Intel Might Drop 10nm Node for Desktop Processors, 14nm Until 2022?

[BiG StroOnZ]

Quote

Yesterday evening the news reached the web that Intel could be skipping their troublesome 10nm node altogether, for desktop processors. And that would mean the next-gen Ice Lake processors.

 

Website HardwareLuxx.de reports that Intel will not manufacture any 10nm processors for desktops, focusing on 7nm chips for the specific segment, set for launch in two years. This means that Intel will confine its 10 nm microarchitectures, "Ice Lake" and "Tiger Lake" to only the mobile platform, while the desktop platform will see derivatives of the 14nm "Skylake" until 2022. Essentially, the report claims that Intel will not launch the "Tiger Lake" and "Alder Lake" chips based on a 10nm process, at least at their previously scheduled time frames. Intel has been challenged to migrate from the 14nm to the more advanced 10nm manufacturing process, at least for desktop chips, which have high clock rates. The company has announced the 10nm Ice Lake processors for mobile devices.

 

Meanwhile, Intel was very quick to out some words on this to Toms hardware: desktop processors based on the 10nm silicon fabrication node are still on the company's roadmap:

"We continue to make great progress on 10 nm, and our current roadmap of 10 nm products includes desktop," the company said in its communication.

 

Source: https://www.guru3d.com/news-story/intel-might-drop-10nm-node-for-desktop-processors.html

Source 2: https://www.tweaktown.com/news/68127/intel-cancels-10nm-desktop-14nm-hold-until-2022/index.html

 

All I know is this; the idea that 14nm with "some tweaks" will be enough for Intel to compete with AMD until 2022, does not seem very convincing Intel. Let's all hope that their quick response claiming that their 10nm node is still currently on track, is actually factual, and not Intel simply trying to draw attention away from the topic. 

[The Benjamins]

How on earth are they going to compete?

 

AMDs 7nm+ zen 3 CPUs will hit mid 2020 and 5nm is on track for 2021.

 

By the time Intel gets there 7nm they will be up against 5nm+

[WereCatf]

If Intel can only manage to make proper desktop-CPUs from 2022 onwards, that's.....going to really suck for some people and be an absolutely fricking fantastic thing for AMD. I am kind of hoping for this to actually be true.

[Dedayog]

Just now, The Benjamins said:

How on earth are they going to compete?

 

AMDs 7nm+ zen 3 CPUs will hit mid 2020 and 5nm is on track for 2021.

 

By the time Intel gets there 7nm they will be up against 5nm+

Intel's 14nm chips are doing very well, even against 7nm.  

 

I'm curious to see what benefits 7nm will give Intel, if they can continue to refine these processes so well or if there is a diminishing returns as we get smaller on current materials.

 

AMD and Intel are going to bump up against a hard limit soon, making them basically equal.  Who finds the next perfect material first?

[greenhorn]

If this is true, then AMD can comfortably make even larger inroads into Intel's desktop CPU market share. I just hope they don't "pull an Intel", e.g. deliberately stall their progress until the competition can keep up again (which can make sense financially, as the gaps to previous products that are still on sale will be small enough to still be able to sell them at a decent price).

[Princess Luna]

4 minutes ago, The Benjamins said:

How on earth are they going to compete?
By the time Intel gets there 7nm they will be up against 5nm+

The same way how 14nm++ is still on pair with AMD's 7nm, like it or not but Intel architectures are often denser and higher performer, specially with all the refining they put it through.

[The Benjamins]

1 hour ago, jstudrawa said:

Intel's 14nm chips are doing very well, even against 7nm.  

 

I'm curious to see what benefits 7nm will give Intel, if they can continue to refine these processes so well or if there is a diminishing returns as we get smaller on current materials.

 

AMD and Intel are going to bump up against a hard limit soon, making them basically equal.  Who finds the next perfect material first?

 

1 hour ago, Princess Luna said:

The same way how 14nm++ is still on pair with AMD's 7nm, like it or not but Intel architectures are often denser and higher performer, specially with all the refining they put it through.

 

 

I don't think there 14nm is doing well, its a far cry from that, besides a 5-10% gaming advantage on the very high end intels CPUs use 50% more power with 50% less cores to get nearly the same per core perf and %50 less all core perf.

 

Its not even remotely close.

 

EDIT:

 

its more like 50% more cores (30-40% more multi perf) at 10% less power, with nearly the same single core perf

 

[SkyHound0202]

In Intel's defense, using a mature node that is already in mass production is less risky than an experimental smaller node, and it makes economical sense.

 

Considering the case of Broadwell (14-nm Haswell) and Cannon Lake (10-nm Skylake), neither of which reached the same frequency as previous generation, which nullified their process advantage at a comparable IPC. Intel is betting on ensuring the longevity of 14-nm process so that it can "buy itself time" before 10-nm is ready.

 

But since Intel can't even cope with the 14-nm shortage, I don't have much faith in Intel's 10-nm development.

[The Benjamins]

1 minute ago, VegetableStu said:

sounds about right, the definitions are similar o_o

 

remember intel 10nm ≈ TSMC 7nm

https://fuse.wikichip.org/news/525/iedm-2017-isscc-2018-intels-10nm-switching-to-cobalt-interconnects/6/

I thought intels 7nm is about the same as TSMCs 5nm not 5nm+

[Princess Luna]

9 minutes ago, The Benjamins said:

intels CPUs use 50% more power with 50% less cores

Spoiler

¯\_(ツ)_/¯

 

I get it you are the usual AMD fan but how about don't just go shooting random numbers just to try making a point your brand is better.

[Taf the Ghost]

The HardwareLuxx image isn't quite correct. Comet Lake/Rocket Lake have replaced the Icelake/Tigerlake sequence on Desktop. (Ice/Tiger still mobile.)  Tigerlake will launch in 2020 (early silicon is already running around). Alder Lake is the only one up for debate right now on whether it might make Desktop. That'll be 10nm++ designs, but it's always been likely Intel would simply not release another Desktop revision until Meteor Lake is ready to launch in volume in 2022. Even if really late in the year.

[Taf the Ghost]

13 minutes ago, VegetableStu said:

sounds about right, the definitions are similar o_o

 

remember intel 10nm ≈ TSMC 7nm

https://fuse.wikichip.org/news/525/iedm-2017-isscc-2018-intels-10nm-switching-to-cobalt-interconnects/6/

Except it really isn't. Intel had to dial back 10nm really far to get it out the door. Far enough that no one seems to know what the node even is anymore. It's a shrink from their 14nm, but TSMC/Samsung should have a significant density advantage in actual shipping silicon.

 

Edit: It'd probably be safe to call Intel's 10nm as equal to something like the Foundry 9nm. Denser than Foundry 10nm but not as dense as Foundry 7nm.

[The Benjamins]

12 minutes ago, Princess Luna said:

  Hide contents

¯\_(ツ)_/¯

I might of went a bit over board but you only showed half of the story, with no source link. what is the 3900x vs 9900k preformance in specviewperf?

 

UPDATE: SPECviewperf is a GPU test so I don't get why its used?

https://www.hardwarezone.com.sg/feature-amd-ryzen-9-3900x-review-ryzen-7-3700x-intel-core-i9-9900k-gaming-cpu-performance-specs-features/cpu-rendering-encoding-benchmarks

 

Spoiler

 

 

 

https://www.techspot.com/review/1869-amd-ryzen-3900x-ryzen-3700x/

Spoiler

 

[GoldenLag]

So how is intel going to make GPUs on 10nm if they cant create large enough dies for CPUs?

 

Perhaps 10nm is better at big dies and lower clocks and voltages?

[porina]

Some existing thoughts from my post on this topic yesterday.

 

 

Since then it is interesting to see reactions both to the original claim and Intel's apparent denial of it. The question people are poking into now is what does Intel mean by making a desktop part on 10nm. Some speculate they will only repurpose the 10nm server parts for HEDT, which I think is likely. Another theory is that Intel might only produce lower end desktop parts at 10nm, smaller dies not dissimilar to the mobile ones. Again, a possibility but I don't feel this adds value unless they focus only on more power efficient parts for specific use cases.

 

 

[Taf the Ghost]

Just now, GoldenLag said:

So how is intel going to make GPUs on 10nm if they cant create large enough dies for CPUs?

 

Perhaps 10nm is better at big dies and lower clocks and voltages?

The node yields. It's just the original yields were about .5%. Not 5% but .5% and they couldn't yield a iGPU at all. They've dialed back a lot of the technical advances and while they can show off silicon at the density they've claimed, reality is that it isn't that much more dense than 14nm and clocks a lot worse. And that was just to get the node out the door. Yields are still going to be terrible and a single server die likely is yielding maybe 20% per wafer. Maybe. Full die SKUs from Icelake Server parts are going to be extremely expensive and really only fit in a niche roll. 

 

What's more fun is Intel's 7nm is their first EUV node. And they're going to be doing a lot of layers with EUV. Which means the whole "7nm in 2022!" mantra may not happen. Even with two full years to get there. It will be some large, generational leaps, but only because their design teams haven't had a node to launch anything new on in ages.

[Taf the Ghost]

1 minute ago, porina said:

Some existing thoughts from my post on this topic yesterday.

 

 

Since then it is interesting to see reactions both to the original claim and Intel's apparent denial of it. The question people are poking into now is what does Intel mean by making a desktop part on 10nm. Some speculate they will only repurpose the 10nm server parts for HEDT, which I think is likely. Another theory is that Intel might only produce lower end desktop parts at 10nm, smaller dies not dissimilar to the mobile ones. Again, a possibility but I don't feel this adds value unless they focus only on more power efficient parts for specific use cases.

 

 

They could launch some Pentiums with the defective mobile dies and call it "10nm on Desktop". They launched 10nm Cannonlake the last week of December 2017 just to say 10nm was shipping for revenue. 

[BuckGup]

24 minutes ago, Princess Luna said:

The same way how 14nm++ is still on pair with AMD's 7nm, like it or not but Intel architectures are often denser and higher performer, specially with all the refining they put it through.

Yes and no. AMD jumped 15% IPC in one generation. Intel has jumped 9% in the past 6 years

[S w a t s o n]

You guys remember the 5775C and the like one person that actually bought one, and the like 2 reviews it got (mostly for the igpu)? If we see 10nm desktop it's going to be a repeat of that, I fully agree with Charlie here. Go ahead, think back to 5th gen i5/i7....intel would rather you not and just remember good ol skylake aka 6th gen.

[GoldenLag]

8 minutes ago, Taf the Ghost said:

yielding maybe 20% per wafer.

The 28 core atm is reportadly 35% good one (rumour and about 6 months old now)  Dunno if that was the 8180 or just 28 core

 

 

If the density isnt much better. I find it unlikely for a 28 core to get that great yields

 

There is offcourse salvage and lower clocks. Kinda like how Jaguar chips had extra CUs for better salvage yields.

Edited October 15, 2019 by GoldenLag
Fairly certain its 35 now that i think about it

[porina]

Just now, BuckGup said:

Yes and no. AMD jumped 15% IPC in one generation. Intel has jumped 9% in the past 6 years

Intel have made a similar jump... but you can only get it on Ice Lake mobile. Sunny Cove compared to Old-lake is a comparable improvement shown by Zen 2 to Zen(+). Another question being thrown around is if Intel really can't shift volume to 10nm, would they bring Sunny Cove (or successor) to 14nm? It wont gain from the process improvement but will at least keep the architecture up to date.

[The Benjamins]

1 minute ago, porina said:

Intel have made a similar jump... but you can only get it on Ice Lake mobile. Sunny Cove compared to Old-lake is a comparable improvement shown by Zen 2 to Zen(+). Another question being thrown around is if Intel really can't shift volume to 10nm, would they bring Sunny Cove (or successor) to 14nm? It wont gain from the process improvement but will at least keep the architecture up to date.

intel might of gained 18% ipc on ice lake but they also lost about 10-15% in clocks.

[Taf the Ghost]

2 minutes ago, GoldenLag said:

The 28 core atm is reportadly 25-35% good one (rumour and about 6 months old now)  Dunno if that was the 8180 or just 28 core

 

 

If the density isnt much better. I find it unlikely for a 28 core to get that great yields

 

There is offcourse salvage and lower clocks. Kinda like how Jaguar chips had extra CUs for better salvage yields.

We still don't know what size of dies Icelake server will launch with. Right now Cooper Lake and the next platform (Eagle Stream?) is in the validation. We haven't seen anything above 8 cores on Icelake in any of the filings or testing results, to my knowledge. 

[porina]

3 minutes ago, S w a t s o n said:

 

You guys remember the 5775C and the like one person that actually bought one, and the like 2 reviews it got (mostly for the igpu)? If we see 10nm desktop it's going to be a repeat of that, I fully agree with  Charlie here. Go ahead, think back to 5th gen i5/i7....intel would rather you not and just remember good ol skylake aka 6th gen.

*raises hand* I bought one. Well, I got the 5675C when it was current, and only more recently did I manage to find a cheap 5775C to go with it. The initial transition to 14nm was also a problem for Intel, very different from the 14nm we have today. Having said that, the 5675C and 5775C were odd balls in that they were more or less the mobile parts done on desktop. They had the L4 cache of Iris Pro, and this made them very interesting CPUs. Skylake was better on architecture, but that cache meant Broadwell desktop parts (not HEDT) punched far above their weight in memory intensive compute applications.

 

I actually would like to see that. Take Ice Lake, put it on a desktop package and let me overclock it. 

[porina]

2 minutes ago, The Benjamins said:

intel might of gained 18% ipc on ice lake but they also lost about 10-15% in clocks.

Yes, but we have to consider architecture and process as two parts. Intel 10nm sucks, but what would that same architecture do on 14nm?