Kaby Lake information.

[Fjodor2001]

Hey, if you want to think Zen based APUs for mainstream PCs will have HBM2, then I am not going to try to stop you...

The trend is towards higher integration. The reason is obvious: It's cheaper to produce a single chip containing APU with HBM, than a separate APU/CPU + discrete GFX card with HBM. Especially since the iGPU part of the APU will be wasted silicon in the latter case, and the iGPU is taking up an ever larger part of the APU die area.

It's just a matter of when it'll happen. First gen Zen APUs, or perhaps second...? That we don't know yet.

 

[witeken]

Sure looks like it. Very nice, good to see Intel doing this.

You can also increase perf by increasing clock by 100MHz or claim the same with 1% IPC boost.

 

[ShintaiDK]

They can do variants both with and without HBM.

An APU with 4C/8T + HBM could be priced quite high and still be competitive, because it would compete with a separate CPU/APU + discrete GFX card which costs a lot.

So speculation and no fact? We can just as well dream that Intel will have HMC memory attached.

 

[AtenRa]

HBM is an expensive, high end technology and from the slides that have leaked the Zen based APUs are clearly aimed at mainstream price points, below Summit Ridge in terms of price/performance.

Summit Ridge may come close to $500-1K price for the 8-Core SKU, even if you have a 4GB HBM ZEN APU in 2017 at $300 it will still be below Summit Ridge.
We dont know the specs of the 2017 ZEN APUs yet so dont rush to dismiss the HBM yet.

HBM2 is a technology for high end dGPUs, FPGAs, etc. It's not for cheap APUs aimed at $500 laptops.

Nobody said anything about $500 Laptops, but it could be viable for 1K laptops, AIOs, workstations and servers.

 

[mikk]

I think what's more interesting is how KabyLake will compete with the 14 nm Zen based APUs with HBM in 2017.

It will be even more interesting to see how Zen based APUs will compete against Cannonlake 2017. I don't expect Zen APUs before H2 2017.

 

[tenks]

So this is the PCH that Skylake-E will be sharing with Kaby?

Looks kinda weak for something thats going to support the HEDT platform..

 

[Arachnotronic]

So this is the PCH that Skylake-E will be sharing with Kaby?

Looks kinda weak for something thats going to support the HEDT platform..

Better than X99

 

[beginner99]

Optane support is another word for NVME support in case anyone is confused.

No it's not. It's for support of future SSD products with Intel / Microns new 3D Xpoint non-volatile memory.

Therefore this is IMHO probably the biggest takeaway. This will blow any current SSDs out of the water.

 

[MarkLuvsCS]

Optane is the Intel + Micron next SSD NVM Express technology.

Source

I think it's clear Optane support is necessary to support 3D Xpoint in DIMM slots. Apparently although it's pin compatible with DDR4 the communication is going to be different. This is going to require an updated memory controller so Kabylake will likely be the first to support 3D Xpoint DIMM memory.

Fortunately for older systems 3D XPoint will probably make its way into SSDs via NVME. I would guess Skylake-E may add Optane support since it is coming out likely around the same time.

 

[ShintaiDK]

I would be surprised if the desktop would support Optane DIMMs. On the other hand the prospect is huge. And the same chipset is for the HEDT series. However I do think its just a PR note about NVME in reality.

 

[RussianSensation]

I am thinking there is a mistake in this slide. I think it should be 5K @ 60 fps for single display and 5K @ 30 fps for dual displays. How can you drive dual displays at 5K at twice the fps vs. a single display? Makes no sense to market it like this.

 

[Arachnotronic]

I am thinking there is a mistake in this slide. I think it should be 5K @ 60 fps for single display and 5K @ 30 fps for dual displays. How can you drive dual displays at 5K at twice the fps vs. a single display? Makes no sense to market it like this.

Yeah, it really looks like a typo. Believe it or not Intel slides can sometimes be full of 'em...

BTW, seems like a perfect fit for a future 5K iMac w/o dGPU inside...at least the Iris Pro model

 

[jpiniero]

I would be surprised if Intel made any CPU changes; doing GPU changes alone this quickly is quite a bit of work. Higher/tweaked clock speeds yes.

I would be surprised if the desktop would support Optane DIMMs. On the other hand the prospect is huge. And the same chipset is for the HEDT series. However I do think its just a PR note about NVME in reality.

I do expect Intel to support 3DXP DIMMs in the mainstream line; but Kabylake would be really be pushing it. I was thinking more Cannonlake.

 

[RussianSensation]

Yeah, it really looks like a typo. Believe it or not Intel slides can sometimes be full of 'em...

BTW, seems like a perfect fit for a future 5K iMac w/o dGPU inside...at least the Iris Pro model

Nah, it seems Apple fans think modern ARM CPUs will soon outperform all x86 chips. :sneaky:
http://daringfireball.net/2015/11/the_ipad_pro

I am surprised Intel isn't introducing USB 3.1 natively into the 200 chipset. Somehow I thought that would have been the case in 2016.

 

[tenks]

Nah, it seems Apple fans think modern ARM CPUs will soon outperform all x86 chips. :sneaky:
http://daringfireball.net/2015/11/the_ipad_pro

I am surprised Intel isn't introducing USB 3.1 natively into the 200 chipset. Somehow I thought that would have been the case in 2016.

Intel's native Thunderbolt 3 controller was the first to get to certified with USB 3.1. Intel will be pushing that.

 

[VirtualLarry]

Nah, it seems Apple fans think modern ARM CPUs will soon outperform all x86 chips. :sneaky:
http://daringfireball.net/2015/11/the_ipad_pro

I am surprised Intel isn't introducing USB 3.1 natively into the 200 chipset. Somehow I thought that would have been the case in 2016.

Why bother, when they can charge additional $$$ for their Alpine Ridge controller chip? That means, for mobos that implement 3.1, they can charge for the PCH, and Alpine Ridge.

 

[ShintaiDK]

Remember 2 versions of USB 3.1. And pretty much everyone can and does support G1. While G2 is the quite more rare support.

 

[Arachnotronic]

Why bother, when they can charge additional $$$ for their Alpine Ridge controller chip? That means, for mobos that implement 3.1, they can charge for the PCH, and Alpine Ridge.

Yeah, it seems that Intel knows how to run a business.

 

[Blue_Max]

200-chipset for next year already? I suppose it is only in preparation for the Skylake-10mn version/Cannonlake

That was my thought... seems like this refresh is somehow too early! Just a nice surprise, I guess... something to drive down prices and tempt my next upgrade.

 

[mikk]

http://www.sweclockers.com/nyhet/21...e-med-pakostade-grafikdelar-och-l4-cacheminne


There is a production Roadmap from ww46.

 

[AtenRa]

http://www.sweclockers.com/nyhet/21...e-med-pakostade-grafikdelar-och-l4-cacheminne


There is a production Roadmap from ww46.

Nice one thanks,

So, Broadwell-E is Q2 2016 after all.

 

[Arachnotronic]

http://www.sweclockers.com/nyhet/21...e-med-pakostade-grafikdelar-och-l4-cacheminne


There is a production Roadmap from ww46.

Dang, looks like Fjodor2001/AtenRa were right. Q2 it is for BDW-E. Hopefully early Q2 (i.e. April) rather than later...

 

[Sweepr]

Nice one thanks,

So, Broadwell-E is Q2 2016 after all.

Still not Q3 like you predicted. It was always late Q1 or Q2, no big surprise.

 

[AtenRa]

Still not Q3 like you predicted. It was always late Q1 or Q2, no big surprise.

I was always saying Q2-Q3 2016 depending on production start, and initially Broadwell-E was early Q1 2016 not late.

 

[Sweepr]

I was always saying Q2-Q3 2016 depending on production start, and initially Broadwell-E was early Q1 2016 not late.

There was an earlier leak by Bench-life mentioning Q2 so this isn't new.
Biggest surprise about Broadwell-E is that it brings 10C/20T at the top and a non-EE 8C/16T model, something we were only expecting with Skylake-E, which means they antecipated their Summit Ridge response in a big way.