Kabylake review from [H]. This is really bad :(

[Eddward]

For several years there was a Cannonlake platform on the roadmap as a successor to Skylake. In the middle of last year Intel announced Kaby Lake, as "stopgap" platform because of delayed 10nm which means delayed Cannonlake uarch as well. So what can Intel do in this time frame ? Almost nothing. Even usual +5% IPC in die shrinks requires a few year of development and this was probably already done for Cannonlake. For Kaby Lake there is obviously nothing since this platform has never been in the long term planning. Coffe Lake will be the same, but fortunately for us Intel was forced to add more cores, since there was probably no other option to do atm. No new uarch for better IPC, frequency is on the top and still no new process (and Zen is also coming).
I don't know why are people so sad. Everything has its logic if you can read situation. The root cause of this situation is failed tick-tock cycle or in this case unready 10nm as originally planned a few years ago.
But yes, call Skylake refresh as a 7th gen Core with a new codename is a bit odd. They probably chose this way since we are already 1,5 years from the Skylake launch and another year with Skylake (refresh) will not look appealing for consumers.

 

[frozentundra123456]

For me the article referenced by the OP actually does tell the whole story. If clock speed isn't isolated IPC performance can't be evaluated. If IPC is the same but clocks are say 5% faster for one core over another I think most people around here realize that means, with equal IPC, that the 5% faster core will perform 5% better (about, in reality a bit less). It's the reviews that only compare one part to another when we don't know clocks during testing (or other test specifics), especially these days with variable frequencies, where you can't fully understand what is going on. Those kinds of tests are for non-enthusiasts in my opinion. They don't want to know about architecture, just which is faster from a performance point of view. I can get that information from a crappy Intel propaganda slide.

Equalizing clocks as this review does is like stripping the clothes of the Emperor!

Not really. One has to remember that the vast majority of users don't overclock. If stock clocks are higher and/or the chip can maintain higher turbo longer, there is a net performance increase. Most people don't care whether it comes from ipc or clock speed. Now one could argue that since the core is the same and the process only tweaked, it should not have been called a new generation, but it does not change the fact that testing at the same locked clock speed does not tell the entire story. In fact it strikes me as a sensationalist clickbait article, if you want my honest opinion.

 

[Hulk]

Not really. One has to remember that the vast majority of users don't overclock. If stock clocks are higher and/or the chip can maintain higher turbo longer, there is a net performance increase. Most people don't care whether it comes from ipc or clock speed. Now one could argue that since the core is the same and the process only tweaked, it should not have been called a new generation, but it does not change the fact that testing at the same locked clock speed does not tell the entire story. In fact it strikes me as a sensationalist clickbait article, if you want my honest opinion.

I can see your point but have a different opinion. Kaby Lake is what used to be called a Skylake stepping. The article was simply illustrating that point. I learned a lot from it. Namely that Kaby Lake is marketing and is really just a Skylake stepping to improve efficiency a bit. Intel HAS done this before they've just never called a stepping a new architecture name. That is what is striking to me. And like I said we've been farting around 3-4GHz for like 10 years. No news there. Kaby Lake might overclock a bit better than Skylake. Normal for a stepping.

 

[unseenmorbidity]

obviously anyone with a 6700K should not upgrade.
I would still consider buying a 7700K, despite everything written here, since i have an aging CPU and the gains for me would be considerable.

You should look at the 6700k honestly. You can likely find it and a z170 board for at least a $100 cheaper.

 

[iamgenius]

You should look at the 6700k honestly. You can likely find it and a z170 board for at least a $100 cheaper.

Hmmmmm. I wouldn't. I would still get an i7 7700K, delidd it, and then overclock ^_^ This or maybe a binned 6700K with high overclocking potential. I hate to see me < 5 GHz for 24/7.


Lots of good posts here: love you guys!

 

[BeepBeep2]

Intel's gains with Kaby Lake remind me a lot of AMD's gains with Phenom II X4 C2 / C3 steppings @ 45nm.
The last AMD 45nm chips that were produced all seemed to hit 4.3-4.5 GHz stable on air cooling, one of my X4 955s will still run P95 all day on air at almost 4.5 GHz. It was funny, because the last Phenom II X4s off the line in 2013 were still giving Bulldozer a hard time in many applications.

Outside of the new 10-bit HEVC decoding for Kaby Lake - like The Stilt, I still feel it should have been called 6790K or something. Devil's Canyon had the thermal paste and the significant packaging tweaks with it's own new codename too ... but it stayed in the 47x0K family.

 

[Hulk]

Q6600 G0 stepping efficiency improvements. Check out the power consumption graphs. Look familiar?

http://www.anandtech.com/show/2303/2

 

[vissarix]

To be fair im very disappointed with intel lately...first it was devil canyon now kabylake which gives you nothing but 100 or if your lucky 200mhz more headroom oc.....hopefully zen will shake things up so we get more cores at accessible prices and intell will be forced to come out with something better and lower prices aswell

 

[witeken]

To be fair im very disappointed with intel lately...

Are you very disappointed with Intel, or are you disappointed with the laws of nature?

You're like someone who'd be very disappointed if cars didn't become much faster every year.

For some reason people have this idea that CPUs "should" just become much faster continually, which is not very realistic. Your expectations are too high.

 

[vissarix]

Are you very disappointed with Intel, or are you disappointed with the laws of nature?

You're like someone who'd be very disappointed if cars didn't become much faster every year.

For some reason people have this idea that CPUs "should" just become much faster continually, which is not very realistic. Your expectations are too high.

Since they cant make anything faster why dont they give us more cores at the same price? They simply dont want to...but if zen comes out as competitive watch intel drop 6cores skylake prices like hotcakes...

 

[.vodka]

Since they cant make anything faster why dont they give us more cores at the same price? They simply dont want to...but if zen comes out as competitive watch intel drop 6cores skylake prices like hotcakes...

This, Intel has been executing like crap lately, nevermind them still being on top of the market for the past decade. When is cofee lake due? 2018? 7700k should've been 6c Skylake + improved iGPU on the mainstream 1151 socket. I agree 7700k as it is should've been 6790k, no more. Cofee lake itself seems to be a last minute afterthought to keep 14nm fabs going since cannonlake will be 10nm, I don't think Intel would like to go through another 14nm yield issue problem and product scarcity as they have in the past.

*If* Zen ends up being as competitive as shown so far and priced right, yeah, watch intel prices drop like a rock especially considering there's nothing really new for consumers apart from Skylake-X on the expensive HEDT platform.

On the other hand, I'd love to see what the real, actual, beefy Skylake server cores can do.

 

[Dave3000]

Does anyone here know what the i7-7700k's 4-core turbo bin will be? Sources say the base clock is 4.2 GHz and the turbo clock is 4.5 GHz, but that's too vague. It would be awesome if the stock 4-core turbo bin is going to be 4.5 GHz.

 

[frozentundra123456]

Are you very disappointed with Intel, or are you disappointed with the laws of nature?

You're like someone who'd be very disappointed if cars didn't become much faster every year.

For some reason people have this idea that CPUs "should" just become much faster continually, which is not very realistic. Your expectations are too high.

I agree partially: IPC and clockspeed have pretty much hit the wall, and I dont know what Intel can do there, except maybe eventually next gen materials. But for sure they should have made a mainstream hex core a long time ago, and shifted the rest of the product line down accordingly.

 

[Eug]

From my perspective as a non-overclocking consumer - which represents 99% of the consumer population - Kaby Lake is a huge upgrade, precisely because of the media improvements. Speed improvements are there too, esp. on mobile.

In fact, this is the upgrade I've been waiting for for years. I specifically avoided Skylake because it's still stuck in the past with regards to media support. I'll be buying a Kaby Lake computer this spring.

It would have been very foolish for Intel to call this Skylake, considering that Kaby Lake represents such a huge jump in media support.

 

[Hulk]

From my perspective as a non-overclocking consumer - which represents 99% of the consumer population - Kaby Lake is a huge upgrade, precisely because of the media improvements. Speed improvements are there too, esp. on mobile.

In fact, this is the upgrade I've been waiting for for years. I specifically avoided Skylake because it's still stuck in the past with regards to media support. I'll be buying a Kaby Lake computer this spring.

It would have been very foolish for Intel to call this Skylake, considering that Kaby Lake represents such a huge jump in media support.

Not trying to be a smart-ass here. I'm curious what media improvements in Kaby Lake are going to specifically make a difference as compared to Skylake to your computing experience?

 

[Hulk]

Are you very disappointed with Intel, or are you disappointed with the laws of nature?

Point taken. I'm disappointed that Intel has named a Skylake stepping as a new core. Call it "Skylake plus" and add a "50" to the serious number.

And If I were Intel, from a marketing and profit point of view I'd slow the wheels too. What's the point of competing against yourself? Why offer a cheap 6 core mainstream part when you can sell a more expensive 6 core part? It's not like there is a competitive alternative. Can't fault Intel or anybody. But the situation kind of stinks for us.

Kaby Lake 8 core/16 thread 7708k part for $300? Yeah that would get me driving to Microcenter...

 

[PliotronX]

Not trying to be a smart-ass here. I'm curious what media improvements in Kaby Lake are going to specifically make a difference as compared to Skylake to your computing experience?

Hardware decoding and encoding of 4k YouTube videos...

 

[beginner99]

Pretty boring time to be a CPU enthusiast. At least GPUs are still advancing.

Sadly not in the metric that matters most. performance/$. It's actually getting worse in that regard.

 

[Arachnotronic]

I agree partially: IPC and clockspeed have pretty much hit the wall, and I dont know what Intel can do there, except maybe eventually next gen materials. But for sure they should have made a mainstream hex core a long time ago, and shifted the rest of the product line down accordingly.

That trick only works once. What happens once expectations have now shifted to mainstream hex, rather than quad? You get one big jump, and then you're back to the same old "rut" that you were in before...relying on frequency/IPC improvements to boost performance at iso core counts.

At some point, the "add more cores" trick just stops working, and can even be detrimental to segments performance notebooks (though if Intel is able to get a hex core Coffee Lake into the same 45W TDP for notebooks w/o sacrificing 1-4 core performance, I'll be impressed).

 

[beginner99]

From my perspective as a non-overclocking consumer - which represents 99% of the consumer population - Kaby Lake is a huge upgrade, precisely because of the media improvements. Speed improvements are there too, esp. on mobile.

In fact, this is the upgrade I've been waiting for for years. I specifically avoided Skylake because it's still stuck in the past with regards to media support. I'll be buying a Kaby Lake computer this spring.

It would have been very foolish for Intel to call this Skylake, considering that Kaby Lake represents such a huge jump in media support.

Exactly. While i agree that kabylake is pretty much fail for desktop usage, it's a win in mobile because iGPU (media engine) matters there as no dGPU is available and using software decode is either too slow or uses too much power. I mean we knew exactly that kabylake would be and he know since couple years that intels focus is mobile and that was once more demonstrated. For laptops kabylake is a relevant improvement due to higher clocks and iGPU.

That's why desktop will also get Coffeelake while mobile can use newest Cannonlake and 10nm. Less supply issues for 10nm parts, the can give desktops users something that should not cost very much to create and it helps to amortize the 14nm factories.

 

[Arachnotronic]

That's why desktop will also get Coffeelake while mobile can use newest Cannonlake and 10nm. Less supply issues for 10nm parts, the can give desktops users something that should not cost very much to create and it helps to amortize the 14nm factories.

Coffee Lake will also be used in high performance mobile (15W/28W with eDRAM/45W). Looks like 10nm Cannon Lake will be limited to Core M type products. 14nm will have a loooong life.

 

[mikk]

Cannon Lake is listed for 5.2W and 15W in Roadmaps.

My i7-7700k runs much better by the way. Locked to 4.2 Ghz the VID is 1.20V whereas my i7-6700k had a VID of 1.29V with locked to 4.0 Ghz. As a result power consumption is much lower.

 

[unseenmorbidity]

Coffee Lake will also be used in high performance mobile (15W/28W with eDRAM/45W). Looks like 10nm Cannon Lake will be limited to Core M type products. 14nm will have a loooong life.

Wait, so is cannonlake going to be like a repeat of broadwell!?

 

[Hulk]

Hardware decoding and encoding of 4k YouTube videos...

Hardware decoding for battery life purposes I suppose? Or do you have issues software decoding?

As for the YouTube videos. Do you find the quality of the hardware encoder sufficient? I generally notice it to be of noticeable lower quality compared to Handbrake, especially at the low bit rates required by YouTube. Just curious.

 

[Arachnotronic]

Wait, so is cannonlake going to be like a repeat of broadwell!?

Cannon Lake is HEDT and U/Y series only. Coffee Lake will come next for high performance notebooks/desktops (third gen 14nm product), and after that will come Ice Lake, at which point the high perf CPUs and low power CPUs converge.

So, yes.