Espressif Releases ESP8266-Killer!

It’s no secret that we love the ESP8266 chip, and the community of hackers that have contributed to making it useful. We often joke about this or that new WiFi-enabler being an ESP8266 killer, but so far none have stepped up. Here we go again!

Espressif has released a chip that’s going to be an ESP8266 killer, and no, it’s not the ESP32. The ESP8285 went into mass production in March, and should start to appear in the usual outlets fairly soon.

What makes it an ESP8266 killer? It’s an ESP8266, but with the flash memory onboard. Nothing more, but also nothing less. What does this mean? Tiny, tiny designs are possible. And, if the street price ends up being right, there’s no reason you wouldn’t opt for built-in flash. (Unless you were planning on doing some ROM hacking.)

We’ve found a datasheet, a development board with schematics on GitHub, and another blog-post about the chip. AI Thinker apparently has a module already (pictured).

Has anyone out there had one of these in their hands yet? Any idea where you’d order some?

Thanks [deshipu] for the tip!

135 thoughts on “Espressif Releases ESP8266-Killer!

    1. It looks more compact (which is pretty nice), but I wonder where the wifi antenna will go. I don’t see one above, and while it’s possible this is a module intended to be soldered onto a board with an antenna, I also wonder what it will be like trying to fit an antenna onto the smaller devices with will enable.

      1. Yes I do. I wish I could make it work on my modem emulating firmware too. So far I can’t get my ESP to “hang up” from the computer once connected. I have to get the remote end to drop the connection.

  1. Look, if nothing else has changed, it will be not a killer. Just a new thing that will slowly replace the ESP.
    I mean, how many projects with the ESP can you think of where the size of other modules wast the problem? Also, expect this to be more expensive.

      1. Re: “killer” and clickbait. I’ll admit some guilt there, but my honest opinion is that I would pay a small premium to get significantly more memory than an 01-style module in less than 1/4 the size.

        Did the mammals kill the (large) dinosaurs? Are we talking about semantics? My guess: it _will_ replace the ESP8266 for a number of us. Hence, killer.

        1. IF this tiny package is as programmable as the ESP8266, AND can run more efficiently than existing ESP8266 modules… then yes, 8266-killer, and an enabler of tinier, battery-powered gadgets. Otherwise, for most of us hackers, the current crop are small enough; the necessary batteries are the size limitation.

          But, gonna get me some just the same, uh-huh.

        2. To really be a ‘killer’ it would have to be an FCC certified module with integral antenna. Making the antenna external eliminates all the space savings and mandates much more testing to release a product. Those points alone will relegate this module to a subset of uses, but it enables long distance communication at the same time.

    1. Well, yeah. Wouldn’t be much of a headline if it said “Company Introduces Slightly Different Version of Existing Product,” would it?

      Also, of course it will be more expensive. I can’t imagine it not being more expensive. I can imagine it being $1 or less per module more expensive, though, and that’s something. 1MBit of SPI flash costs about $0.40 in single quantity.

      1. i was thinking the reason why the made it without flash was to minimize cost, if flash in their process was more expensive than flash in external chips. But simplicity can also be a good reason.

    1. Seems to be ~comparable to a US$0.25 coin (about 25mm diameter, and 1 Yuan is ~$0.15), which in turn is about 11% of 2 Euros in value. Canada also has similarly scaled physical representations of currency, though differently valued vs the euro.

      I wonder what a chart of current US$/mm^3 (or euros per mm^3, just normalized to some single currency) of various denominations of various currencies around the world would look like. Is European currency size vs value smaller than average, or is North American larger?

      1. The chart would probably need to include the metal value also, as euro coins are magnetic. They are just plated scrap metal with no intrinsic value. A chart like that will expose the fibre of the fabric for each particular currency.

        1. Euro coins, in US cents per cubic millimeter (1 Euro = 1.1252US$):
          1 cent: 0.00325
          2 cents: 0.00488
          5 cents: 0.00950
          10 cents: 0.01903
          20 cents: 0.02705
          50 cents: 0.05118
          1 euro: 0.11375
          2 euros: 0.19642

          This uses your rough formula for the volume. Taking the composition of the alloys, the approximate diameter of the inner part, and the thickness of its layers into account, a 2 Euro coin should have a volume of ~970mm^3 in contrast to the 1154mm^3 calculated from its diameter and maximum thickness.

          http://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32014R0729&rid=1 Annex I

      2. Some coinage systems have significantly larger values in regular use. As an American, it surprised me when I lived in the UK how much value a handful of change might represent (though to be honest with larger/heavier coins, for the most part). Here in the US ‘pocket change’ is extremely unlikely to contain dollar coins or 50-cent pieces in most places, making a quarter the largest common value represented. A tip jar likely only contains 5-10 dollars if it’s all coins, and your spare change likely wouldn’t be enough to buy even a basic lunch. In the UK with 1- and 2-pound coins in regular daily use (6x and 12x the US quarter, in rough value), I often found myself carrying 5-10 dollars in metallic coinage in my pockets, and the one time I ended up collecting donations for an event the bucket had probably $30-50 in value, despite looking somewhat thin.

        1. On the average, unlikely, given how many people look at it and me strangely when I use one, but if I have any pocket change at all it likely includes at least one dollar coin. No sense in having one dollar bills (nor pennies). I’d be in favor of $5 coins, too, and dropping the $5 paper. I can hardly imagine carrying around those enormous half-dollar coins, though, particularly since they’re only worth two quarters.

  2. ESP8285 Onboard Memory size: 1MB. Most ESP8266 modules come with a 4 MB flash chip. It should be noted that this larger size is needed for OTA updates. This new chip won’t be a killer.

      1. Nope, he didn’t confuse megabits and megabytes. Most modules have 16 megabits or more. 512kb is the realm of old esp-01 modules (and you don’t really want those anyway)

        The link you posted is over a year old.

          1. Indeed… I mean how terrible would the code have to be or tons of garbage would you have to be including to not be able to do wifi + an update system in 1Mb. Which according to that blog posts really only takes about 256KB which leaves 768KB free on the ESP8285. 256KB still sounds like alot to me but I guess that SSL is being done in software or something..

          2. 256KB is a LOT. I can do high speed ADC, a bunch of DSP, including a full DFT, process colors notes, include a webserver running high performance websockets with MDNS and outputting to WS2812’s over I2S and it’s STILL got space to spare!

  3. Could be ESP8266 killer, if Expressif decides to stop making ESP8266. Otherwise, it can be nice to have a smaller chip, but normally, in hoppy projects, not required. So i think i will stick with ESP8266, until ESP32 is released, unless these 8285 are cheaper. Not that i use lots of them anyway.

  4. Sure, but where is my ESP that takes a lot less than 300ms to calibrate? Where is the ESP that can connect in less than half a second to an AP? And the one that can wake up from pin change in sleep? Or the one that actually sleeps with <10uA as specified in the datasheet? Or the one that keeps the RAM while sleeping?
    Those are things that could kill the ESP….

    1. “Or the one that actually sleeps with <10uA as specified in the datasheet?" — That's deep-sleep and it's already there.

      "Or the one that keeps the RAM while sleeping?" — Regular RAM isn't kept, but the RTC-memory does retain its contents even during deep-sleep. It's not much, but it's enough for some things.

      1. i know the <10uA is in the datasheet, but I have not found any module that eats less than 20uA.

        Keeping all the RAM has the advantages of starting from where you left after sleeping instead of going through all the power up sequence.

          1. Not really. You can also send the sleep command to flash chip prior going to deep sleep. I am really within 9uA (well, as far as my crappy technology can measure) while the chip is in deep sleep.
            I am waking it up every 3 minutes for last ~7 months no problem. My only problem with the setup is it consumes so small amounts of energy that I am constantly overcharging the small LiPo I have there :(

          2. You may be right about the flash chip. However, really, the 20u is not the problem, that is still 10 years on a pair of AA. The real problem is the long startup time & calibration along with the long time required to get IP and send some bits of data.

    2. I agree with Bogdan. Lower power consumption, faster wake-up and connect times, more gpios, these are the kind of improvements that would be beneficial for embedded platforms (especially battery operated). And better and more accurate (I would say “real”) specs and hardware features.

      Not more processing power or bundling with BT… If you need more processing power you are designing your system the wrong way.

      But, the ESP8266 might have its market if used just as a wireless module in conjunction with other application CPU. Your remote firmware updates would be stored on the application board and not inside the ESP8266.

        1. Sit.Up is a great idea, actually. It could become a common feature built into office chairs…:D you could configure it for your sit-up needs and your employer could check on you as well….

          I currently have an ESP12 managing 30 multiplexed analog sensors, 2 digital inputs, and an i2c slave mcu that performs external WDT, wake-up alarms, and power management (cheaper than any dedicated chip). And the ESP8266 handles 99% of the work…
          However not having enough gpios available on the ESP8266 forces me to use multiplexers and io expanders, and because of its higher power consumption, I am using an ultra-low power microcontroller to keep everything powered off unless and until there is a need for more horsepower and/or wifi uplink. Battery life about 12 mo.

          I would like to see a next gen ESP SOC with at most 5mA avg current consumption while idle and associated with AP, and 50 uA or less in light sleep…. Not thrilled for lots of flash as currently IPv4/UDP, CoAP server, and APP layer(s) all occupy less than 1 MByte of the flash.

          1. yes, it would be good to have it on every chair, but i don’t really see it happening. Activity trackers are becoming cheaper and people would probably prefer them.

            AS for the ESP, since you already have a micro around, just beef it up a bit to skip all those io expanders and throw in a more low power radio. RFM69 might be a cool device.

  5. So, over 3 months since that announcement and they haven’t released them yet…
    How many, do you think they test for major or minor production flaws?
    1% or .5% maybe even only .01% doesn’t seem like it would take this long for a larger batch to get ready to sell.

  6. According to the datasheet, it has 1 megabyte (not megabit) of flash. If that is correct it could replace the ESP8266 at the right price. Even so it is an incremental improvement.

      1. You’re thinking of megabits per second (Mb/s) which does include time. The onboard flash memory can store one megabit, or ~2^20 bits, of program and other data at any given time if I read this correctly.

      2. Memory chip sizes and data transmissions are both measured in megabits. On one side because they’re usually teamed together in groups of 8 or multiples, and on the other side because network links are generally serial.

        1. Kinda successful? I mean, all of the responses were respectful, and simply corrected your error. No heated debate? No flame-war started? Just an apparent mistake politely fixed?

          Nah. Fail troll.

          But a tremendous win for the HaD comment brigade!

    1. Effort on the documentation would have been better received I think. Like why does the 8266 boot into some stupid 70k serial mode when Tx is pulled low at boot. What’s with the stupid 70k data rate anyway and why do they sometimes take 1/4 of an amp and do nothing when they should have booted.

      I could go on, it’s a long list and all with no explanations that I can find.

  7. Wifi is perfect for hacking, not good for anything else because of the unreliable signal propagation. I predict the IoT revolution will be driven by a combination of Wifi and unlicensed band products (308Mhz and 433Mhz) Not Zigbee and all those unreliable bands.

    1. I predict it will not happen. Even if some magical new standard comes along that makes sense (maybe the 802.11ah), nothing stops them from not having open API or letting 3rd party interact with them.

      1. Nope, not 802.11ah will not be it. Some people really do want the IoT revolution to happen though. They will pontificate long and hard, searching for a solution. Then one day like Newton with his Apple someone will finally figure it out… that will be 802.11aha.

    2. My guess is that it really depends on what the IoT device is actually doing. For simple comsumer goods, like your “smart coffee machine” that already has a fixed power connection, WiFi is probably the best way to go. For more industrial or portable/wide area applications, i think the LoRa-Standards could see a lot of adoption, once the network has been deployed (from what i hear, the telco-companys seem to be setting up the basics for LoRa, at least where i live). Of course, that depends on how they organize access to the LoRa network and how much it costs to deploy your device into that network…

          1. Ehh. Sprite_tm, this is the second time in this discussion that you refer to “we” that could mean ESPRESSIF or possibly someone integrating the CPUs onto boards…. Where do you work (now)?

  8. These ESP8285 chips can be purchased directly from Espressif for $1.10 each delivered to the US. They are as easy to use as the ESP8266 (identical, really). I was disappointed that the six GPIOs normally used for the external flash were not made available for other uses but the internal flash does allow smaller designs. The breakout board I designed with the FTDI chip really makes using the ESP8266/85 much more convenient than fussing with a FTDI connector. Anyway, the 1 Mbyte internal flash is plenty for most applications and I have started using this chip exclusively in my own projects.

    1. “I was disappointed that the six GPIOs normally used for the external flash were not made available for other uses…”

      I suspect the new part is a Chip Carrier – so the GPIO previously used to connect to the external flash are still used for the same purpose, but now the flash chip is inside the carrier along side the ESP8266 chip. The “new” hybrid part is badged ESP8285.

      1. Not “Chip Carrier”, more likely multi-die flip-chip or wire bonding into a single package onto a common substrate. But you are right: the result is just that the external Flash is now integrated into the same package. Someone will have to decapsulate one of these to make confirm!

        However, is it sure that the six GPIOs cannot be shared withe the SPI Flash as before?

    2. Hi Kris,
      sure? 1.10 USD ?
      how much you must take from it? ( quantity )
      i remember, that i paid a little more as 1.10 USD and from quantity, the 1.10 USD is not in the price group list.( its expensive )
      not sure, where you bought. espressif have a direct line ( Betty example ) and they are verry friendly for make a first tier relation price if buyer ask for industry. but honest, 1.10 USD is not my industry knowledge. i know lees price.
      i speak from prices from chn to europe, perhabs USA has other price status.

      best wishes
      rudi ;-)

  9. Actually a true killer would be something that is more industrial and widely used. Checkout KNIT that uses a Marvel Semiconductor Wifi chip MW300 and supports AWS with rich peripherals. Costs about the same in features with popular NodeMCU boards. I heard it can even do multiple sockets. Xiaomi and Hello Barbie use the ‘exact’ same chip.
    https://makerville.io/knit/

  10. The ESP8285 is available directly from http://www.espressif.com. They have a store site but it is in Chinese mostly, but send an e-mail asking to buy them. I bought 100 of the ESP8285 and 100 of the ESP8266 and they delivered same to my house in California for $210 paid via paypal. They also offered to throw in a Wroom module (to entice me to buy more I suppose); I felt it rude to decline their offer even though I don’t like to use the modules (too big).

    Thanks all for the sales response for the ESP8285 Development Board on Tindie! I am all out now but have ordered more pcbs to hand build a few more while I wait for the first 100 unit production batch of the ESP8285 Development Boards from China. I hope to drop the price as much as I can with increasing sales volume. Thanks for your purchases, interest, and patience!

    1. -> They also offered to throw in a Wroom module (to entice me to buy more I suppose); I felt it rude to decline their offer even though I don’t like to use the modules (too big).

      ?
      this is new to me.
      never get this same offer :)
      if so – i would take the ESP8285 Launcher Board ( WROOM-(8286) ) with the esp8285
      have the lauchner with esp8266 and its fine to develop with it, the modul you can snap in/out.
      so i think i will try to update the module by ESP8285 in reflow next times, too.

      ..i must write betty for the offer you get, i will the same

      ;-)

      1. ESP8286 is wrong – sorry:
        EDIT: if so – i would take the ESP8285 Launcher Board ( WROOM-(8286) ) with the esp8285
        TO: if so – i would take the ESP8285 Launcher Board ( WROOM-(8266) ) with the esp8285

  11. I’ll take a look at breaking out pins 9 and 10. Somehow when I read the data sheet I got the impression these pins were not useable; and espressif confirmed this but i suppose I can try and see for myself.

      1. I tested this on my board. I was able to blink an led with pin 9 but not pin 10, not sure wht. Might have been a cold solder joint or maybe using a wire jumper as test lead isn;t precise enough. Anyway, I redesigned the ESP8285 Development Board to break out pins 9 and 10 also and will test it before mass production. Here is the design if anyone wants to build their own:

        https://www.oshpark.com/shared_projects/ZI9PQPVe

        I should have this version available in small quantities in a few weeks time.

  12. hi Elliot

    -> Has anyone out there had one of these in their hands yet? Any idea where you’d order some?

    just in time, got the esp8285 :)
    production is from 052016,

    [fun]
    if you spend “a hendl and a mass radler ” :)
    in chin. garden, i will drive 65 km to you :)
    and give you a packages
    [/fun]

    http://esp8285.de
    ( content is at work next time for datasheet and more )

    modules are on the way just in time too.
    not sure, when it arrival exactly – will update infos then on the web.

    best wishes
    rudi ;-)

Leave a Reply to macegrCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.