Discussion: Intel Cannonlake (Formerly Skymont) & Kaby Lake

[TechyInAZ]

Welcome to the Intel Cannonlake and Kaby Lake discussion thread!

Until a MegaThread is officially created, I hope this thread will be a good spot for discussion on Intel's new code-named Cannonlake and Kaby Lake CPUs. I will also be updating this thread (as much as I can) on the latest news, articles and rumors regarding these new architectures.

*********************************************
Updates:

5/31/2016: Intel’s new “Kaby Lake” CPUs are already showing up in PCs at Computex

6/2/2016: Intel Roadmap Confirms High-End Skylake-X and Kaby Lake-X Processors in Q2 2017 – Mainstream Kaby Lake-S Desktop Chips in Q4 2016

8/26/2016: Intel Kaby Lake Mainstream Quad Core Processors Landing in December This Year – ‘U’ Series Processors In September

9/17/2016: Intel Core i7-7700K Flagship Kaby Lake Processor Listed For Pre-Order – 7th Gen CPUs May Work With 100-Series Chipset

11/13/2016: Intel Kaby Lake - release date, price, specs and rumours

*********************************************

My original review of Cannonlake.

Spoiler

Even with the new 6th Gen Skylake CPUs that have just launched, Intel is continuing its constant future goal of making the microprocessor smaller, faster, and more power efficient than it's previous generations.

Cannonlake: The 10nm CPU

Welcome to Intel's future CPU lineup called Cannonlake. Cannonlake will be Intel's first CPU lineup based on the 10nm die.

Predictions tell that the 10nm tech should improve power savings over the current 14nm tech. However it does make me wonder if the 10nm will have flaws (like leakage), just like most tocks in the tick tock cycle Intel uses. Beyond this, nobody knows much.

Other predictions/rumors have told that Intel will be bringing 8 core CPUs to the mainstream market and 10 core CPUs for the Extreme Edition CPUs.

Performance Boost Over 14nm:

Honestly after seeing the awful performance boost coming from 22nm Haswell to 14nm Skylake, my best guess at performance boosts is a measly 5-7%. However it is to be expected.

CPU wafers are getting so small now that Intel is even struggling to make them. This is because the smaller you get, the more complex it becomes. CPUs are EXTREMELY fragile at 10nm, so for example overvolting will probably kill your CPU faster than in previous generations.

However there are always more ways to make a CPU faster. One way is increasing core clock but I don't think that will be a big contributor since 10nm is so fragile to work with. I believe Intel's best option is to bump the core count like rumors are predicting. If quad cores become the norm and hex + octo cores go mainstream, that will already be a massive boost in performance. This will be especially true since developers for OSes, games, APIs, apps etc are already getting out of single threaded programming.

Release Date:

The official release date as of NOW is 2H of 2017. But I remain skeptical. Remember friends that Intel's planned release dates for Skylake and Cannonlake are WAY WAY off from what is now reality.

In 2014 we should of had Broadwell (5th gen) that COMPLETELY replaced Haswell. Haswell refresh would have never existed because it wasn't needed. Then by 2015 Q1 or Q2 we should of had Skylake, then by 2016 we should of had Skymont (now Cannonlake).

This is what Intel had planned but as you can see it is completely different. It's taking Intel longer to make smaller CPUs forcing them to use "refreshes" to fill in the gap (Kaby Lake...cough cough).

However it is not entirely Intel's fault. Let face it, the lower the nm, the harder it is to make a CPU, which does make sense because your adding in a bunch of extra variables, more IPCs, more fragile transistors etc. which = reliability issues.

Then of course we all have to expect the day in the future that's inevitable. There will be no more shrinking! You can only make a microprocessor so small, which means eventually shrinking a CPU is going to be history.

*********************************************


Thank you for reading! Feel free to discuss cannonlake and kaby lake to your hearts content!

Also, if you have more news on cannonlake or skylake & Kaby lake, please feel free to post the link down below and I will try my best to add it to this thread as soon as possible!

Thank you,
TechyInAZ

 

[BadActor]

It's interesting to see whats going to shake out. There has to be a limit in die size (and we may be very near that) where it won't be economically feasible to make them smaller, especially when your not getting big increases in performance. The conundrum is if you simply add more cores and transistors, how long does it take for the software to catch up to the point that that will give a big enough performance benefit to entice people to upgrade? I could see getting a little longer use out of a system built now or in the near future.

 

[synphul]

I believe Intel has already suggested that 10nm is the exhaustion point for pure silicon. That's why IBM along with others have pushed past to SiGe for their 7nm chips.

http://www.pcworld.com/article/2946124/ibm-reveals-worlds-first-working-7nm-processor.html

Just like ram is growing harder and harder to update along the current dram path and Intel and Micron are looking at something like HMC.

 

[TechyInAZ]

I believe Intel has already suggested that 10nm is the exhaustion point for pure silicon. That's why IBM along with others have pushed past to SiGe for their 7nm chips.

http://www.pcworld.com/article/2946124/ibm-reveals-worlds-first-working-7nm-processor.html

Just like ram is growing harder and harder to update along the current dram path and Intel and Micron are looking at something like HMC.

That's interesting. I did not know sillicon can only be shrunk to 10nm. Lets see if they can make it to 1nm.

 

[synphul]

That's according to intel's remarks, not really sure until then. They may be able to go a bit further but I think this is why IBM and others are already looking to alternatives. Even if it doesn't move entirely from silicon adjustments are having to be made and new materials considered.

 

[iNonEntity]

With all the problems Intel faces in their future CPUs, I find it hard to believe they will be worth upgrading to unless for extremely intensive work. Their improvements seem minimal compared to the issues they propose. Especially since I'm somewhat against upgrading to Win10 for compatibility (I love my Win7 </3) I'm rather green to understanding architecture, though. Maybe they'll blow us away with something crazy. Regardless - I've been saving up for a new computer with Cannonlake and either Pascal or Volta, so I hope it turns out good

 

[TechyInAZ]

http://wccftech.com/intel-skylake-x-kaby-lake-x-q2-2017-roadmap-leak/

Looks like we now have a firm roadmap of kaby lake and skylake-x.

The X sounds so much better than -E. I'm glad Intel changed the lettering.

 

[turkey3_scratch]

I question what besides silicon they could use. For one thing, our silicon already isn't pure silicon, it is p-type and n-type. Germanium has already been used in n-doped silicon. I think p-doped can be something like Boron if I recall.

Any other alternative to silicon would either be a larger element or some compound. Nothing will be ideal, because Silicon is the smalles element on the periodic table with semi-conductive properties. If we are using larger compounds, we're going to have to make our CPUs proportionally larger to account for larger transistors!

 

[TechyInAZ]

I question what besides silicon they could use. For one thing, our silicon already isn't pure silicon, it is p-type and n-type. Germanium has already been used in n-doped silicon. I think p-doped can be something like Boron if I recall.

Any other alternative to silicon would either be a larger element or some compound. Nothing will be ideal, because Silicon is the smalles element on the periodic table with semi-conductive properties. If we are using larger compounds, we're going to have to make our CPUs proportionally larger to account for larger transistors!

So how about FinFET?

 

[turkey3_scratch]

I question what besides silicon they could use. For one thing, our silicon already isn't pure silicon, it is p-type and n-type. Germanium has already been used in n-doped silicon. I think p-doped can be something like Boron if I recall.

Any other alternative to silicon would either be a larger element or some compound. Nothing will be ideal, because Silicon is the smalles element on the periodic table with semi-conductive properties. If we are using larger compounds, we're going to have to make our CPUs proportionally larger to account for larger transistors!

So how about FinFET?

That would still have to have the doped silicon I'd think. How else will the depletion layer be made thinner by having electrons fall into p holes?

Note: My transistor knowledge is little but I know some very specific stuff. Maybe I just sound like a total donkey, dunno.

 

[sam1275tom]

The end is 7nm, that's a physical limit.

 

[Epicness937]

The end is 7nm, that's a physical limit.

but then of course someone will most likely find something new and get it better and such and on and on it goes heck one day people may be able to make artificial brains and put them in pcs

 

[TechyInAZ]

The end is 7nm, that's a physical limit.

but then of course someone will most likely find something new and get it better and such and on and on it goes heck one day people may be able to make artificial brains and put them in pcs

Fortunately silicon isn't the end of the line. With future biological computing being the road ahead, I'm really excited to see how that will turn up.

Basically in the next 10 to 15 years we theoretically should have real supercomputers in ultrabooks via biological chips which use less than 90% of the power and heat today's CPUs use.

 

[turkey3_scratch]

The end is 7nm, that's a physical limit.

but then of course someone will most likely find something new and get it better and such and on and on it goes heck one day people may be able to make artificial brains and put them in pcs

Fortunately silicon isn't the end of the line. With future biological computing being the road ahead, I'm really excited to see how that will turn up.

Basically in the next 10 to 15 years we theoretically should have real supercomputers in ultrabooks via biological chips which use less than 90% of the power and heat today's CPUs use.

I ain't buying it till I see it

 

[Epicness937]

The end is 7nm, that's a physical limit.

but then of course someone will most likely find something new and get it better and such and on and on it goes heck one day people may be able to make artificial brains and put them in pcs

Fortunately silicon isn't the end of the line. With future biological computing being the road ahead, I'm really excited to see how that will turn up.

Basically in the next 10 to 15 years we theoretically should have real supercomputers in ultrabooks via biological chips which use less than 90% of the power and heat today's CPUs use.

I ain't buying it till I see it

same but it could happen
hope it is good

 

[TechyInAZ]

Yeah no kidding. Who knows what the cons are of bio over silicon. Skynet? :lol:

 

[Epicness937]

Yeah no kidding. Who knows what the cons are of bio over silicon. Skynet? :lol:

no one really knows but this point as far as we know 7nm is a limit for silicon but who knows whats next could be a .00001nm 100ghz .5watt 5000 core cpu in 20 years rendering what we have now crap
this is why i have a hard time investing a a powerful pc.....
i mean look at the cpus from 20 years ago only industrial computers have more than one core as far as i know dual cores were not tell the 2000s

 

[turkey3_scratch]

Yeah no kidding. Who knows what the cons are of bio over silicon. Skynet? :lol:

no one really knows but this point as far as we know 7nm is a limit for silicon but who knows whats next could be a .00001nm 100ghz .5watt 5000 core cpu in 20 years rendering what we have now crap
this is why i have a hard time investing a a powerful pc.....
i mean look at the cpus from 20 years ago only industrial computers have more than one core as far as i know dual cores were not tell the 2000s

That's impossible, that'd be smaller than an atom.

 

[Epicness937]

Yeah no kidding. Who knows what the cons are of bio over silicon. Skynet? :lol:

no one really knows but this point as far as we know 7nm is a limit for silicon but who knows whats next could be a .00001nm 100ghz .5watt 5000 core cpu in 20 years rendering what we have now crap
this is why i have a hard time investing a a powerful pc.....
i mean look at the cpus from 20 years ago only industrial computers have more than one core as far as i know dual cores were not tell the 2000s

That's impossible, that'd be smaller than an atom.

i agree
people in the 1800s thought it was impossible to have a pc
only god knows whats coming but my point is that basicly this technology is moveing so fast that we cant keep up

 

[TechyInAZ]

Well the problem is, we don't know what's smaller than an ATOM (basically how an atom is structured (i'm talking about the core, not the rings around each atom)). It won't be until we know that that what your saying is feasible.

 

[Epicness937]

Well the problem is, we don't know what's smaller than an ATOM (basically how an atom is structured (i'm talking about the core, not the rings around each atom)). It won't be until we know that that what your saying is feasible.

technology is scary......

 

[manleysteele]

Intel has already changed the Tick/Tock cadence to add another update at a given process size. That's what Kaby Lake is to Broadwell and Skylake. I know that most people already know this. That's just to catch up the few who don't know. My hope is that intel will spend enough time in the next few processes to fully exploit them. CPU clocks have been flat for quite a while. I don't know that I think they need to get to 5 GHz and above for what I do. I just want it. Waaagh. lol. There are several other shoes that haven't been dropped yet. For instance, when is the mainstream going to get an i7 with 6 cores/12 or more threads? When will we get more PCIe lanes on mainstream parts? Is PCIe 4.0 ever coming out? Why do the present chipsets add so few lanes to the piddling number on a CPU? Price? Why is the Broadwell graphics controller still superior to the Skylake and the projected Kaby Lake?

Here's a more serious question. If 3dxpoint can be made to work with the 1151 on Kaby Lake processor and Union Point (Z270) can the Skylake use it on the same chipset and, if so, why can't it be used on the Z170 chipset?

 

[Epicness937]

Intel has already changed the Tick/Tock cadence to add another update at a given process size. That's what Kaby Lake is to Broadwell and Skylake. I know that most people already know this. That's just to catch up the few who don't know. My hope is that intel will spend enough time in the next few processes to fully exploit them. CPU clocks have been flat for quite a while. I don't know that I think they need to get to 5 GHz and above for what I do. I just want it. Waaagh. lol. There are several other shoes that haven't been dropped yet. For instance, when is the mainstream going to get an i7 with 6 cores/12 or more threads? When will we get more PCIe lanes on mainstream parts? Is PCIe 4.0 ever coming out? Why do the present chipsets add so few lanes to the piddling number on a CPU? Price? Why is the Broadwell graphics controller still superior to the Skylake and the projected Kaby Lake?

Here's a more serious question. If 3dxpoint can be made to work with the 1151 on Kaby Lake processor and Union Point (Z270) can the Skylake use it on the same chipset and, if so, why can't it be used on the Z170 chipset?

i have a answer on why no 5ghz cpus ever heard of the fx 9590......
and yeah there are a lot of cpu questions that are unanswered

 

[8350rocks]

I believe Intel has already suggested that 10nm is the exhaustion point for pure silicon. That's why IBM along with others have pushed past to SiGe for their 7nm chips.

http://www.pcworld.com/article/2946124/ibm-reveals-worlds-first-working-7nm-processor.html

Just like ram is growing harder and harder to update along the current dram path and Intel and Micron are looking at something like HMC.

DDR4 is the current up to speed DDR path, GDDR will end up going toward HBM which already has a JEDEC standard. HMC does not have a standard yet AFAIK. AMD have talked about high end super computer APUs with HBM for CPU/GPU as part of the SoC.

Lots of interesting things ahead in memory...

 

[manleysteele]

Intel has already changed the Tick/Tock cadence to add another update at a given process size. That's what Kaby Lake is to Broadwell and Skylake. I know that most people already know this. That's just to catch up the few who don't know. My hope is that intel will spend enough time in the next few processes to fully exploit them. CPU clocks have been flat for quite a while. I don't know that I think they need to get to 5 GHz and above for what I do. I just want it. Waaagh. lol. There are several other shoes that haven't been dropped yet. For instance, when is the mainstream going to get an i7 with 6 cores/12 or more threads? When will we get more PCIe lanes on mainstream parts? Is PCIe 4.0 ever coming out? Why do the present chipsets add so few lanes to the piddling number on a CPU? Price? Why is the Broadwell graphics controller still superior to the Skylake and the projected Kaby Lake?

Here's a more serious question. If 3dxpoint can be made to work with the 1151 on Kaby Lake processor and Union Point (Z270) can the Skylake use it on the same chipset and, if so, why can't it be used on the Z170 chipset?

i have a answer on why no 5ghz cpus ever heard of the fx 9590......
and yeah there are a lot of cpu questions that are unanswered

I've heard of them.I just can't seem to find any information about them other than a spec sheet.