Intel Skylake / Kaby Lake

[AtenRa]

I dont think its even worth upgrading from sandy bridge.I have 2500K@5Ghz almost 5 years old..

Yeap, SandyBridge owners better upgrade their GPU next year than go with Skylake.

 

[B-Riz]

Lol, Skylake 1151 is underwhelming.

Pretty sure I will go Broadwell-E, Skylake-E next, or Zen.

For gaming, SB is still awesome and, just further cements that a boss GPU is most important.

It was nice to jump to Devils Canyon at 4.5 - 4.6 and be able to keep using the 16 GB o DDR3 I already paid for. :whiste:

 

[Sweepr]

GFXBench results:

Skylake GT2 is 35% faster @ Manhattan offscreen and 68% faster @ T-Rex offscreen.

Well, some of improvements look nice for the folks that use them. For me, I'm a gamer who builds strictly gaming pc builds. I'll just sit on my locked i7 4770, I just don't see any of these chips that offer enough to put out money on it. I certainly am not going to buy new memory on top of everything else for something that appears to be a regression for my usage case in some scenarios. Perhaps the E line will give me a reason at some point.

Look at the tests above and Hardware.fr review. With the right memory kit it's clearly faster than Haswell in CPU-limited games. With that said, Core i7 4770 is a great chip, you can keep it till Cannonlake/Icelake or grab a hexa-core at some point.

It's not just games guys, Skylake-S loves fast memory in applications too:

http://arstechnica.com/gadgets/2015/08/intel-skylake-core-i7-6700k-reviewed

 

[Keysplayr]

Sigh................... Looks like my Sandy and I are going to be together for a while longer. A whole lot of money for a whole little performance improvement. This isn't a tick, or a tock. This is a "doink".
The biggest benefits lies in the platform itself and motherboard features. up to 40 lanes of PCI-e 3.0 goodness for taking full advantage of M.2 native PCI-e storage. Won't steal lanes from your graphics cards. That I must admit is a nice option if you get a board equipped with it.
As far as the CPU itself? Well, looks like a wall has been hit they just can't get around. Or, lack of incentive to work on another "Conroe".

/Sigh..........

Money for new platform now being diverted to new wheels and A/T tires all around my vehicle.

 

[frozentundra123456]

Even against Haswell GT2 is not more than 15-20% faster on average.
Still slower than 1.5 year old 65W Kaveri.

Are we looking at the same graphs? 95 watt Kaveri top of the line is 5% faster in Bioshock, while Skylake is 58% faster in Half Life and 25% faster in GTA V.

And that is the "baseline" quad core igp. To be honest, I am surprised you would even comment on these charts, since they are an absolute humiliation for AMD. Think what iris pro will do.

 

[SlowSpyder]

Just looked at a few dgpu gaming benches. It looks like at low settings Skylake is faster, but at settings you'll actually use it can be slower than previous CPU's. This is a great example of why practical real-world higher setting gaming benches are important in CPU tests.

 

[NTMBK]

Are we looking at the same graphs? 95 watt Kaveri top of the line is 5% faster in Bioshock, while Skylake is 58% faster in Half Life and 25% faster in GTA V.

And that is the "baseline" quad core igp. To be honest, I am surprised you would even comment on these charts, since they are an absolute humiliation for AMD. Think what iris pro will do.

He was referring to the A10-7800 by "65W" part, as it performs much the same as the 7850k.

But yeah, Skylake really pulls away in bandwidth heavy games. That DDR4 is pretty lovely.

 

[Pwndenburg]

Yea, if I was willing to buy an expensive set of DDR4 it would be a significant increase. But, I just don't see it right now. I'm still trying to get over the "bummed out" factor of this review. Looks like the good ol Z87 will keep on truckin for a few more years. I know so little about all this that I don't feel my comment should be taken with much merit here. But, my gut tells me this is the direct consequence of designing chips mobile first. I realize that they must do this. On the upside, we DO have ample power available to us right now. Any improvements at this point seem to just be gravy. This is especially true of someone like me that runs a single gpu setup.

 

[BFG10K]

Erm, remind me when that happened? Because I sure don't remember that.

Discrete GPUs have made similar gains over the same period- compare a HD 6970 with a Fury X- but they took just as long to get there.

Compare a GT-730 to a Titan Black (both released in 2014). The Titan would have no trouble being 4.5x faster, especially if the 730 runs out of VRAM.

If you want to span four years, compare a GT440 (2011) to a Titan-X (2015). The performance difference should be close to a factor of 20x, certainly more than 10x.

EDIT: And I was comparing with the mainstream chip, i.e. Skylake. If you want to compare with Broadwell with eDRAM, it's over 6X faster... and discrete GPUs have not improved any near that much in the same time period.

Go ahead and compare the slowest 2011 IGP to the newest Iris in 2015. It doesn't matter because a Titan-X will smash the GT-440 by much more than that.

You might argue that it isn't fair, but it's reality. A user who purchased a GT-730 back in 2011 has the option to upgrade to a Titan-X in 2015 and get a far bigger performance difference than a 2011 IGP user upgrading to a 2015 Iris will get.

 

[ioni]

So Skylake seems to be slightly worse for gaming, but what about gaming while stream broadcasting? Since it beats DC in basically every non-gaming benchmark, would that make it the better choice for streaming your gaming sessions?

 

[csbin]

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation

lol。。。。。

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up
Haswell to Skylake (DDR3): Average ~5.7% Up.

 

[froggermuted]

So Skylake seems to be slightly worse for gaming, but what about gaming while stream broadcasting? Since it beats DC in basically every non-gaming benchmark, would that make it the better choice for streaming your gaming sessions?

It's not worse for gaming with proper ram, sadly the ddr4 is far away off being on par with the current ddr3. It was similar with the x99.

 

[NTMBK]

Compare a GT-730 to a Titan Black (both released in 2014). The Titan would have no trouble being 4.5x faster, especially if the 730 runs out of VRAM.

If you want to span four years, compare a GT440 (2011) to a Titan-X (2015). The performance difference should be close to a factor of 20x, certainly more than 10x.


Go ahead and compare the slowest 2011 IGP to the newest Iris in 2015. It doesn't matter because a Titan-X will smash the GT-440 by much more than that.

You might argue that it isn't fair, but it's reality. A user who purchased a GT-730 back in 2011 has the option to upgrade to a Titan-X in 2015 and get a far bigger performance difference than a 2011 IGP user upgrading to a 2015 Iris will get.

I am comparing two comparable products, Intel's flagship CPU. (Which once again does not have the fastest IGP available, just like back in the Sandy Bridge days.) You are comparing completely opposite ends of the cost spectrum. That would be like comparing a single-channel, 2W Bay Trail part against the i7-5775C.

 

[Hulk]

Interesting.

It seems like architectural improvements and process shrinks are stalling out at the same pace.

 

[NTMBK]

So Skylake seems to be slightly worse for gaming, but what about gaming while stream broadcasting? Since it beats DC in basically every non-gaming benchmark, would that make it the better choice for streaming your gaming sessions?

Go for a Haswell-E, the extra cores will help with multitasking.

 

[Zodiark1593]

Absolutely no regrets pulling the trigger on my i5-4590. Much, much cheaper for excellent performance as opposed to going to the new platfor m and RAM.

 

[ShintaiDK]

Interesting.

It seems like architectural improvements and process shrinks are stalling out at the same pace.

Look at Hardocp review. Its a memory bottleneck issue.

Memory advancement have really been terrible.

 

[Sweepr]

So Skylake seems to be slightly worse for gaming, but what about gaming while stream broadcasting? Since it beats DC in basically every non-gaming benchmark, would that make it the better choice for streaming your gaming sessions?

It isn't. AnandTech used the slowest DDR4 out there, 2133MHz.

It beats Haswell using faster DDR4 RAM as shown by HardOCP tests above.
Hardware.fr used DDR4 3600MHz and Skylake beat Haswell by 10% per clock (average) in the games they tested.

 

[BFG10K]

I am comparing two comparable products, Intel's flagship CPU. (Which once again does not have the fastest IGP available, just like back in the Sandy Bridge days.) You are comparing completely opposite ends of the cost spectrum. That would be like comparing a single-channel, 2W Bay Trail part against the i7-5775C.

5775C and Baytrail are fundamentally different platforms so a desktop Iris is completely irrelevant because Baytrail can't upgrade to it. Such a comparison makes no sense.

OTOH a Titan-X can be dropped into practically any PEG slot, so the reality is 4.5x doesn't mean much when a dGPU upgrade can offer a much bigger gain.

 

[phillyman36]

Newegg has bundles that don't seem to be available until 8/14. My Microcoenter has no cpu and still only 2 mobo.

http://promotions.newegg.com/intel/15-3811/index.html?icid=324906

Asus still hasn't posted the specs and downloads for any of their mobo.

 

[Head1985]

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation

lol。。。。。

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up
Haswell to Skylake (DDR3): Average ~5.7% Up.

sad

 

[AtenRa]

Are we looking at the same graphs? 95 watt Kaveri top of the line is 5% faster in Bioshock, while Skylake is 58% faster in Half Life and 25% faster in GTA V.

And that is the "baseline" quad core igp. To be honest, I am surprised you would even comment on these charts, since they are an absolute humiliation for AMD. Think what iris pro will do.

I would suggest you not to take toms igpu result seriously,

AMD A8-7600 65W TDP is ~10% faster than 95W TDP DDR-4 Skylake GT2.
AMD A10-7800 65W TDP is 18% faster than 95W TDP DDR-4 Skylake GT2

Total War

AMD A8-7600 65W TDP is ~12% faster than 95W TDP DDR-4 Skylake GT2.
AMD A10-7800 65W TDP is ~13% faster than 95W TDP DDR-4 Skylake GT2

GTAV

AMD A8-7600 65W TDP is ~9% faster than 95W TDP DDR-4 Skylake GT2.
AMD A10-7800 65W TDP is 8% faster than 95W TDP DDR-4 Skylake GT2

GRID

AMD A8-7600 65W TDP is ~19% faster than 95W TDP DDR-4 Skylake GT2.
AMD A10-7800 65W TDP is 23% faster than 95W TDP DDR-4 Skylake GT2

Shadows Of Mordor

AMD A8-7600 65W TDP is 13% faster than 95W TDP DDR-4 Skylake GT2.
AMD A10-7800 65W TDP is ~16% faster than 95W TDP DDR-4 Skylake GT2

 

[AtenRa]

He was referring to the A10-7800 by "65W" part, as it performs much the same as the 7850k.

But yeah, Skylake really pulls away in bandwidth heavy games. That DDR4 is pretty lovely.

No i was referring to the A8-7600, see above post.

 

[Sweepr]

All reviews added to first page!

POV Ray

lol。。。。。

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up
Haswell to Skylake (DDR3): Average ~5.7% Up.

Memory bound as shown above.

 

[frozentundra123456]

He was referring to the A10-7800 by "65W" part, as it performs much the same as the 7850k.

But yeah, Skylake really pulls away in bandwidth heavy games. That DDR4 is pretty lovely.

Yea, well the point was cherry-picking data. In two of three games Skylake was clearly faster, while in the one game he was touting as an AMD victory, the difference was pretty much within the margin of error.