Intel Skylake / Kaby Lake

[TahoeDust]

I removed the TR result that was produced with the lower clocked GPU. I don't want anyone to think I am trying to bias the results.

 

[Phynaz]

After looking at the slides from the ice lake preview, it looks to me like 14++ will be the highest performing process until 10++. Maybe I'm taking the graph too literally, obviously there are other factors in play (performance is not really a scalar metric).

http://images.anandtech.com/doci/11722/kaizad-mistry-2017-manufacturing(1)_29.png

Depends on how you define performance. For sheer processing speed yes, 10nm parts are going to regress from 14nm - we've known that since at least January.

For mobile performance 10nm parts should offer similar processing speed to today's parts while reducing power consumption. Which in that measurement means increased performance.

 

[IntelUser2000]

Depends on how you define performance. For sheer processing speed yes, 10nm parts are going to regress from 14nm - we've known that since at least January.

For mobile performance 10nm parts should offer similar processing speed to today's parts while reducing power consumption. Which in that measurement means increased performance.

I don't think the dual core mobile parts that will be fabbed on 10nm will be slower, rather the opposite. They'll use the lower power to increase frequency even a little, if not just better Turbo. They never regressed on frequency with the mobile parts, mostly increased.

One thing I hope more than anything else with Cannonlake is the graphics. The Gen 10 GT2 graphics is said to have 40EUs. I hope the performance increase is commensurate with the EU count(which would require significant architectural improvements).

 

[TahoeDust]

No, I saw somebody with 2012 got 6100 or something, but 2000 was the best it would do, and I think that was throttling. I saw 2038, and after a minute, it went down to 2000 (XOC precision was running)

You're right. I forgot those were run at 2012 because @Darkhelmutt 's card was throttling. I just ran this at 1999MHz. It would be a more direct comparison to your score.

 

[Markfw]

You're right. I forgot those were run at 2012 because @Darkhelmutt 's card was throttling. I just ran this at 1999MHz. It would be a more direct comparison to your score.

But its still odd that my CPU went up a lot, and no change to the score, but GPU mhz change DID

 

[TahoeDust]

But its still odd that my CPU went up a lot, and no change to the score, but GPU mhz change DID

It's definitely odd. Maybe there is something else in the platform making the difference? All of the intel results are fairly consistent. I wonder where a Ryzen based system would land?...closer to the intel chips or your TR?

 

[coffeeblues]

https://www.fool.com/investing/2017/08/16/intel-corporation-is-probably-going-to-kill-this-p.aspx

The author didn't provide the source for his supposed CFL-X leak, which my search shows to be 2016 wccf article and it doesn't look like CFL-X has been repeated ever since, which wccf never misses the opportunity to do. So he tried to create the impression that CFL-X plan is real so he could have something to doubt and then express his new found personal doubts in Intel's name and invoke the act of "killing it". Internet garbage at it's finest.

 

[dullard]

https://www.fool.com/investing/2017/08/16/intel-corporation-is-probably-going-to-kill-this-p.aspx

Kaby Lake-X are niche chips in a niche market (those who want to overclock as high as possible). But considering that the cheaper Kaby Lake-X only motherboards aren't even on the market yet, I think it is premature to call it a dud chip that Intel will kill off. Since it is a niche market, it will never be the top seller.

The concept is to give chip enthusiasts almost everything that they wanted all at no extra cost: no IGPU using power, faster clocks, more overclocking headroom, better binning, and more TDP. Intel just left off the soldering portion and they'd have answered every single complaint about the Kaby Lake processors.

But, the "no extra cost" part is a problem until the Kaby Lake X only motherboards are here. There is a cost of going to HEDT motherboards, in real money terms and in motherboard teething concerns.

 

[IntelUser2000]

The author didn't provide the source for his supposed CFL-X leak, which my search shows to be 2016 wccf article and it doesn't look like CFL-X has been repeated ever since, which wccf never misses the opportunity to do. So he tried to create the impression that CFL-X plan is real so he could have something to doubt and then express his new found personal doubts in Intel's name and invoke the act of "killing it". Internet garbage at it's finest.

Don't be so quick to judge. Or you might become the "internet garbage" you mention. Actually Ashraf is quite accurate in regards to his information.

https://benchlife.info/intel-coffee-lake-with-14nm-process-will-launch-2018-11192016/

That chart is the same chart that first shown the existence of Coffeelake-S for LGA 115x sockets. It may not finalize as a product, but CFL-X was surely in planning. Every product in that chart except for CFL-X is coming in a few months. For CFL-X it remains to be seen.

no IGP using power,

IGP uses zero power. It's power gated when not used.

 

[PeterScott]

https://www.fool.com/investing/2017/08/16/intel-corporation-is-probably-going-to-kill-this-p.aspx

They should KIWF.

It's a totally brain-dead product. Whoever thought that was a good idea, must work in an echo chamber full of yes men.

 

[coffeeblues]

Don't be so quick to judge. Or you might become the "internet garbage" you mention. Actually Ashraf is quite accurate in regards to his information.

https://benchlife.info/intel-coffee-lake-with-14nm-process-will-launch-2018-11192016/

That chart is the same chart that first shown the existence of Coffeelake-S for LGA 115x sockets. It may not finalize as a product, but CFL-X was surely in planning. Every product in that chart except for CFL-X is coming in a few months. For CFL-X it remains to be seen.

If Intel was planning to release CFL-X they would have put it on a roadmap.

Continuing with the content of the article, I don't see why the unimpressive reviews of KBL-X, which was just a repackaged chip pressured by 7800x and Ryzen, would be significant in the making or not making of CFL-X (even if it would be compatible with and could be transplanted to x299 platform). Notice that the mood of expectations for 8700k as compared to 7700k (from which KBL-X was made) is decidedly different and our local information supplier even went so far as to hint that 8700k would perform no worse than 7800x.

The existence of SKL-X only boards does not make it an absence of boards that CFL-X could be run on, there are still low-end (and even low-end cpus only) boards and even top of the line boards that can run both low-end and high-end cpus.

Intel could make CFL-X if it wouldn't have an inner conflict with creating 3 products eating each other at close to 400$ price point, but not because of KBL-X reviews or board nonsense.

I don't see the substance in the article that would support it's title, so I will say it was written just for the loud headline.

 

[coffeeblues]

I think margins are not gained because the PC market has declined.

Anyway I believe they should sacrifice margins to improve products. They focus too much on margins. 50% margin would be fine, if the revenue is 20% greater than today. Intel's Net Profit sucks even with those amazing margin numbers.

20% higher revenue in a shrinking market is not happening and would turn it in to a growing market all by itself, not even Intel could do it, and the margins for products sacrifice would result in even lower profits if the higher revenue target is not met.

Business exists to make more money for itself, margins for products sacrifice for the benefit of the consumer is a charity.

They are #1 by far in R&D spend. Many more companies do more than them in actual products and money generated. What's wrong with them?

Intel was the largest semiconductor company in the world (until Samsung just recently overcame them in semiconductor revenue) with fairy tale margins and tax optimised profits so nobody did more volume of products and money. What's wrong with you?

 

[Arachnotronic]

Depends on how you define performance. For sheer processing speed yes, 10nm parts are going to regress from 14nm - we've known that since at least January.

This chart most likely refers to transistor drive currents at a given xtor leakage level; this will impact the kind of frequency capability chips based on the processes will be able to deliver.

The second chart shows that capacitance goes down significantly from the 14nm technologies to the 10nm technologies; this should help Intel cram more stuff onto the die (more stuff = higher capacitance) at iso power levels.

tl;dr -- while peak frequency capability goes down, Intel can still drive higher perf by integrating more stuff into the chip, increasing IPC on the core by throwing more xtors at it, more cores, etc. without blowing up the power budget.

So I expect that 10nm+ Ice Lake will move ST perf forward from CFL in the desktop market by pushing IPC up by more than enough to offset what will be a mild reduction in peak frequency capability.

Then at the 10nm++ generation (Tiger Lake), Intel can deliver a nice generational boost by pushing frequencies to new heights.

For mobile performance 10nm parts should offer similar processing speed to today's parts while reducing power consumption. Which in that measurement means increased performance.

True! Peak frequency might not move up much, but the sustained frequency could go up pretty nicely with the 10nm mobile parts. And that's really what matters when you are power limited.

 

[DrMrLordX]

So you are going to use older results when newer information is available?

Link me a new review from a site like AT or what have you. I'm fairly confident AT WILL review Skylake-X for gaming again . . . it's just that they haven't done it yet.

The crux of my argument is that I'm responding to flawed logic and arguments.

Flawed logic and arguments that apprently don't exist . . .

I've not updated to any recent hardware release, I'm on the wait for z390 train, not the wait for z370 train so my patience is rational, not emotional.

What does z370 vs z390 have to do with any of this? The chip is going to run about the same on either platform, assuming that the VRMs are there to get it to 5 GHz or however high it can go.

The same as highly-assumptive conclusion.

BS. Compare a Haswell i3 to a Haswell i7 at the same clockspeed with the same memclock in something like CB R15. The i7 isn't going to be *slower*.

without having any evidence to back it up

Pfft.

Until then 8700k for top dog cpu is expectation, not a fact or conclusion or truth.

Okay.

Zen3 5Ghz 7nm process for best gaming cpu ever.

There you go trying to bring AMD into it.

If the 8700K does not best the 7700K in all areas, it will be a disappointment to me.

Exactly. Same core, improved process, more L3 + 2 more cores, same core interconnects, same cache layout . . . Intel would have to really screw things up to get it wrong.

Coffeelake is what kaby should have been.

It's actually what Skylake should have been, but when you consider how long it took Intel to get 14nm right, maybe that's too much to expect from them.

It seems that the CPU speed and SMT mode had nothing to do with the score. CPU@3400 5700 something score, CPU@4100, exact same score. I tried everything. But when I messed with the mhz of the video card I got up to 2000 and got this:

Have you tried at different RAM speeds to see if it scales with RAM clock? And what happens when you set low custom resolutions?

 

[Shivansps]

The problem with the whole "Intel milked quad cores" thing is that it makes sence only if people upgraded every gen... who did SB->IVY->Haswell->Skylake->Kabylake? I bet that is a minority. Contrary to what people say, the lack of bigger performance increases and no core increase cause people to sit on the hardware more time than normal... i used my 2500K for 6 years, thats the longest i used a CPU EVER. Petty sure people sit on the SB/IVY/Haswell/Skylake cpus waiting for a good upgrade.
And the lower end, Celeron to i3, did get significant perf increase from gen to gen btw...
So what Intel did, from the profit POV, it just dont seem right to me, Intel failed at point number 1 in product marketing, create a demand. Petty sure they hurt their own margins by doing that.

Considering the state of software and core usage, the idea of increasing core count in 2016 with CannonLake was right, they just failed to deliver.

 

[AdamK47]

They should KIWF.

It's a totally brain-dead product. Whoever thought that was a good idea, must work in an echo chamber full of yes men.

Kaby Lake-X is a massively popular processor. It even has it's own thread on these forums dedicated to it. Would a "totally brain-dead product" have it's own thread? I think not.

 

[beginner99]

An i3 or Pentium is enough for the usage needs of the vast majority of users around the world, imho.

That is a major problem. The scissors to what a basic user needs and what an enthusiast wants is getting bigger and bigger. 10-20 years ago everyone was better of with a faster CPU. This doesn't hold true anymore. Your granny using a pc for some office and email can work just fine with a celeron or even atom cpu. There would be little to no benefit to get a 8700k. This poses a serious problem. Either enthusiasts get too little or mainstream too much performance or you need a larger more complex product portfolio with 3 different dies which is costly.

Intel's margins haven't improved at a corporate level since the release of Sandy Bridge.

In fact, Intel's gross margin percentage is down since 2011.

Tell me again how AMD's absence from this market has allowed Intel to increase its margins?

As has been said this is despite wasting billions on stupid purchases and losing billions in the mobile sector while delivering product in a declining market. Keeping 60%+ margins in such an environment is actually amazing. With a competing AMD, they would have gone down.

And to stop you right now, the could remain linear regardless of Ryzen because Intel has in fact stopped wasting money in mobile sector. We can already see that in the most recent upward trend.

 

[mikk]

https://www.chiphell.com/thread-1765892-1-1.html

 

[nvgpu]

Looks like the leaks Sweepr gathered was spot on.

Can't wait for the launch on Monday Aug 21.

 

[Sweepr]

Massively improved MT performance across the board. Interesting that Core i5-8400 delivers biggest generational improvement, this shows how much attention Intel put into the mainstream offerings. With 6 Skylake cores at aggressive clocks and possibly a <$200 price tag it will be a killer choice for a mainstream gaming build.

 

[LTC8K6]

Looks like the leaks Sweepr gathered was spot on.

Can't wait for the launch on Monday Aug 21.

It's fuzzy, but the 8700K base clock does not look like 3.7
Looks either 3.6 or 3.8?

 

[gx_saurav]

Very good line up by Intel IMO. If this is priced well, then for gamers Intel is the way to go forward.

I now wonder where does something like Core i7 7700/7700k fit now? Is it little less than new Core i5 in terms of performance?

 

[jpiniero]

Massively improved MT performance across the board. Interesting that Core i5-8400 delivers biggest generational improvement, this shows how much attention Intel put into the mainstream offerings. With 6 Skylake cores at aggressive clocks and possibly a <$200 price tag it will be a killer choice for a mainstream gaming build.

That's a little unfair because the 7400 has much lower clocks compared to the 7500 or 7600.

 

[LTC8K6]

The unlocked 8350K could be an interesting overclocker.

We would simultaneously have 7740X, 7700K, and 8350K, all 4C/4T and unlocked. The 7740X probably has a memory clock advantage.

 

[R0H1T]

Very good line up by Intel IMO. If this is priced well, then for gamers Intel is the way to go forward.

I now wonder where does something like Core i7 7700/7700k fit now? Is it little less than new Core i5 in terms of performance?

A locked i5 ~ no, otherwise yes since 7700k can be OCed close to 5GHz & that'll beat any locked i5 handily. The locked i7 7700 is such a bad buy right now.