Intel Skylake / Kaby Lake

[ninaholic37]

Ahhh yes I wasn't sure. Thanks for clarifying

I remember going on a skiing trip with my dad in 1991, and we brought along his black-and-white 286 laptop, and I was able to play Winter Challenge for about 45 minutes while he was driving, and it didn't even need to be plugged in! Ever since I always thought laptops were awesome and preferred them over desktops.

Some new ones do feel cheap so I buy used business laptops for <$100 instead.

 

[Dufus]

Wondered if I softened the phrasing too much to adequately describe my utter contempt for the pathetic excuse of a computer that laptops have been/are/will be. Need to work on being more direct with my opinions.

Warning: Enthusiast Rant starts now.

To minimize the limitations, being the antithesis of an enthusiast computer, costs multiple thousands - even the uber lappys still suffer from a terrible thermal envelope, pathetic screens and Playschool style keyboards. And none that exceed the broken format barrier. The only desktop replacement is a better, faster desktop.

The laptop is a betrayal of the Personal Computer, a once noble device that could be tailored to one's budget, needs and tastes - assembled and serviced with one's own hands from carefully selected components.

The laptop is the icon for bland look-alike computing appliances - trading function for portability - performance for form factor - excellence for mass marketing drivel - the laptop is a celebration of conformist mediocrity, all at a higher price. That sort of banal treachery can make a corporation rich.

/rant

Such is the cost of portability.

 

[RussianSensation]

There's some real lack of info on how various motherboards function with regards to PCIe lanes. It's not something I really thought about until I was looking at x99 boards and saw the Asus x99 pro usb 3.1 mention something about sharing bandwidth. I know my next build will use a pcie ssd in addition to my sli 970s and I do want to have access to usb 3.1 without sacrificing any bandwidth for my pcie devices.

There are X99 boards that do not have this limitation. As I already noted, it's possible to run 8x/8x/8x + 4x on a 5820K on some X99 boards which means it will be a peace of cake to run 16x/8x + PCIe 3.0 x4 and have enough for dual USB 3.1s. That's not a reason to discount the X99 platform. I think the bigger factors are what a person does with his/her computer, overclocking and power usage. If the rig is being built primarily for gaming, Skylake is a far more power efficient system and it will win in most games in the short term since most games do not use > 4 cores. However, if someone does things with their PC other than just gaming, intends to not upgrade their CPU for 4-5 years, then there is no question that a 5820K @ 4.5-4.6Ghz is the better longer term productivity+gaming setup. When it comes to productivity like rendering, video editing and so on, i7-6700K isn't even on the map against a 6-core X99 processor. It'll take a max overclocked 6700K just to match a sub-4Ghz 5800/5900 series.

There are some unique advantages of Z170 though such as the ability to RAID 2-3 PCIe SSDs. I'll have to look up if this is possible on X99 but I would imagine Z170 is going to be by far superior in this regard.

 

[steve wilson]

When will faster DDR 4 kits come out? It seems like this is what Skylake really needs to shine.

 

[cmdrdredd]

When will faster DDR 4 kits come out? It seems like this is what Skylake really needs to shine.

How much faster? G.Skill has 3600Mhz kits, but they are outrageously expensive at $400 for 8GB (2x4GB)

There are X99 boards that do not have this limitation. As I already noted, it's possible to run 8x/8x/8x + 4x on a 5820K on some X99 boards which means it will be a peace of cake to run 16x/8x + PCIe 3.0 x4 and have enough for dual USB 3.1s. That's not a reason to discount the X99 platform. I think the bigger factors are what a person does with his/her computer, overclocking and power usage. If the rig is being built primarily for gaming, Skylake is a far more power efficient system and it will win in most games in the short term since most games do not use > 4 cores. However, if someone does things with their PC other than just gaming, intends to not upgrade their CPU for 4-5 years, then there is no question that a 5820K @ 4.5-4.6Ghz is the better longer term productivity+gaming setup. When it comes to productivity like rendering, video editing and so on, i7-6700K isn't even on the map against a 6-core X99 processor. It'll take a max overclocked 6700K just to match a sub-4Ghz 5800/5900 series.


There are some unique advantages of Z170 though such as the ability to RAID 2-3 PCIe SSDs. I'll have to look up if this is possible on X99 but I would imagine Z170 is going to be by far superior in this regard.

Which is why I said there was some lack of info here because I didn't even know the board I was looking at shared USB3.1 bandwidth with PCIe slots until I scrolled down the Asus specifications page and saw the * next to some options. Even searching the internet for the board and trying to see what configurations were possible without sacrificing PCIe speed turned up a blank. I suppose that the number of people running SLI + PCIe SSD + USB 3.1 or using the built in Wifi/Bluetooth is small.

 

[Sweepr]

Did someone post PCLab's review yet?

Lots of good stuff. They did test some of the latest games at 1080p Ultra settings and even compared overclocked Sandy Bridge vs Ivy Bridge vs Haswell vs Skylake vs Vishera.

DDR3-2133 CL10 vs DDR4-2666 CL16 so Skylake-S was still at a disadvantage (next to older platforms), but better than DDR4-2133 crap.

Battlefield 4 Multiplayer

Core i5 6600K 17.1% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo)

Crysis 3 Welcome to the Jungle

Real star of the show is the Core i7 6700K here, 7.6% faster than Core i7 4790K despite 5-10% lower turbo clocks.

Far Cry 4

Core i5 6600K 23.6% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo) and almost as fast as Core i7 4790K (4.0-4.4GHz base/turbo + 8 threads).
Core i7 6700K 22.45% faster than Core i7 4790K despite 5-10% lower turbo clocks.

GTA V

Core i5 6600K 11.85% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo), faster than previous Core i7 4770K too. Core i5 6600K 88% faster than FX8350.
Core i7 6700K 2.15x as fast as FX8350.

The Witcher 3

Core i5 6600K 10.1% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo), on par with last generation Core i7 4770K.
Core i7 6700K 13% faster than Core i7 4790K despite 5-10% lower turbo clocks.
Core i7 6700K 2.06x as fast as FX8350.

Project Cars (Rain)

Core i5 6600K 9.1% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo), on par with last generation Core i7 4790K.
Core i7 6700K 11.1% faster than Core i7 4790K despite 5-10% lower turbo clocks.

Total War: Attila

Core i5 6600K 21.3% faster than Core i5 4690K (same 3.5-3.9GHz base/turbo), faster than Core i7 4790K (4.0-4.4GHz base/turbo + 8 threads).
Core i7 6700K 16.3% faster than Core i7 4790K despite 5-10% lower turbo clocks.

When it comes to CPU intensive games nothing touches Skylake-S right now. I wish they had included Haswell-E too.
We can also examine Intel's IPC advantage. Both Core i7 6700K and FX8350 operate at 4.0-4.2GHz. In some games Skylake-S is more than twice as fast as Vishera.

 

[Sweepr]

Here's a quick clock per clock (IPC) analysis, PCLab has Haswell and Skylake-S at equal clocks in their overclocking results.

Core i5 6600K vs Core i5 4690K at fixed 4.5GHz, Skylake-S is:
16.2% faster @ Battlefield 4 MP
10.9% faster @ Counter Strife Global Offensive
6% faster @ Crysis 3
22.7% faster @ Far Cry 4
13% faster @ GTA V
12.5% faster @ The Witcher 3
9.2% faster @ Watch Dogs
14.7% faster @ Project Cars
14.8% faster @ Starcraft 2
24.5% faster @ Total War Attila

Who said Skylake-S was slower for games again? You have some explaining to do AnandTech.

Edit: Overall gaming performance per clock (IPC) from 14 games/tests @ 1080p Ultra:

Sandy Bridge to Haswell: 13.5%
Haswell to Skylake: 12.8%
Skylake beating Broadwell with eDRAM here.

 

[Blue_Max]

Did someone post PCLab's review yet?

Wow.

Those are some *stunning* gains and it sure makes me finally want to upgrade this (what used to be) perfectly-good 2600k.

Daaaang... that really is impressive. Thanks for sharing the link.

 

[IndyColtsFan]

Wow.

Those are some *stunning* gains and it sure makes me finally want to upgrade this (what used to be) perfectly-good 2600k.

Daaaang... that really is impressive. Thanks for sharing the link.

Yeah, I think this is pushing me towards doing an upgrade to the 6700K rather than waiting longer. My current 2600K system will be 4 years old next month and is the longest I've ever used a single system as my main system.

 

[Nothingness]

Here's a quick clock per clock (IPC) analysis, PCLab has Haswell and Skylake-S at equal clocks in their overclocking results.

Core i5 6600K vs Core i5 4690K at fixed 4.5GHz, Skylake-S is:
16.2% faster @ Battlefield 4 MP
10.9% faster @ Counter Strife Global Offensive
6% faster @ Crysis 3
22.7% faster @ Far Cry 4
13% faster @ GTA V
12.5% faster @ The Witcher 3
9.2% faster @ Watch Dogs
14.7% faster @ Project Cars
14.8% faster @ Starcraft 2
24.5% faster @ Total War Attila

Who said Skylake-S was slower for games again? You have some explaining to do AnandTech.

It looks like you missed Assassin's Creed Unity (2% faster) and Arma III (+10.7%) results

Another thing to note is that the gain for 6700K vs 4790K @4.7 GHz seems to be less, but still very nice.

For the record the RAMs used are: DD3-1866 9-9-10-24 1N and DD4-2666 16-17-17-36 2N. So DDR4 has worse latency (~20%), but higher bandwidth (~40%).

 

[Sweepr]

It looks like you missed Assassin's Creed Unity (2% faster) and Arma III (+10.7%) results

Overall results added (from all 14 tests). About the same clock per clock gains from Haswell to Skylake as Sandy Bridge to Haswell (~13%).

For the record the RAMs used are: DD3-1866 9-9-10-24 1N and DD4-2666 16-17-17-36 2N. So DDR4 has worse latency (~20%), but higher bandwidth (~40%).

Using AnandTech's method:

Normally in our DRAM reviews I refer to the performance index, which has a similar effect in gauging general performance:

DDR3-1600 C11: 1600/11 = 145.5
DDR4-2133 C15: 2133/15 = 142.2

As you have faster memory, you get a bigger number, and if you reduce the CL, we get a bigger number also. Thus for comparing memory kits, if the difference > 10, then the kit with the biggest performance index tends to win out, though for similar kits the one with the highest frequency is preferred.

DDR4-2666/16 = 166
DDR3-1866/9 = 207

 

[Nothingness]

Using AnandTech's method:

DDR4-2666/16 = 166
DDR3-1866/9 = 207

I prefer to use real time latency as Intel does:
That gives:
DDR4 16/(2666/2)*1000 = 12 ns
DDR3 9/(1866/2)*1000 = 9.6 ns
So about 25% more latency for DDR4, while it has 2666/1866 ~= 1.42 so 40% more bandwidth. Can't wait to more complete studies which play on both latency and frequency of RAM for both DDR3 and DDR4

 

[CHADBOGA]

Good to see Skylake looking a lot better with faster ram, it bodes well for the next year or two as faster and cheaper ram comes out.

 

[ShintaiDK]

It would be nice to see where the peak relative benefit is for Skylake in terms of memory. What CAS and what speed. It seems speedwise its on the other side of 3000.

But Skylake really brought back the focus on memory.

 

[SAAA]

Well... first impression of Skylake was really bad, but waiting pays off definitely.

Here's a quick clock per clock (IPC) analysis, PCLab has Haswell and Skylake-S at equal clocks in their overclocking results.

Core i5 6600K vs Core i5 4690K at fixed 4.5GHz, Skylake-S is:
16.2% faster @ Battlefield 4 MP
10.9% faster @ Counter Strife Global Offensive
6% faster @ Crysis 3
22.7% faster @ Far Cry 4
13% faster @ GTA V
12.5% faster @ The Witcher 3
9.2% faster @ Watch Dogs
14.7% faster @ Project Cars
14.8% faster @ Starcraft 2
24.5% faster @ Total War Attila

Who said Skylake-S was slower for games again? You have some explaining to do AnandTech.

Edit: Overall gaming performance per clock (IPC) from 14 games/tests @ 1080p Ultra:

Sandy Bridge to Haswell: 13.5%
Haswell to Skylake: 12.8%
Skylake beating Broadwell with eDRAM here.

I prefer to use real time latency as Intel does:
That gives:
DDR4 16/(2666/2)*1000 = 12 ns
DDR3 9/(1866/2)*1000 = 9.6 ns
So about 25% more latency for DDR4, while it has 2666/1866 ~= 1.42 so 40% more bandwidth. Can't wait to more complete studies which play on both latency and frequency of RAM for both DDR3 and DDR4

Those numbers and the Anandtech review ones suggest that Skylake really loves bandwidth... we need a serious comparison with 6-cores, various DDR4s and mostly gaming focus.

Are there any results for Haswell or older processors that show these kind of gains with better ram? Because all the comparison I've seen were single digit gains at most, not 10%+ jumps in actual gaming...

 

[Dasa2]

I prefer to use real time latency as Intel does:
That gives:
DDR4 16/(2666/2)*1000 = 12 ns
DDR3 9/(1866/2)*1000 = 9.6 ns
So about 25% more latency for DDR4, while it has 2666/1866 ~= 1.42 so 40% more bandwidth. Can't wait to more complete studies which play on both latency and frequency of RAM for both DDR3 and DDR4

funny thing is calculated speeds dont seem to stack up to me when looking at these ddr3\ddr4 tests with aida64 and 7zip\arma http://www.hardware.fr/articles/940-5/cpu-ddr4-vs-ddr3-pratique.html
any ideas on why? maybe a side affect of skylake supporting ddr3\ddr4 is that ddr3 has a higher latency to the cpu

by that its seems that 1866c9 and 2666c16 would have a very similar latency
the bandwidth however is significantly higher not that it seems to be having as much of an affect on performance

Are there any results for Haswell or older processors that show these kind of gains with better ram? Because all the comparison I've seen were single digit gains at most, not 10%+ jumps in actual gaming...

yes look at my post near the end of the last page

 

[Nothingness]

funny thing is calculated speeds dont seem to stack up to me when looking at these ddr3\ddr4 tests with aida64 and 7zip\arma http://www.hardware.fr/articles/940-5/cpu-ddr4-vs-ddr3-pratique.html
any ideas on why?

It's because there's much more to real latency than using CAS and frequency For instance, memory bank opening, efficiency of the memory controller, etc. My latency ratio is most likely very wrong, but it was just a way to get a hint.

by that its seems that 1866c9 and 2666c16 would have a very similar latency
the bandwidth however is significantly higher not that it seems to be having as much of an affect on performance

As I said many parameters are at play here. Perhaps is the DDR4 controller much better?

 

[Head1985]

Still nobody post here this review?
http://www.purepc.pl/procesory/test...56600k_skylake_mocne_cztery_rdzenie?page=0,33

 

[frozentundra123456]

Granted, the results are looking better, but the inconsistency still bothers me, anand's results in particular. Especially since it appears you need very fast, expensive ram to see much benefit. So unless you have money to burn, it still seems a questionably worth it upgrade for anyone with a high clocked SB or better. It also makes x99 very close in price. Are there any results for Haskell E with fast DDR4?

 

[jpiniero]

Granted, the results are looking better, but the inconsistency still bothers me, anand's results in particular. Especially since it appears you need very fast, expensive ram to see much benefit. So unless you have money to burn, it still seems a questionably worth it upgrade for anyone with a high clocked SB or better. It also makes x99 very close in price. Are there any results for Haskell E with fast DDR4?

Haswell-E is Quad Channel, so faster DDR4 ram won't help much.

 

[StrangerGuy]

I really don't think its the DDR4 bandwidth that is holding Skylake back, more like a latency bottleneck as cranking up the frequency also reduces the effective latency.

 

[Enigmoid]

I really don't think its the DDR4 bandwidth that is holding Skylake back, more like a latency bottleneck as cranking up the frequency also reduces the effective latency.

Seems likely.

Hopefully this memory latency problem and the inclusion of AVX3 (L1 and L2 not big enough and need more bandwidth) will force intel to rework caches and increase the size.

Core has only had small changes in the cache and L1 has remained constant in size since Conroe. Its quite amazing the performance that can be extracted with newer vector extensions but IMO sizes will have to go up with AVX3.2.

 

[dark zero]

Seems that AMD is finally being left behind hard in the benchmarks they were great. They got stuck on figthing Sandy... maybe Zen is the very last move.

Meanwhile, what happened to Celeron Skylake? I saw that only there will be up to Pentium, being the lowest tier... maybe the "Celeron" becomes Intel Dual Core and put only on OEM's machines?

 

[Phynaz]

Interesting, Intel announced mobile Skylake Xeons:

http://blogs.intel.com/technology/2015/08/bringing-intel-xeon-to-notebook-pcs/

 

[CakeMonster]

Core has only had small changes in the cache and L1 has remained constant in size since Conroe. Its quite amazing the performance that can be extracted with newer vector extensions but IMO sizes will have to go up with AVX3.2.

Well, I'm guessing the only reason they haven't done it already is that they can get away with it and have bigger margins. Its not like they don't have the space. This is the kind of stuff I would pay extra for if I had the option.