AMD's 64-bit ARM server chip Seattle finally flies the coop ... but where will it call home?

They should start selling them as appliances

Things like Spark or Hadoop clusters-in-a-box, Or Cassandra, ElasticSearch etc.

The density needs to go way higher, though - even a 5x improvement in performance per watt will get drowned out by Amdahl's law when CPU power consumption only represents about 20-30% of total system power usage (DRAM accounts for a significant proportion).

The future can't be prevented. Only delayed.

There will be ARM servers.

In the meantime this will make a nice home NAS.

Nice to see some innovation

scale out vs scale up designs, but more importantly to keep Intel on their toes.

KVM on ARM

Or maybe the future really is rows and rows of generic x86 boxes running KVM on Linux.

Erm... KVM on Linux already runs fine on ARM as well.

if you are not running windows then x86 is not an issue

Intel did a lot of work getting the x86 to perform well and it is the patents on this work which still limit other architectures.

I hope they can survive

It'll be a long time before companies will want to move over to ARM systems. Companies don't care about long-term cost savings when there is the risk that everything will just fall apart. The two biggest costs for a company with a large server real-estate are: maintenance costs of the hardware and development costs of the software that goes on it; the power cost for the systems and initial purchase of the hardware is minuscule. Even if the code itself doesn't need any re-work, you still need to do months and months worth of QA testing before management even considers moving over to a new platform.

While companies love money, they aren't willing to save 5% on operational costs if it means risking the entire things collapsing for even an hour or two. Repayment to clients due to SLAs and lost business, not to mention the bad PR will cost a business far more than the cost savings of going with a different platform.

For the past 15 years, I've been trying to get a client to replace their HP 9000 dinosaurs with x86 hardware and the above have been their counter-arguments. And if I can't get a fairly dynamic company to move away from a completely dead platform and would create close to 85-90% cost savings, what hope does ARM have when trying to get companies to move away from x86?

Will there be a common platform

The big problem about ARM right now is that virtually every computer requires its own OS image. While on PCs you can boot your OS on virtually all PCs with just 2 images, ARM systems are incredibly diverse.

Seriously on a PC you can blindly access some ports and you will always find a serial port. Same goes for basic graphics, USB, SATA or whatever you could wish for. And for things that do not have fixed ports you can query the devices in a standardized way.

This needs to be established first before we can think about having ARM servers. There's no point in having one that can only run version X of operating system Y.

Hopefully eventually such a standard platform will move into the mobile world. This would finally mean that you could choose your operating system independently of your manufacturer. You could even update your software without it. Suddenly a mobile device might actually be useful for a longer period of time.