This was posted 1 year 11 days ago, and might be an out-dated deal.

Related
  • expired

Google Coral USB Accelerator $110.99 Delivered @ RS Components

801
Google Coral USB Accelerator $110.99 Delivered @ RS Components
Go to Deal
Ljamhos on 20/12/2023 - 10:13 au.rs-online.com (4884 clicks)
Last edited 20/12/2023 - 10:20

I've been keeping an eye out for these for my Home Assistant / Frigate setup, but they've been at least $150 even on RS previously. Happened to check and see they are down to $110.99 with free delivery so bought one!

The Coral USB accelerator by Google, allows for an edge TPU coprocessor to be added to any system, enabling high-speed machine learning inferencing. It includes a USB-C socket to connect to a host computer to perform accelerated ML inferencing. The on-board Edge TPU is a small ASIC designed by Google that accelerates TensorFlow Lite models in a power efficient manner: it's capable of performing 4 trillion operations per second (4 TOPS), using 2 watts of power—that's 2 TOPS per watt.

Edit for GST!

Computing Google Google Coral USB TensorFlow Processing Unit

Related Stores

RS Components
RS Components

closed Comments

  • Master Bates on 20/12/2023 - 10:18
    +201

    I have no idea what you’re talking about but I feel like I should buy one.

    • Grey 13 on 20/12/2023 - 10:57
      +6

      Big brain move!

    • MohanM on 20/12/2023 - 11:26
      +18

      It will help google to better spy on you

      • buyer-123 on 21/12/2023 - 11:18
        +1

        not an expert but this seems like a local AI accelerator, with no data being sent to the cloud

      • gadget on 21/12/2023 - 12:34
        +1

        It would help if you wrapped it in tin foil.

    • MrPants on 20/12/2023 - 11:52
      +4

      I have no idea what you’re talking about but I feel like I should buy one.

      Same, but I feel like I need 2.

      • gadget on 21/12/2023 - 12:35

        You need at least 10.

    • chartparker on 20/12/2023 - 14:08
      +5

      For those with QNAP NASes, you can add a Coral USB accelerator to speed up AI functions such as face/object detection in Qumagie and QVR Pro

      • REDRUM on 20/12/2023 - 19:07

        Dont forget Frigate

    • sevendollarsfifty on 20/12/2023 - 15:35
      +36

      He's just using big words to sound smarter. It's really not that complicated. In layman's terms it just reconfigures the modulation integration later between core instances, rendering instantiation outputs by refactoring quantum synapses across larger data saturation quarks.

      • seamonkey on 20/12/2023 - 16:42

        I haven't a clue, but I feel apart from being humourous, it may also be accurate. Is it?

      • nessism on 21/12/2023 - 06:27
        +3

        I think it's more related to to the layering of defactorated laminated flux principles, in a bninary substrate sense, but otherwise your contextual refactoring is cromulent.

      • rezdo on 21/12/2023 - 07:09

        Does that mean I don't have to work anymore?

      • mikeb on 21/12/2023 - 08:33

        Yes I thought that's what he meant….lol

        I think I want one?

  • easternculture on 20/12/2023 - 10:19
    +9

    Will this mine [add random letters]coin crypto

    • aaronsd on 20/12/2023 - 10:25
      +7

      Haha, I know you are joking but I've seen some projects that attempts to do all sorts of weird things with the Coral TPU. Fortunately (or maybe unfortunately for some?) the primary use of this is for improving efficiency of some very specific AI workloads. The most common use of this device is with Frigate via Home Assistant (and some other similar home NVR type software) to add some AI smarts to your home camera security system.

  • AloraanGrogu on 20/12/2023 - 10:19
    +1

    I think that price is ex GST

    • jamhos on 20/12/2023 - 10:26

      Yep oops, fixed! Still cheeeeep.

  • chossenger on 20/12/2023 - 10:21
    +1

    Isn't the $100.90 price excluding GST, so more like $111?

  • RtN on 20/12/2023 - 10:23
    +10

    Does this give me more FPS for overwatch?

    • Sleuth on 20/12/2023 - 10:40
      +19

      Yes, it would increase your Flops Per Second.
      I assume you call your home security system "Overwatch".

      • Sleuth on 20/12/2023 - 10:42
        +4

        I'm running to the ATM machine as we speak.
        SMH my head…

  • clartan2015 on 20/12/2023 - 10:24

    has anybody got this to work with BlueIris + Deepstack? thinking if its an "easy" task to offload the processing to something like this coral

    • jmadden91 on 20/12/2023 - 10:27
      +3

      Should be using codeprojectAI rather than deepstack these days. I believe it supports Coral now

      • clartan2015 on 20/12/2023 - 10:38

        thanks! bought one. hopefully its an "easy" transition =)

        • nessism on 21/12/2023 - 06:29

          Famous last words (I have no idea, just wanted to pretend I do).

    • onlinepred on 20/12/2023 - 10:40
      +2

      After trying deepstack and blueiris with ambitions to get this, I gave up and bought smart cameras haha

      • clartan2015 on 20/12/2023 - 10:44

        oopps! hopefully that codeprojectAI + coral wont be too much trouble lol

        • onlinepred on 20/12/2023 - 10:53
          +1

          It's not, just don't expect anything amazing is all. It's still well below any smart camera, but requires a lot of fiddling, setup, maintenance, etc

          • Bruceflix on 20/12/2023 - 11:01

            @onlinepred: I don't think so. Whilst initial setup may be a bit more complex, once it's set it just works.

            • onlinepred on 20/12/2023 - 11:03

              @Bruceflix: I thought so too, but after an update, I basically had to set it all up from scratch. Also after using deepstack, it basically stopped working, and had to switch to codeproject. Meanwhile, my nest camera has not stopped working, and consistently gives me accurate notifications and barely a few false positives in over 2 years, while I get multiple false positives a day from codeproject.

              • Bruceflix on 20/12/2023 - 11:08

                @onlinepred: My specific case was with Frigate and HA. I haven't had any issues. I have no experience with deepstack or codeprojectAI.

              • JPNx on 20/12/2023 - 11:54

                @onlinepred: What type of security are you running that there's a need to get instant alerts of activity?

    • blues99 on 20/12/2023 - 12:24
      +1

      Mine works great with CPAI and BI

  • jcmpscn on 20/12/2023 - 10:29
    +3

    Nice deal, bought it. Want one for my frigate setup for a while too.

  • beefsandwich on 20/12/2023 - 10:34
    +9

    Thanks OP, this should go nicely with the CodeFlange / SmooshAI setup I have running my local hosted IntegraCel switched Tatercity HelliVer 2.3

    • itsfree on 20/12/2023 - 10:52
      +39

      I don't even know if you're joking lol

    • banana365 on 20/12/2023 - 12:14
      +22

      You go to your room and think about what you've just said. We'll not have language like that in this house.

    • JamesWan on 20/12/2023 - 14:01
      +1

      Make sure it's the left code flange

  • AwesomeAndrew on 20/12/2023 - 10:40

    can this run LLM such as llama2 or mistral?

    • bonechiller on 20/12/2023 - 10:48

      I would love to know this as well.

    • ngengerous on 20/12/2023 - 11:43
      +4

      Unfortuantely not - LLMs are heavily VRAM/RAM reliant, which this thing doesn't have much of.

  • Flying Ace on 20/12/2023 - 10:42
    +23

    Wow this is a thread full of terms and product names I have never heard of before…

    Frigate, Coral, CodeFlange, SmooshAI, IntegraCel, Tatercity, HelliVer, BlueIris, Deepstack, CodeProjectAI 😳

    • WhyAmICommenting on 20/12/2023 - 12:09
      +8

      Pretty sure some of those are made up!

      • mlevit on 21/12/2023 - 11:34
        +1

        They're all made up.

  • Anonymous User on 20/12/2023 - 10:48
    +38

    I feel like I'm too stupid for this post

    • Grey 13 on 20/12/2023 - 10:58
      +3

      Be like @Master Bates: buy now understand later (or never).

  • elitistphoenix on 20/12/2023 - 10:54
    +1

    Tried coupon FIRST15 (15% off orders over 100) said it applied. Still got billed the full amount.

  • Commodore64 on 20/12/2023 - 10:57

    Wow these have gone up.
    I bought my dual edge PCIe (two TPUs) for about $60.

    • Bruceflix on 20/12/2023 - 11:06
      +4

      They were about $30/tpu until demand skyrocketed once it was found to be the ideal component for autonomous, loitering munitions.

    • UltimateAI on 20/12/2023 - 11:53
      +1

      USB version has always been more expensive, the dual M.2 and PCIe models are still < $75 AUD.

      • yamumma on 20/12/2023 - 18:30
        +1

        Not sure why you got negged, but this is accurate. Got both an m.2 and a single PCIe version for about $30ea some time ago when the USB ones were scarce.

        Both are preferable to another USB dongle in my opinion. Removed onboard WiFi and tucked away inside a Dell SFF. Frigate + Coral, works a treat.

        Edit: oh, because you negged the deal?

  • controlc on 20/12/2023 - 11:03
    +2

    Got one for my Frigate setup, thanks OP!

  • SubNoize on 20/12/2023 - 11:05

    (profanity) yeah bro, been waiting for these to drop in price for ages. Can finally run frigate now

    • jamhos on 20/12/2023 - 11:09

      I've been running it anyway and the CPU detection is kinda hilarious… anywhere between 0.02 seconds and 15 seconds, or sometimes not at all. Very excited!

  • coupeborgward on 20/12/2023 - 11:06

    was waiting for this for a long time. This apparently can be used with my qnap to improve the face recognition.

  • hyatc on 20/12/2023 - 11:06
    +14

    For those who don't know what this is
    Imagine your computer is like a chef trying to prepare a lot of dishes quickly. The Google Coral USB Accelerator is like a special tool that helps the chef chop ingredients faster, making the cooking process more efficient. In tech terms, it's a device that helps your computer perform certain tasks, like recognizing objects in images, much more quickly and effectively.

    • Bruceflix on 20/12/2023 - 11:10
      +10

      But what is a chef?

      • WhyAmICommenting on 20/12/2023 - 12:10
        +16

        It’s a bit like a computer..

        • Thiefsie on 20/12/2023 - 14:47

          Gold

          =D

        • nessism on 21/12/2023 - 06:46

          .. that cooks things.

      • toasty on 20/12/2023 - 18:37

        Jerome Nigel McElroy is The Chef

    • rumblytangara on 20/12/2023 - 11:21
      +2

      Offloads specific tasks way from the CPU?

      'cause I read the product sheet and still had no bloody idea what this thing is for, besides "connect to RPi"

    • Bedgrub on 20/12/2023 - 20:16
      +2

      It’s more like the Kleva Slicer of ML: it does light work ok, but it’s not for commercial kitchens.

  • No on 20/12/2023 - 11:15
    +1

    Im currently running frigate with 3 cameras using cpu mode. It works reasonably fast but I do get a fair few false detections.

    Will this help with the false detections or just make it faster?

    Thanks!

    • Woodyg22 on 20/12/2023 - 11:21
      +3

      Just faster, which could allow you to go higher resolution detections, maybe leading to higher accuracy results. Tweaking your filters may also achieve the same result

      • No on 20/12/2023 - 11:31

        Cheers, I think Ill grab one and see if I can improve it. I am currently using the low res sub feeds from my cams.

  • shanlee on 20/12/2023 - 11:16
    +1

    No idea what this is about. but a + for OP's post.

  • AlienC on 20/12/2023 - 11:21
    +3

    Does this let me interfere with the matrix database. I need to speak with Top G.

  • [Deactivated] on 20/12/2023 - 11:22

    Which nvidia GPU can you compare its performance to?

    • pscac001 on 20/12/2023 - 11:39

      That's kind of like asking which Intel CPU a 2060 is comparable to. They are both doing similar jobs (in that they process things), and you can kind of compare them based on a very specific task, but they work in very different ways and very small changes in the workload will have drastically different effects.

      The one thing you can safely say is that for a task that is well suited to running on a Coral (which doesn't even include all machine learning processes), you will get MUCH higher performance per watt.

      • [Deactivated] on 20/12/2023 - 11:44
        +1

        " for a task that is well suited to running on a Coral" which GPU can you compare it to? Just curious. Not gonna buy one.

        update: Anyway cgpt told this:

        When comparing the Coral USB Accelerator to an NVIDIA GPU, it's a bit like comparing apples to oranges because they are designed for different purposes and scales of operation. However, if we were to make a rough comparison based purely on the capability to run lightweight machine learning models, the Coral might be likened to an entry-level NVIDIA GPU like the NVIDIA GeForce MX series or the lower end of the GeForce GTX series in terms of machine learning performance. But remember, this comparison is very approximate and depends highly on the specific tasks and models being run.

        • pscac001 on 20/12/2023 - 11:53
          +2

          You still can't, different models and even different versions of the same model might have drastically different results.

          To give a very, very rough comparison, I know of a couple of people who were running 1070 cards to do object detection on half a dozen cameras, who removed their GPU after installing a coral. It's not really a meaningful comparison in terms of processing capability, but I guess they both achieve comparable results from the user's point of view.

    • UltimateAI on 20/12/2023 - 11:52
      +3

      These are a lot slower than most modern GPU's, however they are extremely efficient, meaning low power and low heat.
      When building devices on 24x7 the reduced heat is noticeable, often meaning I can deploy passive cooling only.

      • jebdra on 20/12/2023 - 19:00

        I'm curious

        Why did you neg the deal?

        Most people who seem to know what OP is talking about (rather than making clever comments) think it's ok

        • jebdra on 20/12/2023 - 19:02

          Apologies. I just read further down the thread
          Please disregard.

        • UltimateAI on 20/12/2023 - 21:03
          +1

          Yeah, if I could revoke the Neg I would, from a NOW price comparison and availability it's the best price TODAY, but is this a deal when it's still above RRP?
          I purchase these regularly and they do come available for under $100 AUD delivered.

          I think OP did a good community service to share with the community that these exist, I am a big fan and at this price for what I do NVR builds this is still a great price.. (Although I mainly get the M.2 versions, cheaper and cleaner), I do also have purchased a few USB ones and they work well for Image recognition.

  • iLikePizza on 20/12/2023 - 11:22

    Thanks, got one!

    Hopefully it works well with Eufy's RTSP streams

    • HunterBargainHunter on 20/12/2023 - 11:49

      Eufy recognition is not too bad (when it works)
      What added value would this be compared to stock eufy app?

      • jamhos on 20/12/2023 - 11:51
        +1

        I've been using an existing Eufy camera with Frigate and Double Take / Compreface. Main advantages are speed (Eufy can be slowwww), local (just because I prefer it), and with Double Take - facial recognition. So I can disable alerts if it's "me".
        But I think you can do facial recognition with the new Eufy homebase? So maybe less of a difference if you have that.

        • iLikePizza on 20/12/2023 - 12:07

          Hows the quality difference with the whole 1080p RTSP thing.

          Actually what cameras are you using?
          Im using the 2k indoor pan and tilt outside.

  • monky on 20/12/2023 - 11:23
    +5

    Thanks OP I was thinking about getting this full price earlier in the week to enable hyperboost on my Wooter v4.2 instance. Gotta figure out how to pair it with my external Fantopro Gigabit first haha. Expecting to pull an extra 25 Gigasharts with this baby 🔥👌

  • samyall on 20/12/2023 - 11:29
    +7

    Word of advice if you do want to buy this: RS sometimes ships with Aramex in Sydney.

    Do yourself a favour and request shipping with Aus Post or Toll so your package actually arrives.

    • monky on 20/12/2023 - 11:32

      Never had an issue with Aramex in Sydney but I'll keep that in mind now ty

      • samyall on 20/12/2023 - 11:37
        +2

        You must be one of the lucky few then.

        It's worth reading their reviews if you ever need to feel better about yourself cause at least you know you are doing better than Aramex.

        • pscac001 on 20/12/2023 - 11:45
          +1

          I found the staff at Aramex super helpful. One of them spent almost 2 hours running around the depot tracking down my package.

          Of course this was after it got to the point where I was thinking "Sure, spending 2 hours sitting at the Aramex depot seems like the best way to get my package". I wasn't the only one there and I'm not sure I was even the one who had to wait the longest.

          • easternculture on 20/12/2023 - 17:39

            @pscac001: It was either that , or be bossed around by his supervisor. Id rather just waste 2 hours.

    • elitistphoenix on 20/12/2023 - 12:11
      +1

      Thanks. Done. Aramex in Brissy is the same.

    • clartan2015 on 20/12/2023 - 23:06

      received update from RS, mine's being shipped with Courier Please

      • elitistphoenix on 21/12/2023 - 12:30

        Same. Almost as bad as Aramex. Guessing I won't see this for a while.

  • wogman on 20/12/2023 - 11:29
    +9

    I'm wondering if this will help speed up my old turbo encabulator?

    For a number of years now, work has been proceeding in order to bring perfection to the crudely conceived idea of a transmission that would not only supply inverse reactive current for use in unilateral phase detractors, but would also be capable of automatically synchronizing cardinal grammeters. Such an instrument is the turbo encabulator.

    Now basically the only new principle involved is that instead of power being generated by the relative motion of conductors and fluxes, it is produced by the modial interaction of magneto-reluctance and capacitive diractance.

    The original machine had a base plate of pre-famulated amulite surmounted by a malleable logarithmic casing in such a way that the two spurving bearings were in a direct line with the panametric fan. The latter consisted simply of six hydrocoptic marzlevanes, so fitted to the ambifacient lunar waneshaft that side fumbling was effectively prevented.

    The main winding was of the normal lotus-o-delta type placed in panendermic semi-boloid slots of the stator, every seventh conductor being connected by a non-reversible tremie pipe to the differential girdle spring on the “up” end of the grammeters.

    The turbo-encabulator has now reached a high level of development, and it’s being successfully used in the operation of novertrunnions. Moreover, whenever a forescent skor motion is required, it may also be employed in conjunction with a drawn reciprocation dingle arm, to reduce sinusoidal repleneration.

    • ebenezer7 on 21/12/2023 - 15:59

      In short, no it won't.

Login or Join to leave a comment
Login Join