Intel announces layoffs for 11% (~12,000) of their workforce

[Silent_Buddha]

Intel just announced their plans to get out of SoCs for phones and tablets entirely:

They aren't getting out of tablets. Only phones. They'll still be continuing development of the Atom line, albeit the higher performance Atom parts, IE - Cherry Trail. And even if they drop that they are still committed to the high end tablets (600+ USD) via their Y series CPUs and professional tablets (900+ USD) U series CPUs.

Once more of the former Android only manufacturer's announce their Windows tablets, we'll see whether Atom will continue or not. So far the one I've seen is using Skylake Y, so if the other Chinese manufacturer's follow suite and there is no demand for Cherry Trail, then it may indeed pull out of the mid-range tablet market.

Actually thinking about it. It might make sense to drop Atom as it is entirely possible that the volume of Y and U series tablets might be starting to eclipse the sales of Atom based tablets as I think much of the growth for Windows Tablets is being driven by the Microsoft Surface line hence why companies are trying to copy the design.

Regards,
SB

 

[Raqia]

Yeah, that's a fair point; seems like Skylake-Y would be an exquisite replacement for atom if they clocked it at 500 mhz or so and created a single, hyper-threaded core block. Their margins would take a massive hit though, and furthermore most mobile SoCs have a good number of important blocks that are not CPU, so Intel would be expending quite a bit money in R&D for something that's outside their wheelhouse. We'll see if Intel goes that route and if that massive die is commercially viable when 4-6 ARM cores fit in the same transistor/die budget.

 

[Gubbi]

I'm wondering if Intel will realign their microarchitectures now they've given up on <2W skus. Skylake scales down to 4.5W (reconfigurable to 3.5W TDP) and all the way up to very highly clocked workstation SKUs. The upper bound on operating frequency dictates the amount of pipelining needed to achive this frequency, - or the amount of power burnt to honour the timing constraints.

I could imagine, essentially the same microarchitecture, implemented for two main TDP/frequency targets. One implementation with a high number of FO4 delays per pipestage (and fewer pipe stages) for a low frequency/TDP target and a fast one (like the current Skylake offerings).

I don't see any room for Silvermont derived SKUs at all going forwards, so unless they beef up their Atom offerings significantly, it'll wither and die.

Cheers