Kaby Lake-E? Any news? Does this even make sense?

[VirtualLarry]

Just curious. Since current rumors essentially put SKL cores in KBL, but with an improved hardware-decoding block in the GPU, and the HEDT parts don't even have a GPU...?

Same cores as SKL, no improved GPU, would that even make any sense to produce, over SKL-E? (Personally, I don't think so.)

Since, then, therefore, the HEDT platform would skip a mainstream generation, would that mean that it would then "catch up" to mainstream? (SKL-E, and KBL, with CNL-E and CNL to follow-up?)

 

[Sweepr]

Cannonlake-EP comes after Skylake-EP.

 

[VirtualLarry]

Does that mean that Cannonlake and Cannonlake-E will release closer together?

 

[jpiniero]

It's possible although there would still be a sizable gap.

 

[mysticjbyrd]

Does that mean that Cannonlake and Cannonlake-E will release closer together?

I seriously doubt it. We don't even have broadwell-e out yet, and kabylake is scheduled to release this year too.

Broadwell-e

Skylake-e

Kabylake-e

Cannonlake-e ~ 2022

 

[zentan]

Cannonlake-e would likely come after skylake-e,there would be no separate Kabylake-e.

Does that mean that Cannonlake and Cannonlake-E will release closer together?

No,doest loot like it.
Skylake-E should come somewhere in 2018(if not delayed).
Cannonlake-e might be around 2020.Cannonlake mainstream parts would be somewhere between 2017 end and 2018.
All of these takes lots of assumptions,so yeah could turn out to be very incorrect.

 

[tenks]

There won't be a Kaby-E, just like there wasn't a Haswell refresh/Devil's Canyon-E.

The HEDT cores are a year+ behind anyway, so Kabylake wont effect that schedule.

Broadwell-E will out soon for the HEDT, followed by Kabylake on the DT later in the year. In 2017 HEDT will get refreshed w/ Skylake-E and coexist with KabyLake on the DT until its replaced bye Cannonlake. Then eventually Cannonlake-E will replace Skylake-E.


*breaths*

 

[ShintaiDK]

Intel said they are working on reducing the time in between client and server. However they will never come at the same time.

 

[Arachnotronic]

There probably won't be a Kaby Lake-E. Cannonlake-EP follows Skylake-EP.

 

[Edrick]

There won't be a Kaby-E, just like there wasn't a Haswell refresh/Devil's Canyon-E.

The HEDT cores are a year+ behind anyway, so Kabylake wont effect that schedule.

Broadwell-E will out soon for the HEDT, followed by Kabylake on the DT later in the year. In 2017 HEDT will get refreshed w/ Skylake-E and coexist with KabyLake on the DT until its replaced bye Cannonlake. Then eventually Cannonlake-E will replace Skylake-E.

I agree.

Now that Intel moved to their 3 step cadence for DT "Process-Architecture-Optimization", I think it is a safe bet that all the past optimizations found in the HEDT line will trickle down into the DT line, but will not require a 3rd step of their own. Meaning that Skylake-E (2017) should include many of the optimizations found in KabyLake DT (2016), non GPU of course.

 

[ElFenix]

There won't be a Kaby-E, just like there wasn't a Haswell refresh/Devil's Canyon-E.

The HEDT cores are a year+ behind anyway, so Kabylake wont effect that schedule.

Broadwell-E will out soon for the HEDT, followed by Kabylake on the DT later in the year. In 2017 HEDT will get refreshed w/ Skylake-E and coexist with KabyLake on the DT until its replaced bye Cannonlake. Then eventually Cannonlake-E will replace Skylake-E.


*breaths*

for the record, haswell refresh wasn't a new chip, just updated bins to take advantage of a more mature process (and some packaging changes for certain parts). intel had done that with sandy and ivy bridge as well, and there weren't sandy/ivy refresh 'e' parts either.

kaby, being a different chip and not just an SKL refresh, won't be getting an 'e' but for different reasons than why SB, IB, and HW refreshes didn't get an 'e' part.


seems to me intel is really starting to drag on 'e' part release windows. SB-E did not take near so long to come out after SB launched, and for nehalem the 'e' part was the first to launch.

 

[CentroX]

Kaby Lake is this year, and Cannonlake is 2017. Wouldn't expect Kaby Lake-E anytime soon.

 

[tenks]

for the record, haswell refresh wasn't a new chip, just updated bins to take advantage of a more mature process (and some packaging changes for certain parts). intel had done that with sandy and ivy bridge as well, and there weren't sandy/ivy refresh 'e' parts either.

kaby, being a different chip and not just an SKL refresh, won't be getting an 'e' but for different reasons than why SB, IB, and HW refreshes didn't get an 'e' part.


seems to me intel is really starting to drag on 'e' part release windows. SB-E did not take near so long to come out after SB launched, and for nehalem the 'e' part was the first to launch.

For the record, Kabylake isn't a new chip either. It's using skylake cores. Hence the comparions to Haswell refresh.
And Like Haswell refresh, the core wasn't changed significantly enough to warrant an EP/E version. Ie. They're not getting an E part for the SAME exact reasons. Thank you very much.



Also, Sandy and Ivy did not get the equivalent of a Devil's Canyon Haswell refresh. There were no Sandy or Ivy skus with better thermals and stock 4ghz clocks.


The point remains, adding a 3rd iteration to the DT line doesn't effect the HEDT schedule.


Thanks

 

[Arachnotronic]

For the record, Kabylake isn't a new chip either. It's using skylake cores. Hence the comparions to Haswell refresh.
And Like Haswell refresh, the core wasn't changed significantly enough to warrant an EP/E version. Ie. They're not getting an E part for the SAME exact reasons. Thank you very much.



Also, Sandy and Ivy did not get the equivalent of a Devil's Canyon Haswell refresh. There were no Sandy or Ivy skus with better thermals and stock 4ghz clocks.


The point remains, adding a 3rd iteration to the DT line doesn't effect the HEDT schedule.


Thanks

Actually, it's not clear if the SKL cores have been "tweaked" in KBL. There will also be changes to the "uncore" (media encode/decode, probably a better memory controller, etc.). Kaby Lake is certainly a different beast from Devil's Canyon.

 

[tenks]

Actually, it's not clear if the SKL cores have been "tweaked" in KBL. There will also be changes to the "uncore" (media encode/decode, probably a better memory controller, etc.). Kaby Lake is certainly a different beast from Devil's Canyon.

While I agree with that 100%, the point remains the same from my previous post. The CPU/core inside Kabylake isn't different enough from a performance perspective to create a new enterprise class chip from. Especially since the EP/E lines have such a long lead time, with all the extra stress testing (I forget what thats called?), there is no way Intel could maneuver that quickly.

 

[Arachnotronic]

While I agree with that 100%, the point remains the same from my previous post. The CPU/core inside Kabylake isn't different enough from a performance perspective to create a new enterprise class chip from. Especially since the EP/E lines have such a long lead time, with all the extra stress testing (I forget what thats called?), there is no way Intel could maneuver that quickly.

Oh, I agree. Especially with the apparent bifurcation of the client/server cores at this point. The term you're looking for is validation

 

[tenks]

Oh, I agree. Especially with the apparent bifurcation of the client/server cores at this point. The term you're looking for is validation

Thank you, I couldn't spit it out for some reason, Validation!

 

[stockwiz]

No reason to upgrade the sandy bridge for another 5-10 years at this rate.

 

[CHADBOGA]

No reason to upgrade the sandy bridge for another 5-10 years at this rate.

I've said before in a somewhat serious manner, I think there is a good chance my i5-3570K @4.0Ghz will last me 10 years.

It will soon be 5 years old(I think) and it still seems fast as hell.

 

[VirtualLarry]

I've said before in a somewhat serious manner, I think there is a good chance my i5-3570K @4.0Ghz will last me 10 years.

It will soon be 5 years old(I think) and it still seems fast as hell.

3570K @ 4.0 is a good, solid, chip. I built a rig for a (now late) friend with that CPU, and overclocked it to the same amount. He wanted an overclocked rig, so I thought that would be an easy and mild overclock, suitable for day-to-day usage without the risk of instability.

 

[CHADBOGA]

3570K @ 4.0 is a good, solid, chip. I built a rig for a (now late) friend with that CPU, and overclocked it to the same amount. He wanted an overclocked rig, so I thought that would be an easy and mild overclock, suitable for day-to-day usage without the risk of instability.

Yep, my thinking exactly.

4 cores @ 4 Ghz.