Servers first on 7nm
7 replies (most recent on top)
They are already ahead on 10nm:
http://www.fudzilla.com/news/processors/40946-qualcomm-to-debut-with-10nm-server-cpu
Just remember IBM used to have a lot of fabs and just made minicomputers and mainframes. They made billions from software too.. so for a while they could subsidize the small factories and fund leading edge R&D but that all went BK a few years ago.
Now look at Intel, volumes dropping like crazy on client, no mobile business to speak of, and modems are made by TSMC.
Volumes dropping, yes server chips can be sold for thousands or more, so even yielding a couple die/wafer intel can make money on 7nm even if they are large. But client selling for 75 bucks you do the math.. and if you need to launch volume for client you need high yields if your yields s--- no amount of pushing wafers can make up. No look at TSMC or Samsung for their cellphone launch they will need to push excessive of 10 million/quarter easily. That requires healthy technology, guess who is going to win the race at the next node. BK'd is intel!
So much for using scale to compete LOL!
10 millions servers chips is nowhere near 200M clients.
Possible, but highly unlikely. If they couldn't meet yield goals on client die at 22nm/14nm/10nm process start, what makes them think they will meet on a much larger server die at 7nm? I don't think this process technology will see the light of day anytime before 2021. By then, TSMC/Samsung/GF will be way ahead and Intel will be in deep financial trouble.
I think op is correct.
It used to that there was actually a compelling reason people upgraded PCs, nothing was more valuable, new generation design on a new technology that got value due to die shrink and new functions. Also you had reasonably good yield and high ASP.
These days new generation client except for the extreme edition don't sell for the same as last generation. Think about that.. if your new product sells for the same price as the last, and the die size isn't smaller and the yield isn't good, LOL maybe really bad how does one make money!
Now with the last few generations it is no secret why the technology is late, yields are terrible. You can't afford to try to transition to 200 million on 10nm or 7nm as with low yields you will BKing it.
Intel's days are numbered.
Source?