I think it's less about yield, than it is about amortizing a large R&D budget across more chips. Keep in mind that they are competing with Intel who ships huge volumes and is using this generations profit to fund the next generation fabs.
So AMD pours everything into a single die and manages to hit the desktop (Ryzen with 1 die), workstation (thread ripper with 2 dies), and server (Epyc with 4 dies). All with a single wafer of silicon, and as a bonus the memory bandwidth, max memory capacity, and total number of cores scales to fit all 3 markets.
Pretty crafty. This is far from new of course, various Power CPUs, Intel cpus as far back as the pentium pro, and of course the previous generation Xeons.
Intel's strategy has been one die for most of their low end/low power chips (max 2 core/4T) and a larger die for their desktop (max 4c/8t) that use the same socket (LGA1151).
Then Intel targets HEDT (High end desktop) and server with the same chip, same die, same socket, just marketing to differentiate the x-servies chips and the regular single/dual socket chips that share chipsets, sockets, and a lga2011 socket.
AMD seems to have scared Intel pretty bad. They have held off on the skylake xeons, only shipping them to cloud providers while they wait for the AMD release before releasing skylake xeons to the masses.
I'm glad to say AMD performance seems pretty good, SpecINT (using GCC) is around 50% faster than the similar Intel chip and SpecFP is even better. Seems fair unless you use the Intel compiler to compile all your binaries anyways. In fact the fastest AMD + gcc-6.2 is faster at SpecFP than the fastest Intel + Intel's compiler (1330 vs 1090 respectively).
Intel also shares dies across their laptop and desktop lines. I'm not sure how many dies there are, but they don't have much more than a 2 core and a 4 core version of each generation of silicon.
Skylake generation servers will have at least three different cores (LCC, HCC, XCC; low core count, high core count, and extreme core count) + a Xeon-D successor (presumably).
That's a desktop chip rebranded as Xeon, with a different set of feature flags. The server chips are a totally different beast, with many more cores, dual socket support, network on chip, AVX 512, upgraded connectivity etc.
These are different skylake xeons. Skylake E3 were available in 2016. They are limited to only 4 cores. The OP refers to the server-grade Skylake Xeons, previously referred to as E5 and E7. They will reach a lot higher number of cores.
Is there any Intel core with 128 PCIe links then? I'm not aware of any. Iirc Intel is at 40 lanes / socket, which is more like 10 per core rather than 128.
No, because if they did it would be wasted. Board manufacturers wouldn't lay out all of those lanes to slots because of the increased expense and limited market at that pricepoint that's interested in saturating all that I/O.
So AMD pours everything into a single die and manages to hit the desktop (Ryzen with 1 die), workstation (thread ripper with 2 dies), and server (Epyc with 4 dies). All with a single wafer of silicon, and as a bonus the memory bandwidth, max memory capacity, and total number of cores scales to fit all 3 markets.
Pretty crafty. This is far from new of course, various Power CPUs, Intel cpus as far back as the pentium pro, and of course the previous generation Xeons.
Intel's strategy has been one die for most of their low end/low power chips (max 2 core/4T) and a larger die for their desktop (max 4c/8t) that use the same socket (LGA1151).
Then Intel targets HEDT (High end desktop) and server with the same chip, same die, same socket, just marketing to differentiate the x-servies chips and the regular single/dual socket chips that share chipsets, sockets, and a lga2011 socket.
AMD seems to have scared Intel pretty bad. They have held off on the skylake xeons, only shipping them to cloud providers while they wait for the AMD release before releasing skylake xeons to the masses.
I'm glad to say AMD performance seems pretty good, SpecINT (using GCC) is around 50% faster than the similar Intel chip and SpecFP is even better. Seems fair unless you use the Intel compiler to compile all your binaries anyways. In fact the fastest AMD + gcc-6.2 is faster at SpecFP than the fastest Intel + Intel's compiler (1330 vs 1090 respectively).