Intel Skylake / Kaby Lake

[zentan]

z170m OC formula review at tweaktown. Seemingly a very good mATX board for enthusiasts who are ok with 2 DIMMs limitation. Memory overclocks is excellent,cpu overclock seems nice as well.

I wasn't surprised when the Z170M OC Formula hit 5GHz; it's an overclocking motherboard, so 5GHz is to be expected.
Stability testing worked well, but it's obvious that the LLC is very aggressive as the VCore is slightly higher than what I chose in the UEFI and it almost throttled on a few cores

The Corsair Vengeance 4GHz kit is the fastest kit I have on hand, and most Z170 motherboards fail to boot with this kit at XMP. However, the Z170M OC Formula handled this kit like a champ; training was extremely minimal. I expect this kit to work well, as many memory vendors have told me their 4.33GHz kits and faster kits work flawlessly with this motherboard.

With its two DIMMs, the Z170M OC Formula blows away most other motherboards when it comes to ease of memory overclocking. Every kit I threw at it, including the Corsair 4000MHz Vengeance, booted the first time with only XMP. It is very rare that a motherboard handles memory so easily, and the Z170M OC Formula is perfect for high-speed memory excursions. We have also been told by more than one memory vendor, that this motherboard is pretty much a staple for reaching speeds over 4.33GHz when using XMP

Most motherboards can pull off close to 4GHz without trouble, but going above 4GHz is where things get tough with XMP and even manual tuning, and it is where the Z170M OC Formula shines. In an era where most motherboards push the CPU and cache clocks to the same degree, memory is what sets overclocking motherboards apart, and the Z170M OC Formula does exceedingly well in that regard

http://www.tweaktown.com/reviews/77...mula-intel-z170-motherboard-review/index.html

 

[Sweepr]

Interesting read:

Skylake Xeon E3s Serve Up Cheap Flops

While the performance of the CPU cores on the Xeon E3-1500 v5 processors is interesting for single-threaded applications, these four cores with eight threads do not provide all that much oomph. Was is intriguing is the fact that the new Iris Pro P580 graphics engine on the chip has 72 graphics cores, up from 48 graphics cores with the P555 engine used in the Broadwell Xeon E3 v4 hybrid chips. We asked Intel for the floating point ratings for the GPUs inside the Skylake-H chips, and it turns out the Iris Pro P580 delivers only 331.2 gigaflops at double precision running at its Turbo Boost speed of 1.15 GHz, but the single precision performance is a very respectable 1.32 teraflops. Yes, that is about 20 percent of the expected single precision performance of the “Knights Landing” Xeon Phi parallel processor, which has 72 active “Silvermont” Atom cores with a pair of 512-bit AVX 3.0 vector engines on each core. The Knights Landing chip is expected to deliver more than 3 teraflops at double precision and twice that at single precision, so it packs a punch to be sure. But for workloads that do not require low latency interconnect and that can fit their data and applications in the relatively small memory footprint of the Xeon E3 chip, this is clearly an option.

This GPU in the Skylake-H Xeon chip offers very compelling bang for the buck for single precision floating point math.

www.nextplatform.com/2016/06/14/skylake-xeon-e3s-serve-cheap-flops/

 

[witeken]

This has got to be definitely the most crazy thing I learn about Intel this year:

Another way to make a chip faster is to add special circuits that only do one thing, but do it extremely quickly. Roughly 25 percent of the E5’s circuits are specialized for, among other tasks, compressing video and encrypting data. There are other special circuits on the E5, but Intel can’t talk about those because they’re created for its largest customers, the so-called Super 7: Google, Amazon, Facebook, Microsoft, Baidu, Alibaba, and Tencent. Those companies buy—and often assemble for themselves—Xeon-powered servers by the hundreds of thousands. If you buy an off-the-shelf Xeon server from Dell or HP, the Xeon inside will contain technology that’s off-limits to you. “We’ll integrate [a cloud customer’s] unique feature into the product, as long as it doesn’t make the die so much bigger that it becomes a cost burden for everyone else,” says Bryant. “When we ship it to Customer A, he’ll see it. Customer B has no idea that feature is there.”

So... the "custom" chips that Intel sends to its biggests cloud providers, are in fact the same die as all the other Xeon, but with the secret feature enabled!D:

http://www.bloomberg.com/news/articles/2016-06-09/how-intel-makes-a-chip

 

[DrMrLordX]

That's odd.

 

[Arachnotronic]

This has got to be definitely the most crazy thing I learn about Intel this year:



So... the "custom" chips that Intel sends to its biggests cloud providers, are in fact the same die as all the other Xeon, but with the secret feature enabled!D:

http://www.bloomberg.com/news/articles/2016-06-09/how-intel-makes-a-chip

Nothing new. In fact, there are special instructions at the CPU level that only certain customers know about from what I have heard.

 

[Burpo]

Nothing new indeed.. Xeons have always been better at compression & encryption. Running Passmark tests side by side between Gulftown & Westmere showed me that. Xeons consistently got higher scores in those tests.

 

[nerp]

That's odd.

It's the most cost effective way to do it.

 

[VirtualLarry]

Nothing new. In fact, there are special instructions at the CPU level that only certain customers know about from what I have heard.

Is it just that the opcode format is undocumented (in which case, "fuzzing" could potentially find it), or are those customers delivered custom platform microcode that actually "unlocks" those opcodes for the customer?

 

[mikk]

iKBLULTGT1 = "Intel(R) HD Graphics 610"
iKBLULTGT2 = "Intel(R) HD Graphics 620"
iKBLULTGT2F = "Intel(R) HD Graphics 620"
iKBLULTGT3 = "Intel(R) HD Graphics 635"
iKBLULTGT3E15 = "Intel(R) Iris(TM) Graphics 640"
iKBLULTGT3E28 = "Intel(R) Iris(TM) Graphics 650"
iKBLULXGT2 = "Intel(R) HD Graphics 615"
iKBLDTGT1 = "Intel(R) HD Graphics 610"
iKBLDTGT2 = "Intel(R) HD Graphics 630"
iKBLHALOGT1 = "Intel(R) HD Graphics 610"
iKBLHALOGT2 = "Intel(R) HD Graphics 630"
iKBLSRVGT2 = "Intel(R) HD Graphics P630"
iKBLWGT2 = "Intel(R) HD Graphics P630"

http://forums.laptopvideo2go.com/to...2120164471-adds-broxton-and-kabylake-support/

 

[DrMrLordX]

It's the most cost effective way to do it.

It might be cost-effective, but having entire feature sets on your CPU that are undocumented (effectively making them hidden) is a bit off-putting, especially when conspiracy wonks taken far-too-seriously by entire foreign governments will conclude that such features might be backdoors of some kind for somebody.

If the stuff is there, document it. People ought to know what it is they're buying, even if the information is in a format that is beyond the comprehension of 99.9%+ of people anyway.

 

[mikk]

The memo referenced two other products – Coffee Lake and Glenview – that Intel hasn't discussed publicly.

http://www.oregonlive.com/silicon-forest/index.ssf/2016/04/intels_new_president_launches.html


Glenview must be a product with graphics because there are entries in 21.20.16.4471. Geminilake as well.

 

[jpiniero]

I still want to see them make a Core phone, even if it would be too pricy for my liking.

 

[Burpo]

Interesting..

Intel HD Graphics Driver 21.20.16.4471 adds Broxton and Kabylake support.

"Haswell and Broadwell support has been completely removed from this driver."

http://www.geeks3d.com/forums/index.php?topic=4597.0

 

[nerp]

It might be cost-effective, but having entire feature sets on your CPU that are undocumented (effectively making them hidden) is a bit off-putting, especially when conspiracy wonks taken far-too-seriously by entire foreign governments will conclude that such features might be backdoors of some kind for somebody.

If the stuff is there, document it. People ought to know what it is they're buying, even if the information is in a format that is beyond the comprehension of 99.9%+ of people anyway.

Some of it might be proprietary and protected corporate trade secrets, so it makes sense that Intel isn't just opening the door to something it might have co-developed with an organization and shared in the R&D.

 

[DrMrLordX]

"Haswell and Broadwell support has been completely removed from this driver."

Huh, interesting. I guess Haswell and Broadwell users are stuck on old drivers ad infinitum? Or are they going to maintain a separate driver development for Gen9 and earlier?

Some of it might be proprietary and protected corporate trade secrets, so it makes sense that Intel isn't just opening the door to something it might have co-developed with an organization and shared in the R&D.

Everything in a Xeon (well, nearly everything) is proprietary. Most companies can't even legally sell a modern x86 CPU. What's odd is that, by your argument, they make more money stopping people from being able to use some part of the CPU baked in at the request of one of Intel's largest customers. You would think that if Xeons had Whizbang Feature X in there, that they could use it as a bullet point and charge accordingly. Google (or whoever else is asking for this stuff to be baked into Xeon designs) must be paying Intel a lot of money to make sure that nobody else has the ability to make use of these undocumented features.

 

[Sweepr]

http://forums.laptopvideo2go.com/to...2120164471-adds-broxton-and-kabylake-support/

Thanks for sharing. Interesting to see Broxton support still there.

 

[zlejedi]

Well they can also increase yields that way. If defect is in one of areas reserved for special customer than cpu still can be sold to everyone else.

 

[dark zero]

Interesting..

Intel HD Graphics Driver 21.20.16.4471 adds Broxton and Kabylake support.

"Haswell and Broadwell support has been completely removed from this driver."

http://www.geeks3d.com/forums/index.php?topic=4597.0

So Broxton is still alive?

Also... Haswell is EOL so no surprise... but Broadwell? What the heck Intel? Going to NVIDIA route too?

 

[mikk]

So Broxton is still alive?

Yes, it's called Apollo Lake. Internally Intel calls it Broxton.

 

[Rngwn]

http://forums.laptopvideo2go.com/to...2120164471-adds-broxton-and-kabylake-support/

Looks like KBL 4+4e is not going to happen in a foreseeable future. *heavily pouting*

 

[Sweepr]

Kaby Lake GPU learns hardware HDR

The new tricks that Intel managed to squeeze in to Kaby Lake are High Dynamic Range (HDR), Wide Color Gamut (Rec.2020) and HDCP 2.2 support. Kaby Lake can now use hardware encode and decode for VP8, HEVC 8b, VP9 and HEVC 10b. Skylake used software encoding and decoding to cope with VP9 and HEVC 10b.

Kaby Lake GPU added support for HEVC10-bit decode and VP9 10-bit decode and just like Skylake, it can also support HEVC 8-bit encode/decode and VP8 8-bit encode/decode.

These new codecs should make Netflix UHD playback possible on Intel's platform. It should also enable 4K@60Hz YouTube content as well as Google Hangout videos. The Chrome browser can also benefit from VP9 hardware acceleration support.

www.fudzilla.com/news/graphics/40959-kaby-lake-gpu-learns-hardware-hdr

 

[AtenRa]

So desktop Kaby retail availability could start from 12th of December 2016 (50th week) until late January 2017.

 

[Sweepr]

Fudzilla speculates that Kaby Lake-X (4C/8T) is LGA 2066, same as Skylake-X ('Skylake-E'):

There will be a Kaby Lake-X 4 core processor with 95+ W TDP using the same LGA 2066 R4 socket ('R4 socket') . Both Skylake-X and Kaby Lake-X support the new chipset that is known as Kaby Lake-X.

www.fudzilla.com/news/processors/40949-skylake-x-10-comes-in-q2-17

If true, they could be pushing some insane stock clocks here (>95W TDP).

Core i7-6770HQ Cinebench R15 OpenGL Result (DDR4-2800)

Almost twice typical HD 530 (45W TDP Skylake-H) performance.

 

[jpiniero]

Fudzilla speculates that Kaby Lake-X (4C/8T) is LGA 2066, same as Skylake-X ('Skylake-E'):

Won't say it won't happen, but that would only make sense if:

1. There won't be an unlocked Kaby Lake-S model
2. The Split is actually happening

I imagine it would still be the same 4+2 KBL die as the other models but maybe they could jack up the clocks to 4.4 if validation allows it even if they have to make the TDP 140. Probably wouldn't get access to the IGP though.

 

[badb0y]

Is it worth it to upgrade from 2600k at 4.7 Ghz to Skylake if all I do is play games.