Intel Skylake / Kaby Lake

[Subyman]

Think this would be a decent upgrade from a 4820k? Itching to upgrade my system and 2H 15 is good timing.

 

[ClockHound]

He is not normalizing by clockspeed. So you get a 14% increase from 4770k to Skylake, on clockspeed alone (base clocks), and he is assuming (optimistically I think) another 12.5% from ipc increase. So 1.14 x 1.125 = 28% actually, because you should multiply not add the gains, but that is how he is gettng the number.

And how much increase do you get in clockspeed alone from 4790k to Skylake?

 

[IntelUser2000]

147.68 GFLOPs (Skylake's theorized score at 3.7 GHz) divided by the equivalent Haswell score at the same clock speed (114.58) = 1.29x higher IPC, again based only on these assumptions of accuracy and scaling. I know SiSoft is far from a perfect source of specs and conditions, but for now it's all we have.

This is stupid.

Theoretical DP Flops of 3.7GHz Haswell = 2x(256-bit FPUs=4 DP FP operations) x 2 FMAs x 4 cores x 3.7GHz = 236.8GFlops

Same thing at 2.3GHz = 147.2GFlops.

Haswell's efficiency = 114.58/236.8 = 48.4%
Skylake ES's efficiency = 91.8/147.2 = 62.3%

Which are ENTIRELY possible because the former is throttling more in AVX operations than the latter because the latter clocks far lower.

And who ever thought GFlops represented average perf/clock anyway? Haswell has 1.5x+ GFlops of Ivy Bridge thanks to FMA, does it perform that high?

Yes! Low turbo (only 5% vs 10% on HW) is a bit concerning vis-à-vis overclocking - may be less headroom than HW

That's actually POSITIVE news. Because 6700K suggests(hopefully) that its not a "factory overclock it and sell it" part aka 4790K. Low Turbo frequency relative to Base frequency suggests that its easier to reach that Base frequency so they can skip the high Turbo. That's why the Desktop chips have high Base frequency and low Turbo gain relative to 47/57W mobile processors(which have similar Turbo but much less Base).

 

[mohit9206]

GTA V on Skylake iGPU Yay or nay?

 

[Sweepr]

He is not normalizing by clockspeed. So you get a 14% increase from 4770k to Skylake, on clockspeed alone (base clocks), and he is assuming (optimistically I think) another 12.5% from ipc increase. So 1.14 x 1.125 = 28% actually, because you should multiply not add the gains, but that is how he is gettng the number.

That's ~6-7% faster than Broadwell per clock. Broadwell has 5.5% higher IPC than Haswell and it's a Tick. Also Core i7 4770K was (on average) 17% faster than Core i7 2700K according to Anandtech so hopefully I'm being conservative, not optimistic.

 

[Racan]

That's actually POSITIVE news. Because 6700K suggests(hopefully) that its not a "factory overclock it and sell it" part aka 4790K. Low Turbo frequency relative to Base frequency suggests that its easier to reach that Base frequency so they can skip the high Turbo.

To me it suggests they've tried to reach the same frequencies as the 4790K and failed, 4/4.2Ghz base/turbo vs 4/4.4Ghz base/turbo.

 

[BigDaveX]

To me it suggests they've tried to reach the same frequencies as the 4790K and failed, 4/4.2Ghz base/turbo vs 4/4.4Ghz base/turbo.

Either that or they're planning to release a Devil's Canyon-type refresh when yields improve, and leaving themselves enough headroom that they'll be able to release a model (something like 4.2GHz base/4.5GHz turbo) that'll look significantly faster than the initial line-up.

 

[jpiniero]

Either that or they're planning to release a Devil's Canyon-type refresh when yields improve, and leaving themselves enough headroom that they'll be able to release a model (something like 4.2GHz base/4.5GHz turbo) that'll look significantly faster than the initial line-up.

The 4/4.2 IS the Devils Canyon type part. My guess would be that they would later release a 4+4e part (along with the L4) later.

 

[frozentundra123456]

Or, they could release a DC type faster clock skylake and then a BDW-k type part on Cannonlake. Who knows, it is all speculation at this point.

 

[Idontcare]

To me it suggests they've tried to reach the same frequencies as the 4790K and failed, 4/4.2Ghz base/turbo vs 4/4.4Ghz base/turbo.

They can probably hit the same clocks, but the operating temps end up punching well above comfortable TJmax limits established to ensure lifetime reliability metrics.

14nm could be heat-density limited in ways we've long worried about.

 

[Arachnotronic]

BTW, straight from an IDF 2015 presentation on AIO PCs...

This product family was not meant to succeed i7-4790K; that's Skylake's job.

 

[Shivansps]

65W and unlocked, kinda making the 65W thing useless. This may be an option for people not wanting to buy DDR4.

 

[IntelUser2000]

To me it suggests they've tried to reach the same frequencies as the 4790K and failed, 4/4.2Ghz base/turbo vs 4/4.4Ghz base/turbo.

Hardly. 6700K naming suggests that it may be easy as making a 4770K rather than 4790K. Or at least easier.

4790K took additional year from 4770K, and it has less headroom percentage-wise, meaning its sort of like a factory-overclocked part. Most parts clock conservatively so it doesn't create issues and they don't get sued. Hence most people get big overclocks.

With 6700K they are able to pull 4GHz right out of the bat, even though there are rumors that 14nm had significant problems and they still have issues. So at least they are doing something right.

 

[ClockHound]

They can probably hit the same clocks, but the operating temps end up punching well above comfortable TJmax limits established to ensure lifetime reliability metrics.

14nm could be heat-density limited in ways we've long worried about.

First global warming, now this.

 

[Sweepr]

I wish M$ releases a Skylake-U based Surface Pro 4. The rumoured 15W dual-core + GT3e Iris Pro SKU would be perfect for the more expensive versions. Maybe Apple is waiting for this SKU to launch the Retina Macbook Air too.

 

[Arachnotronic]

I wish M$ releases a Skylake-U based Surface Pro 4. The rumoured 15W dual-core + GT3e Iris Pro SKU would be perfect for the more expensive versions. Maybe Apple is waiting for this SKU to launch the Retina Macbook Air too.

I bet Apple was the one that told Intel to even make that SKU.

 

[Dribble]

Quick comparison using AnandTech results for Haswell and older chips. Assuming ~12,5% better performance (Core i7 6700K vs Core i7 4790K) and a 12% bump from Core i7 4770K to Core i7 4790K, here's how the new Skylake compares to current Intel processors @ stock:

26% faster than Core i7 4770K.
36% faster than Core i7 3770K.
47% faster than Core i7 2700K.
81% faster than Core i7 965.

I can see lots of Nehalem/SB/IB users finally upgrading.

Not so sure, no one is running their SB/IB stock, once you o/c the difference is much smaller. Either it's got to be 50% faster then my SB or it's got to have 6 cores.

 

[Muadib]

Not so sure, no one is running their SB/IB stock, once you o/c the difference is much smaller. Either it's got to be 50% faster then my SB or it's got to have 6 cores.

Speak for yourself. My 2600k is stock. It has been since I started leaving my pc on 24/7.

 

[Arachnotronic]

Speak for yourself. My 2600k is stock. It has been since I started leaving my pc on 24/7.

I know a lot of people who run "K" series processors at stock; these people simply can't afford any instability/data corruption.

 

[Muadib]

I know a lot of people who run "K" series processors at stock; these people simply can't afford any instability/data corruption.

Exactly! The only reason I got the K was microcenter pricing. I'm hoping that they do something similar with skylake.

 

[Enigmoid]

Hardly. 6700K naming suggests that it may be easy as making a 4770K rather than 4790K. Or at least easier.

4790K took additional year from 4770K, and it has less headroom percentage-wise, meaning its sort of like a factory-overclocked part. Most parts clock conservatively so it doesn't create issues and they don't get sued. Hence most people get big overclocks.

With 6700K they are able to pull 4GHz right out of the bat, even though there are rumors that 14nm had significant problems and they still have issues. So at least they are doing something right.

+1

Need to remember also that the initial number of broadwell 4 core chips is much lower than haswell at release meaning that they have a lot less dies to bin. That they can get such a high percentage with such high clocks means that 14nm is doing fairly decent.

 

[danjw]

Hardly. 6700K naming suggests that it may be easy as making a 4770K rather than 4790K. Or at least easier.

4790K took additional year from 4770K, and it has less headroom percentage-wise, meaning its sort of like a factory-overclocked part. Most parts clock conservatively so it doesn't create issues and they don't get sued. Hence most people get big overclocks.

With 6700K they are able to pull 4GHz right out of the bat, even though there are rumors that 14nm had significant problems and they still have issues. So at least they are doing something right.

While it did take Intel that long to release it, my understanding was it was more of a project because Broadwell was slipping. They only really worked on it for 6 months. Source: http://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/6

Third paragraph on the referenced page:

"This leads on to overclocking, the supposed reason for the release of Devil’s Canyon. Lisa Graff, VP and GM of Intel’s Desktop Client Platform Group, indicated in a press call before Computex that the lead design teams on Devil’s Canyon had six months to make the changes, and that the CPUs were aimed at addressing concerns levied at Intel from users who enjoy overclocking."

 

[Nothingness]

I know a lot of people who run "K" series processors at stock; these people simply can't afford any instability/data corruption.

I am one of them I have never overclocked any of my CPU.

 

[Sweepr]

I know a lot of people who run "K" series processors at stock; these people simply can't afford any instability/data corruption.

I know too. Some people are willing to pay more for better performance @ stock even if there are cheaper chips that offer that kind of performance with OC. Core i7 4790K and FX9590 are two examples.

If Core i7 6700K delivers >50% better performance than the popular Core i7 2600K @ stock a lot of people will upgrade, both Intel and AMD users. And of course there will be some OC headroom too, how much remains to be seen.

 

[utahraptor]

I am one of them I have never overclocked any of my CPU.

I have always been one to buy premium components hoping to overclock and I always fail miserably This time around with a $250 gigabyte board, a high end closed loop cooler and a gold certified PSU I could barely get over stock speeds on my 3770K without redlining on temps I get sick of reading reviews claiming any old budget board will get at least 500 MHz when I fail so often.