>With 256K BW MBps Infinidat achieved 10,024 and E8 37,197 – 37X higher.
You might want to check your arithmetic.
Anyway, IOPs are meaningless, especially as they don't state how big the box is. I can easily double the IOPs of storage model X by simply adding another one, i.e. 2X.
What's important here is one figure, the steady-state latency of 700 us. But even that is unrealistic without the details.
There are also no details regarding the price. Is this before unrealistic data reduction claims or after unrealistic data reduction claims? How does data reduction affect performance? Did they include it in the test?
That said, there's no denying the abilities of NVMe. I do wish that someone independent would run these tests though.