Intel Alder Lake

[WingZero30]

Well it looks like the 'Lakes saga' does not finish after Tiger Lake.

The next micro-architecture after Tiger lake is purported to be Alder Lake.

Ashraf Eassa an intel share holder who regularly posts on The Motley Fool tweeted the following below:

https://twitter.com/TMFChipFool/status/936829260391620609

Sapphire Rapids was supposed to be the successor to Tiger lake but it may be that SR is server only whilst Alder Lake may be the on the mainstream platform; both succeeding Tiger lake on their respective platforms. It's not clear at this stage what process node would be used for Alder lake but Ashraf thinks that it may be 10nm+++ (lol). Only time will tell I guess.

https://twitter.com/TMFChipFool/status/936708804368064513


At this stage the information is too scarce to make any meaningful generalisation so a bucket of of salt is needed.

 

[muon]

Have Intel basically stopped doing major architecture changes now?

https://en.wikipedia.org/wiki/Tick-tock_model

I bet Icelake is Skylake rehashed.

 

[Perfect_Chaos]

Ice Lake is supposedly a decent performance jump is it not? I'd like a good upgrade from my 5820k at least within the next year. Not sure it's coming though. At least core count is going up.

 

[DragonQ]

Ice Lake is supposed to be a new architecture, but then so was Skylake and that was barely 5% faster than Broadwell in most cases, so I have zero expectation.

 

[opethdisciple]

I am so confused with Intel right now. It's as if they themself have no idea what's coming and when and just make it all up.

Probably why they've recruited Raja to the team.

 

[drunkenmaster]

They supposedly started on a ground up new architecture at the start of this year in response finally to Zen and it's unlikely to be ready for 3-4 years so 2020-21 and AMD too are working concurrently on something to replace Zen completely as well as Zen 2 and 3, with the replacement aimed at around 20-21 also.

Everything else is supposed to be core architecture and just iterations, though thanks to AMD we'll be getting small iterative performance improvements along with big core count increase. Intel simply has to get 8 cores into their mainstream platform on 10nm and if that is with a gpu or if they'll go the AMD route.. who knows. AMD is going to certainly bring a 6 core APU with 7nm Glofo, will they try and squeeze 8 core in, not sure. It looks like 6 core per CCX makes the most sense, higher end laptops, top notch low end gaming PCs and then 12 core cpus, 24 core threadripper, 48 core server.

Intel's been maximising profits rather than maximising R&D and taking bold steps with their CPUs for some time and 10nm is still struggling massively despite the lead they previously had. Intel's R&D lead seems to be dropping off, again lots of rumours putting it about failed R&D projects, losing good people and having money men rather than engineers directing where the spending goes.

I really don't think Intel will have anything remotely ground breaking till EUV chips and a ground up new architecture, they got so comfortable with no threat from AMD for so long that they stopped thinking about game changing technology, because that costs too much to achieve when the competition isn't giving you any reason to try to achieve it.

 

[DragonQ]

They supposedly started on a ground up new architecture at the start of this year in response finally to Zen and it's unlikely to be ready for 3-4 years so 2020-21 and AMD too are working concurrently on something to replace Zen completely as well as Zen 2 and 3, with the replacement aimed at around 20-21 also.

Everything else is supposed to be core architecture and just iterations, though thanks to AMD we'll be getting small iterative performance improvements along with big core count increase. Intel simply has to get 8 cores into their mainstream platform on 10nm and if that is with a gpu or if they'll go the AMD route.. who knows. AMD is going to certainly bring a 6 core APU with 7nm Glofo, will they try and squeeze 8 core in, not sure. It looks like 6 core per CCX makes the most sense, higher end laptops, top notch low end gaming PCs and then 12 core cpus, 24 core threadripper, 48 core server.

Intel's been maximising profits rather than maximising R&D and taking bold steps with their CPUs for some time and 10nm is still struggling massively despite the lead they previously had. Intel's R&D lead seems to be dropping off, again lots of rumours putting it about failed R&D projects, losing good people and having money men rather than engineers directing where the spending goes.

I really don't think Intel will have anything remotely ground breaking till EUV chips and a ground up new architecture, they got so comfortable with no threat from AMD for so long that they stopped thinking about game changing technology, because that costs too much to achieve when the competition isn't giving you any reason to try to achieve it.

Maybe they just never had a backup plan for when their process node shrinks stopped actually yielding improvements. A de facto monopoly will do that to an organisation. From a business point of view I bet they're very happy they held back on 6 and 8 core mainstream CPUs since it gives them something to release in the mean time.

 

[AndreiD]

The only thing they seem to be doing right now to combat AMD's more aggressive competition is just accelerating their release schedule, reason why Coffee released so quick after Kaby Lake and why availability is pretty poor still.
We'll see if it works for them or not, but their Q3 2017 financial report looked really strong.

On the foundry side they're probably still ahead of other foundries, albeit the gap seems to be decreasing. They're touting their 10nm node as having more than 2x the density of Samsung's or TSMC's respective 10nm nodes, but that's just marketing so far since we don't have any Intel 10nm silicon out in the wild.
Intel's big push right now is trying to get 3rd party customers for their foundry business and unlike previous tries they seem to be offering specific nodes for ARM designs, 10HPM/GP for high end SoCs (and other applications) and 22FL for cheaper designs. Supposedly they were cost prohibitive previously, I'm curious if they'll get more customers this time, 10HPM/GP already taped out so we might see some chips on Q2 2018. The Pixel Visual Core chip in the Google Pixel 2 phones is Intel fabbed, probably 22FL?