Stories
Slash Boxes
Comments

SoylentNews is people

SoylentNews is powered by your submissions, so send in your scoop. Only 15 submissions in the queue.
posted by janrinok on Wednesday March 05 2014, @08:25PM   Printer-friendly
from the All-change dept.

bob_super writes:

"The rumors are getting more persistent that Altera may abandon the move to Intel fabs and go back to TSMC.

While this seems to be specifically due to delays at the 14nm node, would failing to deal with the 2nd biggest FPGA vendor doom Intel's attempt to open its fabs to others?"

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1) by tibman on Wednesday March 05 2014, @10:30PM

    by tibman (134) Subscriber Badge on Wednesday March 05 2014, @10:30PM (#11556)

    Sounds like TSMC is doing well at 20nm. I didn't understand the sharing of Intel's fab from the article. It isn't mentioned. Sounds like something that everyone should just know. If Altera was supposed to be the first non-Intel company to use Intel's 14nm fab then that is interesting. But seriously, Intel can't expect someone to wait around for the next six months and not make chips.

    --
    SN won't survive on lurkers alone. Write comments.
  • (Score: 2) by TheRaven on Thursday March 06 2014, @02:42PM

    by TheRaven (270) on Thursday March 06 2014, @02:42PM (#11952) Journal

    Intel has been trying to encourage SoC makers to use Atom cores and Intel's fabs, so that they can compete with ARM. The traditional model for ARM licensees is to license a core design (or a few) from ARM, possibly (depending on their in-house expertise, the amount they're willing to pay ARM, and their target market) hack them up a bit, and then stick some extra stuff (custom coprocessors) on the die. Altera does this with quite a few parts, combining 1-4 ARM cores with an FPGA so that you can prototype custom coprocessors (or even ship the FPGA version, if the volume is sufficiently low that it's cheaper than an ASIC).

    For people in the traditional ARM model, ARM is more or less a single supplier, but they have a lot of competing fabs, each of which is interested in new orders. ARM will put their licensees in contact with fabs, but they can also be completely independent contracts. In the Intel world, you get everything together but the down side is that you are a tiny customer for the fabs, whereas Intel's own product lines are a huge customer.

    This is combined with the fact that TMSC has had better luck with their latest die shrink than Intel and so makes using Intel fabs a lot less sensible. This is unfortunate for Intel, because getting companies to use their fabs is a key part of getting companies to produce Atom SoCs, which is core to their plan to compete with ARM. If you're making a new phone, table, or whatever, Atom doesn't look that interesting because it's a single-vendor solution. In contrast, if you buy an ARM SoC from, say, Samsung and have a bad experience with them, you can easily get the next one from Qualcomm and reuse most of your software stack, including third-party userspace software.

    --
    sudo mod me up