Server or Smartphone?
Quad vice Octocore*
64-bit, yep
3GHz vice 2 GHz, there's one.
16nm FINFET vice 14nm FINFET
* Yes, I know.
But still, late = dead.
Broadcom is shutting down efforts to develop its own server-class 64-bit ARM system-on-chip, multiple sources within the semiconductor industry have told The Register. It appears the secretive project, codenamed Vulcan, did not survive Broadcom's acquisition by Avago and is gradually being wound down. Engineering resources …
"ARM-based servers have been hyped in the market for 6-plus years, with little to show for it in terms of real customer adoption," Gina Longoria, a senior analyst at Moor Insights and Strategy, told The Register on Tuesday.
Maybe if someone that anybody ever heard of actually had some for sale, then customers would buy them ?
AMEN!!!
The number of times I've seen what looks to be a good board or system, but then find they'll only sell to you if you own a datacentre and you're planning to buy 100 units and are willing to sign an NDA.
I can walk up to a computer shop, buy the parts for an x86-based server, and build it myself, no questions asked. Getting an ARM-based one requires sacrifice of a few ritual animals and costs an arm and a leg! x86 might cost more to run, but it pays for itself in the lack of general stuff-around.
Until they solve this problem, I'm sorry, but ARM servers are a lame duck!
At my previous job we had wanted to buy a couple for 3 years now, and nothing ever announced was ever apparently actually able to be bought. Lots of press releases, pictures, etc, but nothing actually for sale.
Step one in getting a market is clearly to actually bloody well let people buy the damn stuff.
"very specific requirements which dont match fully match with the Intel Xeon roadmap "
More details very welcome.
Meanwhile based on my past experience, mainstream server builders looking at ARM as an alternative would mostly be concerned about losing "best mates with Intel" status if Intel hear about their infidelity.
Allegedly.
There's also ARM's new owners. That may have been a great idea for ARM shareholders short term, but it's less obvious that it's such a great idea for the future of ARM (and ARM licensees) in new-to-ARM markets.
Well ..... Intel wants every bit of your cash and they dont care too much about how much current/power their customers use - bar setting alight the motherboard.
Big server farms want lots of powerful server using as little as power as possible.
i.e. Intel trade offs between CPU performance and grunt are different to Serverfarm's.
A server farm could be persuaded to ditch the x86 legacy. Intels business depends on x86 legacy; without that thats its just another silicon fab.
Arms dont make chips. Not sure about Softbank; ARM customer toes and all that.