Intel Skylake / Kaby Lake

[coercitiv]

Not really, I already said about HSW power saving features. When software reads the core frequency the core needs to be active in order to do so. IOW when CPU-z is reading the frequency then bclk should not be in a power saving state. So the question remains. Why is it showing 24MHz?

A screenshot of SKL running 100MHz bclk would suffice to answer the question. 😉

Uhm... one of the earlier leaks?

 

[Dufus]

Thanks coercitiv. You know I think I've seen that before, just couldn't remember it lol.

 

[cytg111]

we need to take a 4 month time then wake up to even dream of any of this to happen

Windows10 launch -> Surface4 launch -> Skylake launch.

- Assign your own odds.

 

[Sweepr]

It would be a shame if Surface Pro 4 launches with Broadwell-U in August. I hope they're waiting for Skylake-U.

 

[mikk]

http://browser.primatelabs.com/geekbench3/2776885


Does anyone have a trustable default 4790k result? Geekbench database results are all over the place.

 

[Arachnotronic]

http://browser.primatelabs.com/geekbench3/2776885


Does anyone have a trustable default 4790k result? Geekbench database results are all over the place.

Uh yeah, give me a few minutes to downclock my 4790K back to stock.

 

[Arachnotronic]

Here you go -- stock 4790K:

http://browser.primatelabs.com/geekbench3/2779933

 

[mikk]

Your result is x64. We need a x86 result.

 

[Arachnotronic]

Your result is x64. We need a x86 result.

No problem. I'll run a 32-bit benchmark.

 

[Arachnotronic]

32 bit result:

http://browser.primatelabs.com/geekbench3/2780128

 

[ShintaiDK]

Depending on turbomodes and what else.

But else its 5.6% MT and 7.1% ST clock for clock.

If 4.2 vs 4.4Ghz turbo its 11.2% ST.

 

[mikk]

Turbo enabled?


http://browser.primatelabs.com/geekbench3/compare/2780128?baseline=2776885


AES seems much improved. Some regression there as well, maybe caused by the clock deficit. Of course with only one 6700k result available this is a risk.

 

[Arachnotronic]

Turbo enabled?


http://browser.primatelabs.com/geekbench3/compare/2780128?baseline=2776885


AES seems much improved. Some regression there as well, maybe caused by the clock deficit. Of course with only one 6700k result available this is a risk.

Yes, turbo enabled.

 

[Sweepr]

32 bit result:

http://browser.primatelabs.com/geekbench3/2780128

Can you run it at 4.0 or 4.1GHz (fixed)?

 

[Arachnotronic]

Can you run it at 4.0 or 4.1GHz (fixed)?

Sure.

 

[frozentundra123456]

Lets hope final results are better than that. It is only a 7 to 10 percent gain depending on turbo behavior, and that is for 2 generations. Even worse, it is not a consistent across the board gain. Most of the results are pretty much a wash, with the gains coming mainly from AES. The more leaks we see, the less optimistic I am becoming.

 

[Arachnotronic]

4790K @ 4GHz, 32-bit mode:

http://browser.primatelabs.com/geekbench3/2780330

 

[Nothingness]

Depending on turbomodes and what else.

But else its 5.6% MT and 7.1% ST clock for clock.

If 4.2 vs 4.4Ghz turbo its 11.2% ST.

If you remove the memory score, the ST score is lower.

 

[R0H1T]

If you remove the memory score, the ST score is lower.

They're higher cause of DDR4 I bet, it seems that Skylake refresh is the one to get, just like Devil's Canyon, unless of course one's in desperate need of it 😛

 

[Nothingness]

Yeah, memory bandwidth looks very good 🙂

 

[ShintaiDK]

They're higher cause of DDR4 I bet, it seems that Skylake refresh is the one to get just like Devil's Canyon unless of course one's in desperate need of it 😛

Its quite clear that its hold back for a refresh. But thats what we need to get used to after Atoms kicked Core desktops down on the node priority list.

The DDR3 SKUs should also be 1866Mhz standard. At least thats what i saw on the Skylake NUCs.

 

[frozentundra123456]

Well, maybe the memory bandwidth will help igpu performance if anybody really cares on the desktop. Is mobile skylake DDR4 also?

Even broadwell was accepted to have about 5% better ipc than Haswell. Maybe skylake final shipping products in real x86 apps will show up better. I dont really know how relevant Geekbench is for x86, but it seems to be becoming the go-to measurement.

 

[Enigmoid]

If these are any indication of real world performance Skylake is shaping up to be pretty sad. Then again geekbench is a terrible measurement.

 

[ShintaiDK]

Well, maybe the memory bandwidth will help igpu performance if anybody really cares on the desktop. Is mobile skylake DDR4 also?

Y/U seems to be DDR3L and H DDR4. While S can be both.

 

[LTC8K6]

If these are any indication of real world performance Skylake is shaping up to be pretty sad. Then again geekbench is a terrible measurement.

http://www.legitreviews.com/wp-content/uploads/2015/05/6700k_skylake_geekbench.png

Who's right?