Intel Skylake Platform

[DSC]

 

[Deleted member 13524]

What does the "2+2", "4+2" stand for?

Is it number of cpu cores and GPU performance rating? "x cores + GT y", as in "2 cpu cores + GT2"?

 

[Blazkowicz]

That's it, annoucing Broadwell GT3e on the LGA desktop, and Skylake GT4e.
Intel is effectively saying that the use of eDRAM die will get more common, though still high end.

 

[iMacmatician]

From CPU-World: "Intel Skylake desktop CPUs to launch in Q2 2015."

According to Intel's desktop roadmap, Skylake desktop Core i5 and i7 microprocessors, using Skylake-S core, along with 100 series chipset will be launched in the second quarter 2015. Broadwell desktop i5/i7 CPUs will be also released in Q2 2015. What is interesting is that all unlocked processors at launch will be Broadwell based, and all locked parts will be built on Skylake-S core.

Any reason for this strange split?

 

[pjbliverpool]

Still no new signs of AVX-512 in Skylake then? I guess my dream of a 2TF Skylake-E is pretty much dead then.

 

[eastmen]

From CPU-World: "Intel Skylake desktop CPUs to launch in Q2 2015."

Any reason for this strange split?

Dunno , maybe clocking is bad on the samples due to the redesigned gpu in skylake ??

I believe broadwell is getting a redesign and then another redesign is going into skylake.

 

[Thorburn]

Dunno , maybe clocking is bad on the samples due to the redesigned gpu in skylake ??

I believe broadwell is getting a redesign and then another redesign is going into skylake.

Can't imagine Broadwell would get a redesign at this stage, do you mean a respin?

 

[eastmen]

No , I mean the design of the gpu is different than haswell and skylake has even more changes to the gpu than broadwell does.

So if skylake may not clock as well due to the completely new gpu design

 

[Grall]

That's just unfounded speculation. Why would skylake have clocking problems and why would you think it to be the GPU's fault? There's no information whatsoever available to back such a claim up.

The GPU runs on a different PLL compared to the CPU cores anyway. They're not going to affect one another, and Intel probably knows by now how to design silicon within a certain clock envelope. It's pretty much a function of the laws of physics, after all.

 

[Raqia]

Still no new signs of AVX-512 in Skylake then? I guess my dream of a 2TF Skylake-E is pretty much dead then.

Here you go:

https://gcc.gnu.org/wiki/cauldron20...AVX-512_Vector_ISA_Kirill_Yukhin_20140711.pdf

AVX-512 (nice jump a la Xbox to Xbox 360) is shared with KNL. In addition to wider and more registers, "scatter" seems to be the big new instruction. There's some new masking instructions too.

 

[mczak]

Here you go:

https://gcc.gnu.org/wiki/cauldron20...AVX-512_Vector_ISA_Kirill_Yukhin_20140711.pdf

AVX-512 (nice jump a la Xbox to Xbox 360) is shared with KNL. In addition to wider and more registers, "scatter" seems to be the big new instruction. There's some new masking instructions too.

This is not "some new masking instructions" all instructions just take an additional (boolean per lane) mask reg which decides if the contents of a lane get updated. Just like GPUs work. Should make for very nice sw renderers (AVX2 could use masks with gathers only already, but they used ordinary simd reg for the mask not special boolean reg).

 

[Raqia]

This is not "some new masking instructions" all instructions just take an additional (boolean per lane) mask reg which decides if the contents of a lane get updated. Just like GPUs work. Should make for very nice sw renderers (AVX2 could use masks with gathers only already, but they used ordinary simd reg for the mask not special boolean reg).

Got it; thanks.

 

[pjbliverpool]

Thanks for the link, unfortunately no specific details in there (that I saw anyway) as to whether this will be coming to the desktop CPU's. At least now we know it will be in Xeons for sure though so I guess I could always pick up a server CPU for my desktop lol.

 

[Raqia]

Thanks for the link, unfortunately no specific details in there (that I saw anyway) as to whether this will be coming to the desktop CPU's. At least now we know it will be in Xeons for sure though so I guess I could always pick up a server CPU for my desktop lol.

I doubt the desktop core itself will be much different from the Xeons though Intel has been known to fuse off features lower tiers of products.

 

[CarstenS]

Skylake in Q1 2015 for developers, H2/2015 mass production and end of 2015 "launched":
http://www.pcgameshardware.de/IDF-I...Live-Demo-Benchmark-3DMark-Ice-Storm-1135431/

Oh - and Skylake's graphics will be DirectX12 feature level 12... officially, kind-of
http://www.pcgameshardware.de/Direc...rafik-erst-echte-DirectX-12-Hardware-1135515/

Sorry for the shameless self-ad, but I need the money.

 

[pjbliverpool]

I didn't realise that was your site. Great work! Aside from the fact I can't read German it's one of my favourite benchmarking sites.

 

[CarstenS]

Thanks - glad you like it. But it's not mine, I just work for them.

 

[fehu]

Too bad that we will be stuck with haswell for another year

 

[eastmen]

Too bad that we will be stuck with haswell for another year

Na , holiday 2015 is when I"m going to get myself a new surface so it should work out really well ! I just hope it doesn't get delayed like broadwell that would stink

 

[fellix]

More platform details for Skylake

• EIST modes will be extended to the system memory
• The new PCH will upgrade it's PCI-E connectivity from 8x2.0 to 20x3.0 lanes
• An upgrade to the LPC interface via the new eSPI standard
• The abandoned FIVR will be replaced with the latest mobile voltage specification IMVP8

An unconfirmed change is ditching of the EHCI standard, which means no more native support for the pre-USB 3.0 devices. In this case, users wishing to re-install Windows 7 or older from an USB drive would be out of luck (dust off that old DVD burner). USB inputs won't be recognized either, while PS/2 will still work.