Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/for-which-nodes-does-intel-foundry-service-apply.16934/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

For which nodes does Intel foundry service apply?

Mediatek is using 14nm, so presumable it's only for old nodes?
Mediatek is not using 14nm. They are using intel 16, aka 22FFL, aka 14nm transistors with fewer metal layers and with relaxed dimensions that are closer to to 22nm than 14nm. Per intel/Pat IFS is already offering i16, and will offer i3/18A when they become available. Pat has also said that if customers really want to they can use i4/20A. Besides the nodes they are also offering EMIB (already in use by graviton 3) and FORVOS, as well as their mask services (if memory serves IBM is the main external customer for this)
 
Last edited:
Mediatek is not using 14nm. They are using intel 16, aka 22FFL, aka 14nm transistors with fewer metal layers and with relaxed dimensions that are closer to to 22nm than 14nm. Per intel/Pat IFS is already offering i16, and will offer i3/18A when they become available. Pat has also said that if customers really want to they can use i4/20A. Besides the nodes they are also offering EMIB (already in use by graviton 3) and FORVOS, as well as their mask services (if memory serves IBM is the main external customer for this)
Thanks for the answer. For advance nodes I assume Intel's own products would have priority over external customers.
 
Thanks for the answer. For advance nodes I assume Intel's own products would have priority over external customers.
Time will tell. Pat says he wants intel to compete with external customers for IFS business. But presumably intel will get a similar status as Apple gets with TSMC (first dibs on capacity, not having to pay ahead of time, etc). I have to assume that intel is not foolish enough to make the same mistakes they made with their first shot at foundry with how poorly they treated their customers.

As for the early access we are kind of seeing it already with i4/20A, but I assume this isn't just intel wanting an "edge" over their customers. I assume they are using intel products as pipe cleaners and as a way to make sure their PDKs are ready for the prime time (also similar to Apple and TSMC and TSMC running FPGA/bitcoin mining chips in the early stages of a new process).
 
If Intel allowed a big customer use their leading node ahead of their own products it would delay Intel's product and they would lose competitive advantage against AMD. I would highly doubt Intel would do that.
 
If Intel allowed a big customer use their leading node ahead of their own products it would delay Intel's product and they would lose competitive advantage against AMD. I would highly doubt Intel would do that.
I don't see this as an issue. There are two possible ways this could happen.
1. Intel's newest node is so bad intel doesn't want to use it for their client or DCAI products. If this is the case what does intel care if someone else uses it first.
2. The yields are so poor intel doesn't want to use the node yet. If this is the case than the node is likely not even in HVM yet, and if that is the situation why would an AMD want to use the node if it isn't even HVM ready.

A more telling question is if customers can use it at the same or similar time to intel. If a customer does ask to use 20A and accepts the risks and difficulties of being an early adopter/partner can they get it at the same or similar time to intel? Or do they have to wait for more capacity to open up as IFS ramps the node?
 
Maybe I didn't explain well. A new node has limited capacity as it ramps up. Usually, Intel does not even have enough capacity for their own products on the new node. If they let an external customer use it then there's even less available and that's going to delay role out of Intel's products.
 
Maybe I didn't explain well. A new node has limited capacity as it ramps up. Usually, Intel does not even have enough capacity for their own products on the new node. If they let an external customer use it then there's even less available and that's going to delay role out of Intel's products.
Yeah that wasn't what I thought you meant 😅. Then I think we agree on this matter. If they want IFS to do well then they will have to ramp fast on these new nodes (no 14nm like ramps). Maybe this might be the case with i3 given how similar it is to i4, allowing all of the original and new i4 fabs to quickly ramp to i3. By the time 20A/18A start revving intel should start seeing it's new fabs in Ireland, Arizona, and Israel coming online and these might allow for a fair amount of early IFS clients on these nodes.
 
Back
Top