Intel Surrenders To ARM

“Intel has discontinued three of its offerings for the Internet of Things and embedded device markets.
 
The chipmaker said in a series of low-key product updates that it would be killing off the Edison, Galileo and Joule compute modules and boards over the second half of the year.”
 
See Intel: Joule’s burned, Edison switched off, and Galileo – Galileo is no more
When you need small cheap computers to do small cheap computing tasks, don’t call on Intel. They are bloated from feeding at the trough of Wintel all these decades. I recommend ARMed computers for just about everything. I’m replacing aged Intel/AMD products with new ARMed devices. It costs less, takes up less space and uses less energy to get the job done. Of course, I run Debian GNU/Linux and not That Other OS…

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.

This entry was posted in technology and tagged , , , , , , , , , . Bookmark the permalink.

13 Responses to Intel Surrenders To ARM

  1. ram says:

    Just the other day I visited a vendor that showed me their latest build. A high end Intel “Kaby Lake” main processor and dual high end NVidia SLI boards. Water cooled. Runs Linux. Approximately 45 TeraFLOPs of computing power for under A$5000. That is alot of “grunt”/$ and pretty good on the “grunt”/Watt rating as well.

  2. oiaohm says:

    Grece read your own link.
    40GbE is essentially four lanes of 10GbE
    The link mentions exactly what I am talking about by a 4x 10G. Of course Grece does not fully read what he is quoting.

    So its rather dated, and the major players in today’s cloud market are all in at 25G.

    No learn to read AWS is the accept-ion not the rule. AWS is using custom built network chips. Annapurna ASIC supports 25GbE,
    Competitor to standard conforming intel solution.

    So was AWS using standard conforming parts no. Stardard conforming parts for 25GbE only appeared at the start of this year.

    See, and correct me if I am wrong, but what Fifi is really talking about is 10Gbase-CX4
    Wrong. 10G x4 setup on soc is in fact a single 40G connection based out of standard conforming connections.

    There is not a single soc chip with it. << this statement of mine is still true. AWS 25G has to connect on the PCIe from the CPU and that consumes like the intel one x8 pcie. 50G also consumes one x8 pcie connection.

    major players in today’s cloud market are all in at 25G.<< this again is backbone or pcie connected not part of a soc chip. So all major players are using cpu with integrated 10G in places because you cannot get integrated faster at this stage.

    Basically clueless. I said that everything above 10G is backbone stuff.

    In fact 40G built out of 4 10G connections can have lower latencies than 50G. More interconnecting paths.

  3. Grece says:

    Of course they are true Robert, but stating what I said to be irrelevant is lame.

    The crux of position is based on Fifi’s argument that “4 x 10G Ethernet is really for sure something targeting server room.”

    See, and correct me if I am wrong, but what Fifi is really talking about is 10Gbase-CX4 which was introduced at the turn of the century approximately. So its rather dated, and the major players in today’s cloud market are all in at 25G.

    http://www.datacenterdynamics.com/content-tracks/colo-cloud/aws-how-to-manage-mega-growth/97431.fullarticle

    How many times would you want to update your network Robert?

    Which is more efficient? 10G -> 40G -> 50G versus 25G -> 50G

  4. oiaohm wrote, “4 x 10G Ethernet is really for sure something targeting server room.”

    and Grece replied, “10G Ethernet?? LOL..
     
    THE standard is 25G as of last year fool.”

    Of course, both statements could be true but Grece never misses an opportunity to attack the messenger with something irrelevant when he could contribute something to the message.

  5. oiaohm says:

    Grece I said server room. There is a difference between server room and data centre. Server room can be quite small.

    https://en.wikipedia.org/wiki/25_Gigabit_Ethernet
    Read carefully 25 Gigabit is only in data centre scale server rooms. Small business server rooms are not data centres.

    Also note the only 25 Gigabit is a PCIe x8 card. There is not a single soc chip with it.

    https://www.theregister.co.uk/2017/02/23/intel_scales_atom_to_16_cores_updates_xeon_socs/
    Yes the new atoms are designed to take that card with there max of 16 lanes of PCIe 3.0.

    4 x 10 G is also interesting this shows Grece be a complete idiot as normal you need something as a bridge device between older to newer.

    Of course 40G, 50G and 100G exist now, but when was the last time you ever stepped foot in a data-center? Oh that’s right never!
    If you had you would have known the problem with 40, 50 and 100G.

    You don’t find 40, 50 and 100 G embedded in soc server chips due to the nightmare variation in standards. Were have two vendors providing 40,50 and 100G and their hardware in fact be incompatible with each other. So not suitable to be embedded in a soc.

    https://www.eurotech.com/en/products/CPU-111-10 Reality a lot of processing hardware is 10G even if you have 100G transports in a server room.

    Also Grece don’t start claiming that I have not been somewhere. Reality is large percentage hardware of server rooms and data centres are 10G and will be 10G into the very distance future. Doing diagnostics on 40-100g stuff does create some serous problems.
    https://www.napatech.com/the-challenges-of-100g-network-analysis/

    Those who have worked in data centres know this. So if anyone was insane and made everything high speed in a data centre would only making life insanely hard to preform diagnostics. Yes 100g and 40g in backbones make sense.

    Up to 50G networking only requires a PCIe x8. 100G takes PCIe x16.

    Please note the 25G by intel is designed as a break out from 100G cable. So cards on the end of 1 cable. 50G specification will be more useful 1 card half the 100G transport.

    THE standard is 25G as of last year fool.
    No that is just a break down of 100G transport. If using Intel spec stuff. Please note cannot be last year. –January 2017– How do you have a standard deployment when in 2016 ordering anything other than samples of 25G from intel was impossible. Yes Jan 2017 is when 25G from intel went into mass production.

    Grece basically 25G something might become standard in server rooms depending on how sales this year go. Of course next year with the 50G standard we could be poking fun any everyone who was idiot enough to buy 25G cards. Right you would be one of those idiots Grece.

  6. Grece says:

    10G Ethernet?? LOL..

    THE standard is 25G as of last year fool.

    Of course 40G, 50G and 100G exist now, but when was the last time you ever stepped foot in a data-center? Oh that’s right never!

  7. oiaohm says:

    http://www.anandtech.com/show/11144/intel-launches-16core-atom-c3000-cpus-for-nas-servers-vehicles
    Better details on the new atom. Historically something built like this would have a Xeon title. 4 x 10G Ethernet is really for sure something targeting server room.

    I would suspect a lot has to deal with the fact AMD has been able to release some decent chips so Intel has to focus on protecting their existing markets.

  8. DrLoser says:

    Not sure what happened there. Probably my fault.

    Something that might actually make a profit.

    Heaven forfend that “profit” should enter into the equation, of course.

  9. DrLoser says:

    Interestingly, however, and as per your cite, Robert, it appears that Intel are redirecting their low-end development effort towards something that might actually make a profit.

    Any thoughts upon this pathetic attempt to make a profit, rather than to oblige nasty little free-loaders like yourself with cheap farty things, you wretched useless old man?

  10. DrLoser says:

    Well, they will work for a living like everyone else who was not granted a monopoly.

    Or, to put it another way, Robert:

    “Holy shit, Intel is doomed. They’ve stopped offering junk no one wants, crap no one cares about, and pieces of shit that should have never existed in the first place. How will they ever recover?”

    I know you are a desperate senile old fool with no relevant qualifications to talk about IT, Robert, but the very least you could do is to address the fricking question.

    Which yo didn’t. As always.

  11. Kurkosdr says:

    I for once agree with Pog that pulling out from the IoT market after losing badly in mobile and tablets is a bad sign for Intel and a sign of inability to execute.

  12. Ivan wrote, “How will they ever recover?”

    Well, they will work for a living like everyone else who was not granted a monopoly.

  13. Ivan says:

    Edison’s aim was to put x86 chips into both the wearables the … Galileo line was similarly pitched as Intel’s challenger to the Raspberry Pi in the low-cost microcomputer space … Joule, meanwhile, had been pitched as recently as last year as a capable IoT platform.

    Holy shit, Intel is doomed. They’ve stopped offering junk no one wants, crap no one cares about, and pieces of shit that should have never existed in the first place. How will they ever recover?

Leave a Reply

Your email address will not be published. Required fields are marked *