AMD plans to tap Globalfoundries and TSMC for 7nm
AMD plans to tap Globalfoundries and TSMC for 7nm
With 14nm, AMD used Globalfoundries almost exclusively, creating both their Ryzen CPU products that their Polaris and Vega GPU products using the same node, making this partial move to TSMC an interesting change. Both nodes will be used to create different product lines, potentially to allow AMD to create a larger volume of products during the early life of these 7nm process nodes. Â
Below is a statement from Lisa Su regarding the use of both foundries, which comes from her recent interview with Anandtech. With the release of 7nm, both TSMC and Globalfoundries have managed to close the gap between themselves and Intel, the traditional leader of the Foundry market, taking away one of Intel’s prime advantages over their competition.Â
 Â
  Question: With GlobalFoundries 14nm, it was a licensed Samsung process, and 12nm is an advancement of that. 7nm is more of a pure GF design. Is there any change in the relationship as a result?
Lisa Su:Â So in 7nm, we will use both TSMC and GlobalFoundries. We are working closely with both foundry partners, and will have different product lines for each. I am very confident that the process technology will be stable and capable for what weâre trying to do.
Question: Do you feel like that the numbering system of process technology gets people confused?
Lisa Su:Â Yes, it probably does! You know, I donât get to name processor technologies, but I will say though Mark Papermaster made the point today that the 7nm node is a very competitive node, and we do believe that the gap to the competitionâs process technology is closing and thatâs a good thing, right?
Â
At this time the performance difference between TSMC’s 7nm and GlobalFoundies 7nm is unknown, though both nodes are expected to offer similar improvements to today’s 14/16nm products. AMD will not be creating any single product on both nodes, so it is unlikely that we will ever get a proper opportunity to do a like for like comparison between both nodes on a shipping product.Â
Â
Â
Another interesting tidbit from Anandtech’s interview is a comment that AMD is likely to separate compute and traditional graphics chips into individual products, creating more “purpose-built products” to act as “special accelerators”. This is Lisa Su’s comment after she was asked if there was room for Gaming and Compute products to be branched off from each other.Â
  I think there is. You will see us move on this, and weâre very committed to gaming so thatâs not going to change, but you will see us do some more purpose-built products for the compute side of things.
Â
At this time it is unknown what AMD’s plans for the future are regarding Radeon, especially after its merger with AMD’s Semi-Custom unit and the arrival of RTG’s new leadership team. Gamers will undoubtedly be pleased to see compute and gaming graphics undergo some separation, as both markets have very different hardware requirements.Â
This differentiation between gaming and compute can already be seen within Nvidia, with Tensor Cores and FP64 compute being priorities within high-end Tesla GPUs and absent in gaming products. Nvidia creates specific products for certain markets, whereas AMD’s Vega GPUs have to somehow suit the needs of both.Â
You can join the discussion on AMD’s plans to use both TSMC and Globalfoundries’ 7nm process nodes on the OC3D Forums. Â