The fact that Microsoft already has numbers for how Milan-X performs in their Azure cloud means that AMD / Microsoft must have been working together months ago. That's the most fascinating part of this story IMO.
---------
The technical information about which workloads benefit from the 1.5GB of L3 cache (woah momma. That's a lot of L3 cache) on these dual-socket Milan-X servers is also interesting for sure and worth a read.
Sure. But then "some customer" ran a whole bunch of EDA RTL "benchmarks" on these CPUs. Hmmmm... who could it have been?
There's something highly amusing about this whole situation. Its clear to me that AMD sold Microsoft a whole bunch of Milan-X chips early. Then AMD rented those servers from Microsoft to run electronic design tools (such as RTL: register transfer language) simulations to design their next CPUs.
The fact that Microsoft already has numbers for how Milan-X performs in their Azure cloud means that AMD / Microsoft must have been working together months ago. That's the most fascinating part of this story IMO.
---------
The technical information about which workloads benefit from the 1.5GB of L3 cache (woah momma. That's a lot of L3 cache) on these dual-socket Milan-X servers is also interesting for sure and worth a read.