Friday, July 27th 2018

Intel Stuck with 14nm Processors Till Holiday 2019

Wrap your head around this: at some point in 2019, AMD will be selling 7 nm processors while Intel sells 14 nm processors. That how grim Intel's 10 nanometer silicon fabrication process development is looking. In the Q&A session of its Q2-2018 Earnings Call, Intel stated that the first products based on its 10 nm process will arrive only by Holiday 2019, making 14 nm micro-architectures hold the fort for not just the rest of 2018, but also most of 2019. In the client-segment, Intel is on the verge of launching its 9th generation Core "Whiskey Lake" processor family, its 5th micro-architecture on the 14 nm node after "Broadwell," "Skylake," "Kaby Lake," and "Coffee Lake."

It's likely that "Whiskey Lake" will take Intel into 2019 after the company establishes performance leadership over 12 nm AMD "Pinnacle Ridge" with a new round of core-count increases. Intel is also squeezing out competitiveness in its HEDT segment by launching new 20-core and 22-core LGA2066 processors; and a new platform with up to 28 cores and broader memory interface. AMD, meanwhile, hopes to have the first 7 nm EPYC processors out by late-2018. Client-segment products based on its architecture, however, will follow the roll-out of these enterprise parts. We could see a point in 2019 when AMD launches its 7 nm 3rd generation Ryzen processors in the absence of competing 10 nm Core processors from Intel. Posted below is an Intel slide from 2013, when the company was expecting 10 nm rollout by 2015. That's how much its plans have derailed.
Add your own comment

72 Comments on Intel Stuck with 14nm Processors Till Holiday 2019

#51
Vya Domus
StrayKAT
I think they understand timing their products better than anyone.
And yet , their struggle with 10nm shows that wasn't the case , not in this particular situation. I think you are smart enough to realize people's doubtfulness of their abilities doesn't come out of nowhere.

Also I don't think people realize the magnitude of these "simple delays". They pour mountains of cash into these things and they plan products for years to come on their basis , that's a big deal. AMD, as poor as their track record has been with new nodes they have never screwed up this much.
Posted on Reply
#52
StrayKAT
Vya Domus
And yet , their struggle with 10nm shows that wasn't the case , not in this particular situation. I think you are smart enough to realize people's doubtfulness of their abilities doesn't come out of nowhere.

Also I don't think people realize the magnitude of these "simple delays". They pour mountains of cash into these things and they plan products for years to come on their basis , that's a big deal. AMD, as poor as their track record has been with new nodes they have never screwed up this much.
I'm not trying to downplay it. I just think the gloating is silly. Not just on this site. It's everywhere. People are somehow forgetting who they're talking about here.

Although, I'd probably laugh more about it, if it was a leak/and they were trying to hide this.
Posted on Reply
#53
Vya Domus
StrayKAT
People are somehow forgetting who they're talking about here.
Well , I don't , they are the ones who designed things like Netburst and offered bribes in the range of billions to OEMs so they can make sure their inferior products are being bought.

Sorry but Intel's track record is full of questionable products and strategies. If you want to say that people forget who they are in terms of their ability to remain market leaders, sure go ahead but as to anything else ? Nah , their past is littered with costly failures , ex. mobile SoCs.

Also, they predicted P68 based CPUs to hit 10 Ghz by 2011 ... yeah.
Posted on Reply
#54
StrayKAT
Vya Domus
Well , I don't , they are the ones who designed things like Netburst and offered bribes in the range of billions to OEMs so they can make sure their inferior products are being bought.

Sorry but Intel's track record is full of questionable products and strategies. If you want to say that people forget who they are in terms of their ability to remain market leaders, sure go ahead but as to anything else ? Nah , their past is littered with costly failures , ex. mobile SoCs.
That's just capitalism. I hope you're not hoping for salvation or something from AMD. It's always/eventually going to be the same... unless you break capitalism itself. But I digress.

I'm just saying don't forget who they are in terms of size/resources. This is like watching people get stoked when another team scores a touchdown on the New England Patriots (or insert the equivalent here). Where they suddenly treat that perpetual behemoth like an amateur.... only to get disappointed moments later.
Posted on Reply
#55
Vya Domus
StrayKAT
That's just capitalism.
This is a matter of technical aspects though and I feel like you confused the two. Being doubtful of that is perfectly valid.
Posted on Reply
#56
StrayKAT
Vya Domus
This is a matter of technical aspects though and I feel like you confused the two. Being doubtful of that is perfectly valid.
Admittedly, I am confused.. since you mentioned the underhanded marketing/distributor tactics (which I don't like either).

edit:

Sidenote -- Despite naming differences, apparently Intel's 10nm process is similar AMD's 7nm. Can someone explain/confirm that?

Either way, they'll probably hit the same roadblock eventually at 5nm.
Posted on Reply
#57
Caqde
StrayKAT
Admittedly, I am confused.. since you mentioned the underhanded marketing/distributor tactics (which I don't like either).

edit:

Sidenote -- Despite naming differences, apparently Intel's 10nm process is similar AMD's 7nm. Can someone explain/confirm that?

Either way, they'll probably hit the same roadblock eventually at 5nm.
Here is a comparison of Global Foundries 7nm and Intel's 10nm -> https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

Here is another article containing a chart with comparison's of all the leading 7nm/10nm players in the market.
https://www.semiwiki.com/forum/content/7602-semicon-west-intel-10nm-gf-7nm-update.html
Posted on Reply
#58
StrayKAT
Caqde
Here is a comparison of Global Foundries 7nm and Intel's 10nm -> https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

Here is another article containing a chart with comparison's of all the leading 7nm/10nm players in the market.
Thanks.. briefly read through the first link, but I think I get the gist of it (not that I understand the whole process). Your second link doesn't seem to be showing up.
Posted on Reply
#59
Caqde
StrayKAT
Thanks.. briefly read through the first link, but I think I get the gist of it (not that I understand the whole process). Your second link doesn't seem to be showing up.
Sorry I edited my post with the proper link.
Posted on Reply
#60
Smartcom5
Vayra86
[…] They are going to have to play the pricing game.
Which they can only lose.
They just can't compete nor win against AMD's CCX-stroke of genius and their Magic Glue™

dorsetknob
An unmentioned consideration might be the urgent need to Redesign the CPU to avoid those all-ready Discovered Back-doors
like Spectre and Meltdown ( AND THOSE THAT INTEL HAVE NOT DISCLOSED).
WE ALL KNOW that Intel have to rework upcoming CPU's to mitigate those flaws found in all Current designs
… which – if they re-establish that physical barrier (gimme a rope here …) between given address-spaces (kernel-/user-space), will hurt their Chips performance really badly – if they truly revert that change they did back then.

efikkan
[…]
Remember that Ice Lake is a major architecture unlike Zen2, so we should expect some IPC gains to offset the lower clock speed. …
I never understood why everyone keeps saying that though?
For me, that ever mentioned Ice Lake as well as Tiger Lake ain't going to be any stellar either – as they most definitely¹ are just going to be another (though, refined) Skylake-designs (being downscaled for 10nm) and newer reincarnations of respins and expansions of (at that time) up-to-date Core-I-variants – that's what they're telling by their names alone, or do they?

Intel always gives every architecture unique names after given schemes like city names (Core → Kentsfield, Yorkfield, Clovertown, Harpertown, Tigerton, Dunnington), rivers (Nehalem → Nehalem, Westmere) or atm lakes (Skylake → Skylake, Kaby Lake, Coffe Lake, Ice Lake, Tiger Lake).

I firmly believe that Ice Lake and Tiger Lake pretty sure ain't going to be any new µArch but refined Skylake-designs.
¹ If Intel will be going to release a ground-braking new µArch and design (as it was rumoured at the launch of Ryzen) with their approach in getting rid of most x86-backward compatibility (whereas no-one really knows by now if it delivers any greater gains in IPC anyway …), they surely won't label it as some Skylake-spin-off but would have gave it a truly unique new name-scheme. That's for sure, isn't it?

Ice Lake will be de facto just another 10nm-refresh and a derivative of Skylake, just like Cannon Lake is was supposed to be – with what Intel would have even been come through with, if AMD wouldn't have been risen Ryzen again …

Speaking of 'woods of trees', anyone knows what has happened to Cannonball?! It never seemed to appear ever again!


In this sense

Smartcom
Posted on Reply
#61
hat
Enthusiast
Intel would have kept right on cruising if it weren't for Spectre/Meltdown, and AMD. Between the two, they're probably scrambling at red alert trying to get their shit together...
Posted on Reply
#62
efikkan
Smartcom5
For me, that ever mentioned Ice Lake as well as Tiger Lake ain't going to be any stellar either – as they most definitely¹ are just going to be another (though, refined) Skylake-designs (being downscaled for 10nm) and newer reincarnations of respins and expansions of (at that time) up-to-date Core-I-variants – that's what they're telling by their names alone, or do they?

Intel always gives every architecture unique names after given schemes like city names (Core → Kentsfield, Yorkfield, Clovertown, Harpertown, Tigerton, Dunnington), rivers (Nehalem → Nehalem, Westmere) or atm lakes (Skylake → Skylake, Kaby Lake, Coffe Lake, Ice Lake, Tiger Lake).

I firmly believe that Ice Lake and Tiger Lake pretty sure ain't going to be any new µArch but refined Skylake-designs.
So, your whole argument rests on the names of the generations?
Ice Lake was going to be their next architecture according to their old tick-tock plan, then Sapphire Rapids the next architecture after that(~2021).

Smartcom5
If Intel will be going to release a ground-braking new µArch and design (as it was rumoured at the launch of Ryzen) with their approach in getting rid of most x86-backward compatibility (whereas no-one really knows by now if it delivers any greater gains in IPC anyway …), they surely won't label it as some Skylake-spin-off but would have gave it a truly unique new name-scheme. That's for sure, isn't it?
You are confused about CPU architectures and ISAs(Instruction Set Architecture). Intel is not going to abandon x86 anytime the next ten years. At the moment, both Skylake and Zen families are RISC-like implementations of x86, and the legacy x86 instructions is not a major problem. x86 is still far more performing than any of the "true" RISC variants in existence.

Ice Lake is going to be another CPU architecture, still implementing x86 ISA, continuing to improve the way Skylake, Haswell, etc. did before it.
Posted on Reply
#63
R0H1T
efikkan
You are confused about CPU architectures and ISAs(Instruction Set Architecture). Intel is not going to abandon x86 anytime the next ten years. At the moment, both Skylake and Zen families are RISC-like implementations of x86, and the legacy x86 instructions is not a major problem. x86 is still far more performing than any of the "true" RISC variants in existence.

Ice Lake is going to be another CPU architecture, still implementing x86 ISA, continuing to improve the way Skylake, Haswell, etc. did before it.
It's not going to be anything significantly different, from lets say Broadwell->Skylake. I'm not sure where you got the impression that major uarch changes (tock) improve the IPC tremendously, they don't! The only thing ICL has going for it is AVX512 for MSDT, servers - I'm not sure Intel even has a joker up their sleeves.

The last major uarch changes were conroe, then Nehalem. SB increases the IPC a lot but I don't remember anything radical being introduced, except AVX.
Posted on Reply
#64
efikkan
R0H1T
It's not going to be anything significantly different, from lets say Broadwell->Skylake. I'm not sure where you got the impression that major uarch changes (tock) improve the IPC tremendously, they don't! The only thing ICL has going for it is AVX512 for MSDT, servers - I'm not sure Intel even has a joker up their sleeves.

The last major uarch changes were conroe, then Nehalem. SB increases the IPC a lot but I don't remember anything radical being introduced, except AVX.
Your argument centers around "major changes". I don't know what you mean by that, but I'll explain what I do.

You don't have to add new instructions to increase performance. Of recent architectures (Sandy Bridge, Haswell and Skylake), Sandy Bridge was the largest improvement, and it was not due to AVX. Sandy Bridge did improve the execution ports and the prefetcher. Haswell and Skylake did similar improvements, just slightly smaller. These are improvements which boost IPC, and have nothing to do with AVX, or special acceleration for AES, compression and video, which these CPUs also have.

Ice Lake will be a new architecture, it will redesign or extend most parts of the CPU design, unlike Ivy Bridge and Broadwell, which did isolated improvements in addition to a node shrink. Ice Lake will rebalance and calibrate the entire front-end and pipeline, including memory controller and cache. It may also add more computational resources like Zen did, but we have no details on that yet. There are also a few new instructions in Ice Lake, one of them is related to saving cycles when copying memory.
Posted on Reply
#65
R0H1T
efikkan
Your argument centers around "major changes". I don't know what you mean by that, but I'll explain what I do.

You don't have to add new instructions to increase performance. Of recent architectures (Sandy Bridge, Haswell and Skylake), Sandy Bridge was the largest improvement, and it was not due to AVX. Sandy Bridge did improve the execution ports and the prefetcher. Haswell and Skylake did similar improvements, just slightly smaller. These are improvements which boost IPC, and have nothing to do with AVX, or special acceleration for AES, compression and video, which these CPUs also have.

Ice Lake will be a new architecture, it will redesign or extend most parts of the CPU design, unlike Ivy Bridge and Broadwell, which did isolated improvements in addition to a node shrink. Ice Lake will rebalance and calibrate the entire front-end and pipeline, including memory controller and cache. It may also add more computational resources like Zen did, but we have no details on that yet. There are also a few new instructions in Ice Lake, one of them is related to saving cycles when copying memory.
Tock, part of Intel's Tick Tock which is dead atm.

I didn't say it was due to AVX but adding AVX (also QuickSync) was perhaps the biggest change from Nehalem. Apparetly I was wrong ~ https://www.anandtech.com/print/3922/intels-sandy-bridge-architecture-exposed

The gains will be minimal, unless Intel has found a serious bottleneck that limited core's performance. SB->HSW was a small leap, HSW->SKL a smaller leap & I expect SKL->ICL to be even less wrt IPC increase. This of course ignores the spectre & meltdown hardwired fixes, that may skew the final numbers in ICL favor.

You mean Fast Short REP MOV ~ we'll see how it works in real world, also if there's any potential spectre vulnerabilities.
Posted on Reply
#66
efahl
If you had the choice between a race car with steel (omg how ancient!) body work and carbon fiber, but the steel one was faster, would you still buy the carbon fiber one? Implementation technology is purely secondary, I buy for performance.
Posted on Reply
#67
EsaT
StrayKAT
I'm not trying to downplay it. I just think the gloating is silly. Not just on this site. It's everywhere. People are somehow forgetting who they're talking about here.
While people laughing at AMD's "Fail"dozer wasn't silly?
I think just for fairness it's right that Intel is now getting laughed at.

I mean compared to AMD they're huge, with huge amount more resources and talent.
And despite of all that they kept consumers at four cores for decade in high end and two as standard.
(making sure game developers didn't have too much of incentive to go looking for new ways to use extra threads)
Also 10nm mass production seems to be now at best four years late from original plans.
And it's not even sure if it's going to be such as originally advertised.
That just looks more and more like Olkiluoto 3 nuclear power plant.
Supposed to now start operating decade behind original plan... Assuming it doesn't keep status quo of still more delays.
(world's second most expensive building in construction costs)
Posted on Reply
#68
Vlada011
Maybe Intel can't do nothing 2 years after they launch i9-9900K, but if his price replace i7-8700K we talk about seriuos and very usefull processor.
After customers OC him on 5.0-5.2GHz for all cores on custom loop and that will be possible if Intel avoid thermal paste that will be best processor mixed for gaming/surfing/little video editing./little youtube streaming, etc...
He will not be as Threadripper but he will finish jobs even for rendering and video editing very good.

Example if some Intel fan want to replace his computers and don't know what to use,...
1. Build small ITX/Mini ITX RIG with Seasonic 600W Titanium Fanless and i9-9900K for gaming and UHD BluRay 4K and fun.
2. Wait Intel X599 to build multi core monster

I have impression that AMD will be better performance for price only if you 24h use all cores 16-32
And Intel for sure could launch 10nm only 18 months after last 14nm CPU.
Maybe is 14nm but will be beast, and his total picture depend only from price.
Asking 500-550$ they are in problems and for 8 months they will need to drop price on 4350-400$.
But if they ask 400$ immediately to be honest I would rather go on i9-9900K than any Ryzen 2, even if I need to pay clear 100 more.

I don't belive AMD have so better and dominat core and his 4.2GHz give you better impression in different applications than Intel 5.0GHz, same number of cores.
I hope only reviews will show good chance for overclocking all cores on higher speed without surprise with VRM and powerconsumption, thermal paste, etc...
Posted on Reply
#69
trparky
ShockG
Look at the TDP for the 8-core Core i7 9900K. It's still 95W with two more cores and a higher clock speed
You seem to forget that that's only the TDP at base clocks, when it speeds up to boost speeds the TDP goes through the roof.
Posted on Reply
#70
Vlada011
But TDP will not go over 120W in Turbo specifications.
Even less, I belive arround 110W.
Posted on Reply
#71
EsaT
Vlada011
But TDP will not go over 120W in Turbo specifications.
Even less, I belive arround 110W.
Already 8700K goes to 150W class power consumption if BIOS isn't limiting boost clocks because of temps/to comply with TDP limit:
https://www.tomshardware.com/reviews/intel-coffee-lake-i7-8700k-cpu,5252-12.html
So that imaginary all 5+ GHz octa core Intel of yours will be more like 200W power consumption CPU when put under full load.
More reasonable power consumption will have to wait for Intel's 10nm.
Posted on Reply
#72
hat
Enthusiast
This is all starting to sound similar to Netburst... high clocks, heat, and power consumption... only it's actually a little faster this time around. AMD just needs a nice little shove with Zen 2 and we're back to Pentium vs Athlon.
Posted on Reply
Add your own comment