Calxeda’s end is not the end of the ARM Server. Yesterday at OCP there was a bunch of ARM news and ZDNet’s Larry Dignan posts.
ARM server army revs up at Open Compute Project powwow
AMD announced their ARM processor.
There is sadness in Calxeda’s lack of funding to continue its efforts on ARM servers.
Calxeda Closes Its Office: The Sad Story Of One Of The Most Innovative Companies In Years
But, over a year ago, I started to doubt whether Calxeda would survive. I would chat with some server hardware folks and one of the ideas we would discuss is out of the 8+ companies out there with ARM chips there would eventually be only three who survive, and whether Calxeda would be one of them was not clear. We had more faith in Samsung to be a player than many of the others.
Almost everyone focuses on ARM as if the ARM chip has magical powers that can beat the x86 chip once someone ships an ARM chip for servers. ARM is just a chip. x86 is a chip. I decided to look for a recent paper that compares ARM vs. x86 and found this paper.
A Detailed Analysis of Contemporary ARM and x86 Architectures
Emily Blem, Jaikrishnan Menon, and Karthikeyan Sankaralingam
University of Wisconsin - Madison
The results show. (note: ISA is Instruction Set Architecture)
We ﬁnd that ARM and x86 processors are simply engineering
design points optimized for different levels of performance, and
there is nothing fundamentally more energy efﬁcient in one ISA
class or the other. The ISA being RISC or CISC seems irrelevant
So if ARM vs. x86 is irrelevant what was the battle to fight? It is about creating a new Fabric for Servers. Read the press statement from Calxeda. Their first two paragraphs say Fabric.
Over the last few years, Calxeda has been a driving force in the industry for low power server processors and fabric-based computing.
The concept of a fabric of ARM-based servers challenging the industry giants was not on anyone¹s radar screen when we started this journey.
In April 2012 Intel announced it had licensed Cray’s supercomputing fabric.
Intel announced the C2000 Atom to go against ARM. In this blog post I referenced posts on Barron’s blog by analyst who discussed the ARM vs. Atom.
There are probably many other small things that made it difficult for Calxeda. It’s hard to be a disruptive force especially in Intel’s turf.
I was lucky to get to know the folks at Calxeda early on. Back when I first talked to them in Apr 2010 they were called Smoothstone.
At same time i was discussing the idea that Intel Atom chips would be put in Servers.
The efforts by ARM did help to wake up Intel that low power servers were a growing market. Back in 2010 it was all about the # of cores. Now Performance per watt is a standard measure. RAM and HD are more energy efficient. SSD is more common. PUE’s of 1.10 are no big deal. Power supplies are 90% energy efficient. Batteries are put on server boards by Googl and others.
A lot has changed in 3 1/2 years and during this time the demand for ARM based servers is higher, but no big buyers. Why? Intel made changes along with the rest of the server ecosystem to use less power.
Intel Atom’s are used in low power cold storage systems.
And the x86 code base made it so much easier to recompile code for new chip designs to improve performance vs. porting to the ARM instruction set. 64 bit processors are just showing up in ARM. Who wants to run a server on a 32 bit chip even if it has 4 cores.
GigaOm’s Stacey Higginbotham reports on Calxeda shutting its doors. It is kind of sad, but most of my friends who think about chip designs agreed that Calxeda had a tough road to success. Why? Even with $100 Mil, it doesn’t last long when you are trying to create Silicon.
We’ll see where the ARM server does take off. In the meantime ARM dominates the mobile space.
I was chatting with an analyst friend today and he asked me what I thought about Verizon choosing Seamicro. Here is the press release.
Verizon Selects AMD’s SeaMicro SM15000 for Enterprise Class Services: Verizon Cloud Compute and Verizon Cloud Storage
We shared our knowledge of how many sea micro boxes were being used at various customers we knew and our numbers matched.
Here is the question that almost never gets asked. In the same way that Airbus made a major win over Boeing getting JAL to buy the A350 and the news hits the WSJ. There is clear announcement of the number of planes bought, and the options acquired for future deliveries.
JAL ordered 31 A350 jetliners carrying a catalog price of ¥950 billion, with an option to buy 25 more of the long-distance planes, the companies said Monday. Deliveries will start in 2019 and roll out over six years.
When the hardware vendors start having press releases that identify the number of units sold over what period then we can judge whether the deal is a big one or not.
What we don't know is what is the rest of the hardware in Verizon's Cloud. Is SeaMicro 1% of the units? 2%? 5%? 10%? 25? of the total server count. We don't really know and if you don't know, then why should you think the announcement has an impact on the way you do business.