved, increasing RAM frequency offers diminishing returns.

I'm not worried at all about 2 channel memory bandwidth.

If I were, I'd go with Kabylake-X.

 

[Ajay]

I'm also expecting impressive memory speeds with the 8700K.

That's up to the IMC and the quality of the motherboard. I'm expecting that most vendors will offer a revisions of their Z270 motherboards for enthusiasts since both the Z370 and Coffee Lake will be socket compatible. I'm not sure if X299 boards support more memory speeds because they have more layers and present cleaner signals due to routing changes. Most reviews sites don't talk about board layers anymore

 

[PeterScott]

Hopefully, Kaby-R should push the Mac Mini to quad cores... I always figured this would only happen after Intel started phasing out dual cores in the iCores. Mini is way overdo for an update, Quad Kaby-R could make it meaningful.

 

[IntelUser2000]

https://translate.google.com/translate?hl=en&sl=ja&tl=en&u=http://pc.watch.impress.co.jp/docs/column/ubiq/1076326.html

The translation also tells us due to PL2 power state being 44W(previous 15W chips were 19W) the Notebook platform has to be redesigned because it would require substantial increase in power delivery and thermals.

The performance in early benchmarks are between i5 6300HQ and i5 7300HQ in the Cinebench benchmark. The real world all-core Turbo is then ~2.9GHz.

It continues the reason for having Turbo. KBL-R would bring greatest benefits in multi-threaded workloads that run for short-term.

There will be no IPC increase, only additional cores.

Traditionally, process refinements like Kaby Lake and Coffee Lake have been referred to as CPU stepping instead of new codenames.

Close, but not exactly. Process refinement is a natural thing, while these changes are deliberate.

While you could say Pascal was a die-shrink of Maxwell, Nvidia states they had to do significant circuit work to enable high frequencies(that also means without the work, 16nm Pascal may not have been much faster than 28nm Maxwell).

So in terms of work done, compared to a decade ago even a shrink requires lot more work.

For Kabylake(and likely Coffeelake) it has media features backported from Cannonlake. I would count KBL-R and Devils Canyon as a process refinement product.

 

[Arachnotronic]

The translation also tells us due to PL2 power state being 44W(previous 15W chips were 19W) the Notebook platform has to be redesigned because it would require substantial increase in power delivery and thermals.

The performance in early benchmarks are between i5 6300HQ and i5 7300HQ in the Cinebench benchmark. The real world all-core Turbo is then ~2.9GHz.

It continues the reason for having Turbo. KBL-R would bring greatest benefits in multi-threaded workloads that run for short-term.

Probably why Intel is bothering to release CNL-U 2+2.

 

[IntelUser2000]

Probably why Intel is bothering to release CNL-U 2+2.

Yea, and the Gen 10 graphics.

 

[Arachnotronic]

Yea, and the Gen 10 graphics.

Oh, and LPDDR4, too, should improve efficiency.

 

[pantsaregood]

I'm not worried at all about 2 channel memory bandwidth.

If I were, I'd go with Kabylake-X.

I'm not understanding what you're getting at then.

You believe Coffee Lake will bring higher IPC because of RAM bandwidth.

Kaby Lake and Skylake are already pushing up against the limits of DDR4, so Coffee Lake won't be able to provide higher bandwidth.

Now you say you don't care about increasing RAM bandwidth? I don't understand.

EDIT: Forgot, but Kaby Lake-X has the same dual-channel controller as Kaby Lake does, so that's also an irrelevant point. The point I'm trying to get at is that Skylake and Kaby Lake, short of some miracle advance in DDR4 manufacturing, have already pushed dual-channel DDR4 to its practical limit. If you're concerned with getting more bandwidth (as you implied by saying Coffee Lake would allow higher RAM bandwidth), then the only real option is to move to a quad-channel platform.

Also, process refinement and die shrinks don't inherently require tweaking the way Pascal did - it was not a direct die shrink of Maxwell. An example of a direct die shrink is Prescott-Cedar Mill. Cedar Mill was only "better" due to lower power consumption, which was a direct result of the die shrink.

Intel's deliberate changes from Skylake-Kaby Lake-Coffee Lake are quite similar to what Intel would traditionally call CPU steppings - such as Cedar Mill's B1, C1, and D0. They were all process refinements, but Intel made no effort to market them as new products.

 

[IntelUser2000]

Intel's deliberate changes from Skylake-Kaby Lake-Coffee Lake are quite similar to what Intel would traditionally call CPU steppings - such as Cedar Mill's B1, C1, and D0. They were all process refinements, but Intel made no effort to market them as new products.

I told you about the backported media. The part about Nvidia was just an example(most people cluster Pascal as a straight shrink Maxwell BTW, because they don't know or care). AMD had to use transistors to improve clocks for Vega too. If the new processes automatically bring much higher clocks as some claim, why did they have to do this? Because they don't, or its quite small.

Also AMD was the one that opted for more continuous improvements versus Intel where they front load everything at once and steppings brought minimal 100MHz changes on a refresh. If you can improve a lot that means you are either at the top of the execution game or you just suck in the beginning. The part that Intel brought process fame wasn't just bringing new node generation, but actual transistor performance was far better than the competition.

Now, Intel has to do it as well, because their first 14nm was sucky.

The most recent example of refinement is the RX 580. Sure it allows peak frequency to increase but power to increase significantly as well. It's pretty clear if you look at results alone today's refinements are at a shadow of what it used to be.

Everyone has to do this. Process shrinks bring far less gains than they used to, in addition to requiring much more effort. More effort, less gain.

 

[IntelUser2000]

Oh, and LPDDR4, too, should improve efficiency.

How much are we expecting for CNL GT2 performance? Since they have CFL-U 15W/28W with GT3e, I think they can't exceed those.

I am guessing 10% low end and 30% high end for estimates. That suggests a minimal increase in terms of performance per EU.

 

[nvgpu]

http://www.androidpolice.com/2017/0...-version-of-google-home-alongside-new-phones/

Wonder if the new Chromebook Pixel will be using KBL-R 8250U, previous 2015 model had a BDW 5200U or 5500U in the LS model.

 

[IntelUser2000]

http://www.androidpolice.com/2017/0...-version-of-google-home-alongside-new-phones/

Wonder if the new Chromebook Pixel will be using KBL-R 8250U, previous 2015 model had a BDW 5200U or 5500U in the LS model.

Does it even make sense to have an expensive Chromebook?

Most of the Notebook seem to say "$999 and up". It's ridiculous. I know the Notebook market is greater, but it doesn't seem to be worth it for the price. Desktops are far better IMO. Go for Desktop + Tab/Smart versus Laptop. The former is far more mobile/ergonomic/comfortable.

 

[TheF34RChannel]

I like these RAM discussions! All I want to know is the speed I need for my gaming purposes (highest = better isn't exactly true as at a certain point it adds nothing but additional monetary costs) without the need for any tweaking in order to get it working. Who knows... But yeah, depends on the board and IMC alright.

 

[kirbyrj]

Does it even make sense to have an expensive Chromebook?

Most of the Notebook seem to say "$999 and up". It's ridiculous. I know the Notebook market is greater, but it doesn't seem to be worth it for the price. Desktops are far better IMO. Go for Desktop + Tab/Smart versus Laptop. The former is far more mobile/ergonomic/comfortable.

I disagree. I have a desktop for home and a laptop for travel. I haven't had a tablet in years. Everything I'd want a tablet to do, I do on my cell phone. Everything that my cell phone is too small for, I take out my laptop.

 

[pj-]

Does it even make sense to have an expensive Chromebook?

Most of the Notebook seem to say "$999 and up". It's ridiculous. I know the Notebook market is greater, but it doesn't seem to be worth it for the price. Desktops are far better IMO. Go for Desktop + Tab/Smart versus Laptop. The former is far more mobile/ergonomic/comfortable.

I agree with kirbyrj. I tried the "desktop + tablet" thing for a couple years and it sucked. I had a cherry trail 10" full windows 10 tablet for years and it was ONLY good for passive content consumption. Typing this post on it would be a nightmare. I bought a 2-in-1 laptop fairly recently and it's so much better. Aside from the keyboard making it easier to type, its presence allows the screen to be supported even when I'm not typing. At 15", it's far too large to actually use as a tablet, but the "tent" and "reverse" modes mean I don't have to use it as a tablet. I set it on my lap or nearby and it holds itself up. If I need to type something more than a few words I just fwip the keyboard around and type 10 times faster than I can with an on screen keyboard.

Desktop + laptop is the best if you can afford it.

 

[IntelUser2000]

Deleted

 

[IntelUser2000]

I disagree. I have a desktop for home and a laptop for travel. I haven't had a tablet in years. Everything I'd want a tablet to do, I do on my cell phone. Everything that my cell phone is too small for, I take out my laptop.

Alright. I get it for Travel.

Typing this post on it would be a nightmare.

If you care about mobility, you can get a keyboard for your Tab. Actually the Atom based 2-in-1s that come with a keyboard is perfect.

 

[pj-]

Alright. I get it for Travel.



If you care about mobility, you can get a keyboard for your Tab. Actually the Atom based 2-in-1s that come with a keyboard is perfect.

The keyboards I've seen for tablets don't interest me. They are typically cramped because of the size and they generally don't support the screen, or if they do it's only at a single fixed angle. Even if there was a keyboard that made it bearable to type a few sentences on the tablet, there's no way it would ever be usable for stuff like coding which I can and have done on my laptop a few times. I will take the tablet on my next vacation because I will be less bummed if it gets lost or stolen but after that I will probably give it to a friend for their kids to use. I don't care about mobility but if I did I would just get a smaller laptop.

The whole tablet category is already on the decline because people are realizing tablets are dumb. Android sucks as a tablet OS, ipads have good hw but are just big iphones, and cheap chinese windows tablets are now being replaced with decent macbook clones.

 

[IntelUser2000]

The whole tablet category is already on the decline because people are realizing tablets are dumb. Android sucks as a tablet OS, ipads have good hw but are just big iphones, and cheap chinese windows tablets are now being replaced with decent macbook clones.

That's true of the PC market too.

August IDC Data shows 37 million Tablets shipped, while for PC market it is 60 million. Assuming 60% share for Laptops that puts Tablet = Laptop

The whole deal with declining market can be explained single handedly: Saturation. There's too many computers nowadays. It's bursting from every hole and pocket. Even in third world countries. People in Affrican countries that take the spotlight when talking about hunger and crime issues have a Smartphone!

That discussion can be a thread by itself though.

 

[Bouowmx]

I know Wikipedia is freely editable but a user added Kaby Lake-R per-core Turbo frequencies.

Core i5: Regardless of number of active cores, Turbo frequency is the same? Now the 8250U's can't be right because 1.6 GHz + (100 MHz × 19) = 3.5 GHz, which is greater than its specs, 3.4 GHz. But still?

Core i7: A more typical behavior: more active cores, lower frequency.

Now I await reviews about sustained performance.

 

[IntelUser2000]

I know Wikipedia is freely editable but a user added Kaby Lake-R per-core Turbo frequencies.

Core i5: Regardless of number of active cores, Turbo frequency is the same? Now the 8250U's can't be right because 1.6 GHz + (100 MHz × 19) = 3.5 GHz, which is greater than its specs, 3.4 GHz. But still?

Core i7: A more typical behavior: more active cores, lower frequency.

Now I await reviews about sustained performance.

It seems right. ExtremeTech has the information too: https://www.extremetech.com/computi...ocessors-built-kaby-lake-add-additional-cores

I don't think the all core Turbo is sustainable unless all you are using it is on a bursty workload anyway. It has to come back down to meet PL1 Turbo(long term). Even then, the power management has to be able to handle instantaneous 44W PL2 requirements.

 

[beginner99]

The 8700k will probably be the only cpu I will not fiddle with even for 100Mhz more, for day to day use.

I will probably get it and see what happens if I up the multiplier by 3, eg. 5 ghz single-core turbo and be done with it if it works good enough. As far as I understood that would also mean 4.5 ghz all core turbo.

 

[moonbogg]

So 24 PCI-E lanes on the 8700K?