ARM server chip upstart Calxeda bites the dust in its quest for 64-bit glory ARM server chip designer Calxeda has shut down as one of its executives told The Register: "We simply ran out of money." The plucky company, founded in 2008, had banked $90m in funding during its lifetime, and carried the torch for taking ARM-compatible system-on-chips into servers. But in the past 30 minutes, news broke that …
More WMD for patent trolls on the auction block.
Hard to not see it coming though with HP Moonshot coming slow, switching to Intel for "strategic" reasons, and SeaMicro bought out by AMD it appears that the legacy data center chip providers have put a temporary delay on the ARM server chip threat.
Now it is time for Facebook and Google to form a joint venture to capture the IP assets before the patent trolls seal off this path forever. Hope they learned their lesson after what happened with Sun and Novell and don't let this fall into the wrong hands.
"Hard to not see it coming though with HP Moonshot coming slow, switching to Intel for "strategic" reasons, and SeaMicro bought out by AMD it appears that the legacy data center chip providers have put a temporary delay on the ARM server chip threat."
Do you really think AMD cares what processors it sells? x86 is the baby of Intel and they want their design to be used and that is why they concentrate on x86 so much. AMD has made a living out of selling x86 processors but if there is money to be made with ARM, why wouldn't AMD produce them as well?
Yeah, but AMD doesn't need Calxeda for that. AMD has taken an ARM license. With Calxeda dead AMD doesn't have to let their ARM server be anything more than a low volume premium niche product to protect their Opteron sales. Calxeda was hoping to generate real volume with breakthrough performance per watt AND density AND dollar, turning the server world on its head. Disruptive. It is not so often that we see the synchronous activities of such bitter rivals arrive at the happy - and completely coincidental - demise of a threat to them both.
So this is probably related to the dual stories from the last few days where both Google and Facebook are looking at having their own ARM server chips built to spec. Because they aren't going to get that disruptive deal from Calxeda.
You have no idea how much history repeats itself.
AMD developed the problem plagued K5 processor. Another company called NexGen was struggling/died. AMD bought it and rebranded their processor as K6 aka Athlon and the rest was history.
http://en.wikipedia.org/wiki/AMD_K6
Original HP notebooks were POS. After the purchase of Compaq, the Presario range injected the much needed know how to make reliable notebooks. The same can be said of their DL series of servers, orginally from Compaq. HP original Netservers was a joke.
I am sure there are more examples.
I would actually expect nVidia to take interest in IP on the table. Currently their ARM and GPU lineups are somewhat diverged, but IIRC the intent was to put ARM cores inside CPUGPU. There might be some IP in Calxeda to help with this, even if not directly then for patent exchange, at least.
It's right up there with getting involved in land wars in Asia - just about the time your heroic efforts have started to move the world you'll run out of money, energy, or the patience of partners, and then usually someone still clever but a little less pioneering will reap the benefits.
Yeah, you have to wonder who on earth started that story.
Just because a Google engineer approached ARM to get some samples to do some price/performance analysis for their custom liquid cooled DC is hardly anything conconrete. Any sane internet engineer would do the same if they had that sort of scale.
It was a pipe dream from day 1, I'm weary that some of these tech stories now-a-days are trader/media manipulation attempts on stock prices. (not saying that was - but sometimes - the link is obvious on certain tech stocks)
HP can be so stupid. To name a project/product moonshoot is saying to potential customers that it will fail but we have a 1 in a billion chance it might be worthwhile. If the prior sludge CEO hadn't blown 11B on a stupid sw company they would have bought calxeda.
Someone has to keep Intel honest and if it's not going to be ARM then it has to be Power.
with AMD and Itanium practically dead x86 is really just "Intel" now. So much for x86 not being Proprietary !
Karl...you should go back to IBM....I hear they miss you in the Power division.
cheers guys....Allison
Hi Alison, I can't comment directly for the Moonshot team but I believe on the Caldera front the target was their upcoming 64bit update rather than anything shipping so would have been tricky.
The ARM based TI solution was announced as in the labs for testing at Discover. Its primarily focused on DSP workloads. If you look at those signed up to Pathfinder on HP.com you can see ARM licensees such as Applied Micro and Cavium. We've announced Intel and AMD x86 solutions as well because Moonshot is workload centric rather than general purpose, so the intent is to have a broad ecosystem of x86, ARM and other solutions to give the right solution to a number of tasks. If your particular interest is ARM then there will be multiple ARM options don't worry (and with AMD and NVIDIA being on Pathfinder that may already be two more large and established companies).
I don't think you'll find a wider ecosystem of SoCs anywhere :). Hope this helps change your mind on HP here. Cheers, Alex
Calxeda dead? How can this be? We are repeatedly told that ARM based servers will be little bundles of energy efficient goodness surrounded by open-source wizardry and pixie-dust.
Surely some white-knight venture capitalist must be prepared to pony-up another $90M to reap the fruit from the tree of success ... Nope! Its all smoke and mirrors apparently, quel surprise!
And, of course nothing to to with the recently-spun "Twin ogres of Google and Facebook" story about them just dying to eat ARM based servers ... Nope! , that was smoke and mirrors too, it seems and related to attempts to garner funding . Shame on you Investment bankers :Trying to make me pour my pension fund into an "evidence-free" ARM based black hole.
Merry Christmas SeaMicro :-) and a Merry Christmas to Qualcomm and TSMC for 2016
some white-knight venture capitalist
They don't exist. they would only buy to sell at a profit within a relatively period of time. Another possibility is selling to prospective customers: HP, Dell, IBM, etc.
ARM servers are coming, not least because they will be available from multiple vendors which should keep the market open.
A real shame for the employees, just before Christmas too. I hope this was done whilst there was enough cash in the bank to pay their final salaries and redundancy.
As for potential purchasers, we have Google as a major company that could do with owning its own hardware company for server uses (and search appliance uses), and of course Facebook (with that employee on the board of Calxeda) would also be interested.
This also shows how hard it will be for AMD to compete in the ARM Server market with their forthcoming ARM server chips, but at least AMD have industry contacts and fabric (Seamicro). Indeed this could be what put investors off sticking another $30m into the company.
90/5 years = $18 mil per year.
$18 m / 130 employees = $138k / year (not really though, I'm quite sure the executive staff took millions / year)
They burned through that much money and they're now in hole. It always amazes me how management always allows it to get this far. Having worked in a similar situaltion myself, management is always overly optimistic and convinces the investors right up until the very abrupt end.. I commend them for their relentless pursuit, but at some point reality has to (or should) prevail.
If you're ever in a situation where, one day, you notice that the executives start to shutffle in meetings between their offices quickly, quietly and shutting doors behind them and this goes on for a couple of days, start getting your CV together, you're going to need it.
Wow.
Something tells me that one or more of the following occurred:
1) Mismanagement on a gross scale
2) Lots of perks for the executives
3) Changing requirements (hardware and software). Nothing burns through money faster than a new spin of an ASIC or chasing ever changing hardware - constanly dealing with parts going obsolete.
Ahhh, the life of working at startups (I remember them well), where things can blow up quickly when this happens.
A pint to commisurate with my brethern (the lowly 99%)
$90m was not all salaries.
Calxeda designed and got fabbed two complete server SoC's (ECX-1000 w. Cortex-A9 and ECX-2000w. Cortex -A15) with their own fabric (to connect multiple SoC's). That cost a lost.
They are (were) a startup with a specific idea, what should management have done before the money ran out? Started chasing some other idea?
Calxeda is not the only company developing ARM-based server SOC's.
Facebook has no interest in ARM, just the HW guy there ...
I remember when AMD released Opteron, they had quite a bit more of the market for a couple years for themselves, before intel game back with the Core CPU.
Intel is now moving as slowly as possible, because AMD cannot really catch up, although since their ships are in this years consoles, I am sure they will have some more cash for R&D ... still, if a guy throws shitloads of cash at ARM, I am sure Intel will laugh them out the room ... until they come back with something that will kill i7.
First they ignore you, then they laugh at you, then they fight you, then you win.
Same is happening with OpenSource ... it simply takes some time to tilt a monster like that, but the day Microsoft's Server/Office marketshare only starts sliding, they are dead. Look at ie ... nobody wants it anymore, despite expensive advertising on TV ... for a product they are giving away ... ROFL.
"Facebook has no interest in ARM, just the HW guy there ..."
The big data centers are gunning for open servers as a method of not only squeezing Intel but providing alternative avenues of server configurations and architectures.
Facebook started the Open Commute project with whole purpose to move to an open agnostic CPU/Sever card which they eventually would be able to swap out directly with another variant whether it be based around ARM, Intel etc. They are also carrying out other changes such as moving their Web PHP engine to ARM
http://tinyurl.com/knkvutr
The problem here for Intel is that it a sea change in market. Instead of a ecosystem thats tied to a x86 design it would be a 'open for all' . To compete Intel will have to completely re-evaluate their prices/margins which is going to hit them hard
Calexga were in the unfortunate position of being to early to the market with no major monetary support except Venture Capitalists. Once AMD, Marvell etc started throwing there hats into the ring then it was always going to be harder and harder to keep the money flowing in as it made it less of an investment and more a high risk gambol for paying off.
Open Commute? What a great idea! Wonder no-one ever thought of it before...
I still have an ALR promotional T-shirt I wear when painting or doing anything messy. The slogan across the back says "Just upgrade the CPU!" as that was ALR's Big Idea, PCs with an exchangeable CPU card rather than a proprietary socket on the motherboard. It didn't work out for them for a whole lot of reasons and I suspect it won't work in the server market either, given the operating life of a server in a datacentre is three or four years before the I/O, network, memory interface etc. is obsoleted by the Next Big Thing(s) and a swappable CPU isn't going to fix that. Easier and probably cheaper to just swap out the server.
Do you actually know what the Open Commute project is ?
The server card is not just the CPU, it contains the memory, I/O Interfaces, Network etc. Its basically a full motherboard designed to be fully interchangeable with any other.
http://tinyurl.com/phsyyue
http://tinyurl.com/q2jcx9g
Large company Data centers are just thousands of these server cards in racks so making everything highly modular would allow them to mix and match components to their own specifications with no problems of matching CPU architectures etc.
The pros of ARM is also its weakness. True that there are several ARM CPU vendors and everybody wonders if that is good for server designers and buyers.
But having too many people dividing the precious resources to take on mighty Intel that services 10Mu servers each year means no ARM vendor will ever win! Intel loves having so many ARM vendors in the server space as it knows no one will succeed for lack of critical mass.
The reason ARM got successful in mobile is because Intel rejected that business as low price, low margin ten some years ago and did not realize until mich later how big and how fast mobile will take off. That is the hope ARM servers guys had walking in. But Intel has learned from mobile and it just will not let anybody come take its bread away in servers.
Too bad for Calxeda, they were good.
But being good is not good enough in the tech business. You also need to figure out which way the wind is blowing and get there fast. When you're wrong, you'll need to lay it all on the line while you find your way.
Choosing to close shop when cash is around is the surest sign the business had lost its vision long before.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
