prismatics

mjtomlin said:

lkrupp said:

My 27” iMac 14,2 (late 2013) is getting long in the tooth but I will not wait until the latter half of 2020 to find out if Macs are moving to ARM. The 2019 iMac may be my last Intel Mac but I don’t really care.

Really? My 27" iMac (11,1 late 2009) is still chugging along just fine. Although it is stuck on High Sierra - the first time I've ever owned a Mac that doesn't have the latest OS.

deegee1948 said:

Great hire! This guy has some big chops, AMD, INTEL, and ARM! Just watch what’s coming!

Possible in-house x64 based CPUs?

No. Intel and AMD x86 platform patents license agreements perfectly show how hard it is. The only other company allowed to make x86 products is VIA, but they're embedded products centered.

wizard69 said:

prismatics said:

Don‘t expect 10 nm to come anytime soon for devices that are more complex than dual core ultra low power, maybe quad core. The defect density in 10 nm will never get to levels where it is viable to replace 14 nm.

Of course Intel will keep saying that their process is progressing and healthy, but it makes absolutely no sense from an economical point of view.

Nobody that is digging deeper into that matter believes what Intel is saying.

10 nm will never make Intel any money. However, it would have if we lived in a reality where AMD didn’t exist any longer.

TSMC is doing fine with their quasi equivalent processes.  Even if Intel’s labs are too dirty for high yields they can always go to the bleeding edge with AMDs chiplet approach.  

The difference between Intel and TSMC is that Intel uses self-aligned Quad Patterning with cobalt interconnects while TSMC does so with double patterning (their 7nm ArF node). The ugly thing is, that with Quad Patterning, you have more than double the steps required to fully process a wafer, increasing the machine time per wafer and the likeliness that while exposing the wafer to whatever is needed to process it, defects develop. Its not that their fabs are dirty, its that there is no way Intel can ramp up 10 nm to _mass_ production (e.g. high performance Desktop, Notebook and Xeon parts) before 7 nm will be there. TSMCs next upcoming N7 Pro will give you EUV, eliminating multiple patterning steps, reducing the total required step count by more than half of the steps, increasing yield.

By the way, ASML is the single company that remains which can produce tools for the upcoming processes; It's same for everywhere else where you have costs that are rising. If Intel folds too much in the next years, they will lose much of their manufacturing capability as new nodes are exponentially rising in cost.

Meanwhile, Intel continues to believe, or rather must communicate to investors, that the self-aligned Quad Patterning 10 nm process will be the way, which is, factually incorrect. By the way, Intel starts EUV at 7 nm, and I expect them to be back in game when they can release the process beyond risk production.

cropr said:

prismatics said:

Don‘t expect 10 nm to come anytime soon for devices that are more complex than dual core ultra low power, maybe quad core. The defect density in 10 nm will never get to levels where it is viable to replace 14 nm.

Dell has already announced XPS 13 machines with  i7 and i5 Ice Lake CPUs, just contradicting what you are claiming

It’s a U suffix part which is not designed for high performance but low power consumption.

mdriftmeyer said:

Apple should move to Zen2 on all but their Macbook's and then when Zen2+ comes out make it complete. The TDP of Zen2, IPC, performance already has surpassed anything Intel can compete against. The margin is only going to widen with TSMC 7nm+ for Zen2+ combined with NAVI APUs.

Nothing tethers Apple to Intel any longer. The Thunderbolt licensing is now fully open and it is time to move on. The yields are phenomenal on TMSC 7nm with both GPUs and CPUs for AMD utilizing it. Costs are considerably lower than anything Intel offers to boot.

Server Market share is going to plummet with ROME EPYC's arrival.

Of course this would be a nice proposition, but Apple seems not interested in general purpose computing.

Their directions they are aiming at are entirely different; T2, Marzipan, Code Notarization, Custom ARM Chips, Increasingly unserviceable Machines, No General Purpose Mac Pro and Flash Prices Extortion on MacBook demonstrate this perfectly.

Don‘t expect 10 nm to come anytime soon for devices that are more complex than dual core ultra low power, maybe quad core. The defect density in 10 nm will never get to levels where it is viable to replace 14 nm.

Of course Intel will keep saying that their process is progressing and healthy, but it makes absolutely no sense from an economical point of view.

Nobody that is digging deeper into that matter believes what Intel is saying.

10 nm will never make Intel any money. However, it would have if we lived in a reality where AMD didn’t exist any longer.