Hacker News Reader

Spec sheet for the CPU is there. I just don't see how IO doesn't end up murdering performance. 16MB shared L2 is across all the cores?

The cores are so wimpy that 16MB is an appropriate amount. For comparison a Xeon D only has 12MB.

Xeon D's also have a 256KB L2 though then 1.5MB per core shared L3. Whereas this is 333KB per core L2 which is presumably shared.

Not really informed enough to say it would be the performance bottleneck but it seems like it would be to me in most server environments.

The Anandtech article linked above explains that the Arm cores differ from the Xeon, since the pipeline architecture is not as complex.