Guess the 2019 Intel desktop product!

[jpiniero]

Allright, let's play a game. Given all the problems Intel is having with 10 nm, it sure looks like there's going to be some sort of uncertainty about what Intel might do.

 

[moinmoin]

Something called 9700K.

 

[dark zero]

CannonLake but only for Core i7.

 

[ksec]

Just playing what if....

Intel has very decent results, so they decide to announce the 10nm delay. If they fix it fast, there will be 10nm Mobile product this year. And possibly 10nm Desktop next year. If they don't, my guess is 10nm will be Mobile and DC next year, Desktop will be whatever "lake" they can come up with.

 

[jpiniero]

Just playing what if....

Intel has very decent results, so they decide to announce the 10nm delay. If they fix it fast, there will be 10nm Mobile product this year. And possibly 10nm Desktop next year. If they don't, my guess is 10nm will be Mobile and DC next year, Desktop will be whatever "lake" they can come up with.

BK said they aren't going to HVM until 2019, and you have to add at least 3 months from the point they go HVM before product is available. And they don't know when HVM will happen because of the yield, it could be first half or second half.

 

[FIVR]

Most likely it will be a "14nm +++" ring bus 8 core along with a refresh of the 8700k and 8700. Intel will claim revolutionary new performance on offer with 33% more cores and they will show graphs that make it clear that intel is giving you more cores now than they ever have before.



Whether this marketing offensive + an 8 core actually works to take back marketshare from Ryzen... I have my doubts.

 

[FIVR]

You really have to wonder whether intel would be better off persevering their margins by continuing to sell the same dies for the same (high) prices, while cutting payroll simultaneously. That way they could even have "margin expansion" plus net income growth while still losing marketshare.

 

[Mockingbird]

It's obviously going to be Coffee Lake or the next 14 nm refresh.

Intel's 10 nm is no where near (high volume manufacturing) HVM ready, and even after it is HVM ready, Intel will need a while to stockpile.

 

[Arachnotronic]

You really have to wonder whether intel would be better off persevering their margins by continuing to sell the same dies for the same (high) prices, while cutting payroll simultaneously. That way they could even have "margin expansion" plus net income growth while still losing marketshare.

Downward spiral, you cut too deeply and you ruin your future product pipeline, which ultimately causes you to see substantially lower profits.

There's a reason that chip companies are spending more each year -- this stuff gets harder and harder and to just tread water you need to spend more. Cutting costs is the last thing you want to do in an increasingly competitive environment.

 

[cmdrdredd]

I'm thinking something like an 8700k with 8 cores instead of 6. Nothing revolutionary, no huge jump in performance, just cheaper variations of their 8core CPUs for mainstream.

 

[beginner99]

Another 14nm "lake" 8-core. They aren't even sure when in 2019 10nm will be ready. If they aren't sure of that, how can they be sure of 2019? Plus we already know first 10nm iteration is actually worse for HPC, lower clocks and 10nm+ more or less matching current 14nm++. Anyway since output will probably still be somewhat limited, they will go mobile first with 10 nm. Since 8700k was released in fall 2017, the will need something new latest in H1 2019. Since there is no guarantee they can manage this, they for sure started a 8-core 14nm project. Whether it's easier to expand 6-core CFL or backport icelake to 14 nm, I can't say but I'm pretty certain we will see another 14nm product in high end desktop.
Ultimately the high clocking 14nm++ and coffelake are actually a issue for intel here because they risk the chance of ST performance regression if they jump to 10 nm too early.

 

[LTC8K6]

Full lineup of refreshed coffee lake chips. 14nm is still holding it's own. All it needs is an 8 core version and a refresh to keep Intel in the lead.
Probably will have refreshed IGP as well.

 

[NTMBK]

Never mind desktop, I want a mobile processor that actually supports LPDDR4.

 

[LTC8K6]

Never mind desktop, I want a mobile processor that actually supports LPDDR4.

Don't we have LPDDR4X now at .6V?

 

[raghu78]

Another 14nm "lake" 8-core. They aren't even sure when in 2019 10nm will be ready. If they aren't sure of that, how can they be sure of 2019? Plus we already know first 10nm iteration is actually worse for HPC, lower clocks and 10nm+ more or less matching current 14nm++. Anyway since output will probably still be somewhat limited, they will go mobile first with 10 nm. Since 8700k was released in fall 2017, the will need something new latest in H1 2019. Since there is no guarantee they can manage this, they for sure started a 8-core 14nm project. Whether it's easier to expand 6-core CFL or backport icelake to 14 nm, I can't say but I'm pretty certain we will see another 14nm product in high end desktop.
Ultimately the high clocking 14nm++ and coffelake are actually a issue for intel here because they risk the chance of ST performance regression if they jump to 10 nm too early.

I think this is Intel's big problem other than the fact that Intel needs a 10+ process which works and yields reasonably well. CFL is a double edge sword. It helps Intel in keep a significant ST perf lead but also makes it that much harder for ICL to overtake CFL by a good margin (>5%). If clocks fall by 10% (both stock and max) while IPC goes by up 10-12% Intel is left with a negligible gain in ST perf. Moreover 14++ yields are going to be much higher than any 10nm related process from Intel for the next few years (atleast 2-3).

 

[Midwayman]

Can we hope for enthusiast chips with no igpu wasted space?

 

[raghu78]

Can we hope for enthusiast chips with no igpu wasted space?

I think Intel's hand will be forced by AMD once 7nm Zen 2 launches in 2019 and we could see such a CPU from Intel by 2020.

 

[PeterScott]

2019 desktop will be 8 Core 14nm++(+). Who knows/cares what it will be called. I didn't vote because no option fit that well.

I can't see any way that 10nm will be shipping HVM for desktops.

Given previous 10nm roadmaps, it was pretty clear, that first and maybe even second generation 10nm wouldn't reach desktop clock speeds, meaning even if 10nm has HVM shipping in 2019, it will be for mobile parts.

 

[mikk]

Icelake (as originally intended with 8 CPU cores/48 EU max)

Icelake GT2 features 64 EUs max. Also we have no clue for the core count on Icelake-S. We could only guess that the core count is at least 8 considering that Coffeelake already comes with 8C this year.

 

[turtile]

It's already confirmed that they will do another 14nm refresh called whiskey lake.

 

[mikk]

Whiskey Lake is for mobile not desktop.

 

[epsilon84]

Isn't the 8 core CFL supposed to launch this year along with the Z390 chipset?

I can understand that being Intel's flagship desktop part from late 2018 into 2019, but will Intel really ride 14nm into the sunset for the entirety of 2019?

 

[PeterScott]

Isn't the 8 core CFL supposed to launch this year along with the Z390 chipset?

I can understand that being Intel's flagship desktop part from late 2018 into 2019, but will Intel really ride 14nm into the sunset for the entirety of 2019?

If 10nm is not ready for desktop HVM, there isn't much else they can do. I doubt they will try to get TSMC to fab them chips.

 

[IntelUser2000]

You really have to wonder whether intel would be better off persevering their margins by continuing to sell the same dies for the same (high) prices, while cutting payroll simultaneously. That way they could even have "margin expansion" plus net income growth while still losing marketshare.

I find your sudden switch on the tone of your posts interesting.

It'll be better for Intel to do the opposite. Sacrifice some margins so they can devote more into R&D and future products.

I vote for Icelake (as originally intended with 8 CPU cores/64 EU max).

Can we hope for enthusiast chips with no igpu wasted space?

Another person that believes in the flawed theory making a dedicated enthusiast die would make sense.

 

[jpiniero]

Isn't the 8 core CFL supposed to launch this year along with the Z390 chipset?

It appears so.

I can understand that being Intel's flagship desktop part from late 2018 into 2019, but will Intel really ride 14nm into the sunset for the entirety of 2019?

Totally depends on whether they can get the yield up. And since Intel admits they aren't sure, hence the uncertainty.