Intel Cannonlake SoC will have 4-core, 6-core and 8-core versions

[Nothingness]

The code jpniero posted means that Cannonlake has all features of server Skylake, which in turn has all features of client Skylake. It should be noted the split beteween client ans server Skylake was done in the same patch that added Cannonlake, so there might be a future split for Cannonlake.

 

[ShintaiDK]

It could simply also be that all Cannonlake cores are the same, unlike Skylake. And the AVX512 part gets disabled if so in client. But I do favour the split to come later theory.

 

[DrMrLordX]

The code jpniero posted means that Cannonlake has all features of server Skylake, which in turn has all features of client Skylake. It should be noted the split beteween client ans server Skylake was done in the same patch that added Cannonlake, so there might be a future split for Cannonlake.

Right, it's just that some might not know that fallthrough causes all the other cases to execute, if the topmost case is true. So if the CPU is identified as CK_Cannonlake, then all the statements under the CK_SkylakeServer and CK_SkylakeClient area also executed.

If CK_SkylakeServer is true, then it and all the code from the CK_SkylakeClient execute, etc.

 

[jpiniero]

It should be noted the split beteween client ans server Skylake was done in the same patch that added Cannonlake, so there might be a future split for Cannonlake.

Actually it's been in there since 2014, they just renamed stuff/re-arranged things.

I do believe in the core split. However Skylake was in such upheaval that it is possible that the mainstream/server split was a one time thing.

 

[Arachnotronic]

The Linux code sure makes it look like it has all of AVX-512 plus some new ones. Plus SHA which might be useful to some people. Of course it might be disabled or some other method on mainstream parts. Also interesting that while they split Skylake Client and Skylake Server they did not do that for Cannonlake.

Nice, CNL will get SHA acceleration in hardware.

 

[Sweepr]

Nice, CNL will get SHA acceleration in hardware.

Found this:

This is a patch recently made ​​the Clang compiler LLVM project which unveiled the new instruction sets to be supported by the architecture Cannonlake . We discover that the CPU will support instruction sets AVX512IFMA and AVX512VBMI (in addition to AVX-512 extensions already built into Skylake processors) and seven new instructions SHA-1 and SHA-256.

...It remains to what extent these new instructions accelerate processing algorithms SHA-1 and SHA-256, two of the most used in cryptography algorithms. According to our information, according to preliminary tests, the implementation of the SHA-256 algorithm with the new instructions would be up to 3.6 times faster than the conventional one in SSSE3 ...

 

[tweakboy]

Reminds me of Corona Lake.. gl

 

[Sweepr]

Intel 2016 Mobile Roadmap

https://benchlife.info/intel-14nm-kaby-lake-and-10nm-cannon-lake-still-on-schedule-03152016

First taste of Intel's 10nm: Cannonlake-Y/U launching in Q3-2017 (August/September?).

 

[moonbogg]

So they are giving mainstream more cores? Or is that not the case...

 

[Magic Carpet]

So they are giving mainstream more cores? Or is that not the case...

I'd like to find out that as well. Just sold my Haswell rig for good $.

 

[NTMBK]

Intel 2016 Mobile Roadmap

https://benchlife.info/intel-14nm-kaby-lake-and-10nm-cannon-lake-still-on-schedule-03152016

First taste of Intel's 10nm: Cannonlake-Y/U launching in Q3-2017 (August/September?).

Full 3 years between 14nm launch and 10nm launch.

 

[Sweepr]

Full 3 years between 14nm launch and 10nm launch.

To be fair Broadwell-U only showed up after CES 2015 and Core-M had very little presence in 2014.

So they are giving mainstream more cores? Or is that not the case...

Only rumours by now, but I'm sure Bench-life will answer this question sometime this year.
Considering they have 25W 4C/8T (Skylake-H) and 45-65W 16C/32T (Broadwell-DE, 'Xeon D') variants and very small dies even with iGPUs I think they're ready to make the move at 10nm.

 

[NTMBK]

To be fair Broadwell-U only showed up after CES 2015 and Core-M had very little presence in 2014.

And do we have any evidence that the same won't happen with Cannonlake?

 

[Sweepr]

And do we have any evidence that the same won't happen with Cannonlake?

Time will tell, but you can't say 'full 3 years' right now.
H2-2017 was narrowed down to Q3 for Cannonlake-Y/U, which is positive news (could be Q4/late 2017).

 

[mikk]

Cannonlake-U in Q3 2017 would be nice or even in Q4 2017, although it is too early to tell. This is Intels schedule currently but we should be aware from almost all prior generations that such schedule isn't set in stone.

I don't think there is a core increase for CNL-Y/U, I hope that 40 EUs for GT2 is real. H265 hybrid encoding requires much more than 24 EUs with Skylake.

 

[moonbogg]

3 years? Jesus. I'll have evaporated into Hawking radiation by then.

 

[Arachnotronic]

Full 3 years between 14nm launch and 10nm launch.

Lol @ Broxton in late 2016/2017. What a joke the Atom team is.

 

[zentan]

Yeah,Cannonlake-U would probably be the dual cores with better clocks may be.4 cores might only come to -H and higher.Unless Intel's 10nm is something radically more efficient than it's 14nm -U would remain dual cores only.Probably quad core i5s could come to somewhat lower ranged market than it is now.Theoretically they could make quads within 15W envelope but with lower clocks/turbos that 2c/4t parts.Don't think intel would like a situation where a dual core in the same envelope is quite better in ST but losing in MT to the quad.

 

[Sweepr]

I don't think there is a core increase for CNL-Y/U, I hope that 40 EUs for GT2 is real. H265 hybrid encoding requires much more than 24 EUs with Skylake.

Too bad. Even with 40 EUs (beefed up, Gen 10), 2C+GT2 Cannonlake should be tiny. 28W Skylake-U (soon Kabylake-U) SKUs already operate at mid-3GHz, not sure if there's a lot of room for improvement here. AMD could take advantage of this by releasing <35W 4C/8T Raven Ridge APUs. Still, if they go this route, I expect many 20-35W quad-core 'H' SKUs and at least some 6C/12T models for larger laptops (45-55W TDP) - these could end up in desktops as well.

 

[Sweepr]

Our mainstream chip with >4C/8T is not Cannonlake, it's Coffee Lake.

 

[DrMrLordX]

It'll be interesting to see how/if they deliver on that. Still waiting on desktop Skylake GT3e/GT4e.