You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Next, AMD announced that Epyc Rome will be manufactured at TSMC 7nm, and chip is already sampling. This explains why Epyc 2 will come before Ryzen 2. Also, sticking with TSMC means AMD will be able to maintain its process node advantage vs. intel at least until 3nm. AMD “Rome” EPYC CPUs to Be Fabbed By TSMC
AMD is taking all its chances agains Intel, due to Intel's huge monolithic server dies they usually introduce Xeon last on their process nodes. AMD thanks to is CCX architecture is launching Epyc 2 first. That plus the choice of TSMC will give AMD a huge lead.
How do you think Intel will react? Do you believe Intel will wait until 2020 to have an answer for Epyc 2?
Other than the off "2018" comment late in the INTC call they were very clear that: "The systems on shelves that we expect in holiday 2019 will be client systems, with data center products to follow shortly after." Now "shortly" could be six months and it is not all in their hands as customers need to qualify these things. That could take 6 months. INTC can bomb prices, which may happen in Q4 to deliver their projected 59% gross margins. Swan was very clear in the call ASPs would be down.
The other things INTC management is trying to do is to tie communications and memory into their CPUs in sort of a system sell. I wonder how free MU is to work with AMD on equivalent solutions. Renduchintala noted:"we're really focused on is delivering product leadership generation after generation and that's at the system level. While processors are a very important part of that recipe, so are other ingredients as well, such as product architecture, silicon design and packaging."
Shenoy noted: "We have a leadership roadmap, not just in CPU but in ... a broad set of additional silicon, from ASICs to FPGA to silicon photonics to memory. And our approach has been over time to increasingly stitch all of those assets together to deliver a better solution ... that's what really sets us apart ... none have that broad portfolio." Now I question the value that brings to DIY cloud vendors, but that is their stated competitive stance. Oddly, they did not mention CPU superiority as a big deal!
Well, AMD is together in CCIX consortium with MU, Xilinx and Mellanox, so like Intel, somebody could try to integrate CPU, memory, FPGA and communications rigth?
Besides, I think something is brewing at CCIX, it says "New website is coming soon".
I assume there's 2 possibilities:
1) Somehow they decided the current website is ugly, not up to par and has to be restyled,
2) They actually reached some sort of milestone, something they've been working on @7nm probably? And they want to present it on a new website: The old one was about what may come and about possibilities about what could be done. Let's hope the new one is about what is ready!
