Microchip To Acquire Atmel For $3.56 Billion

Just last week, there was considerable speculation that Microchip would buy Atmel. The deal wasn’t done, and there was precedent that this deal wouldn’t happen – earlier this year, Dialog made an approach at Atmel. Now, though, the deal is done: Microchip will acquire Atmel for $3.56 Billion.

There are three main companies out there making microcontrollers that are neither ancient 8051 clones or ARM devices: TI’s MSP430 series, Microchip and Atmel. Microchip has the very, very popular PIC series microcontrollers, which can be found in everything. Atmel’s portfolio includes the AVR line of microcontrollers, which are also found in everything. From phones to computers to toasters, there’s a very high probablitiy you’re going to find something produced by either Atmel or Microchip somewhere within 15 feet of your person right now.

For the hobbyist electronic enthusiast, this has led to the closest thing we have to a holy war. Atmel chips were a little easier (and cheaper) to program, but were a little more expensive. Microchip’s chips have a very long history and proportionally more proper engineers who are advocates. PIC isn’t Arduino, though, a community that has built a large and widely used code base around the AVR family.

Microchip’s acquisition of Atmel follows several mergers and acquisitions in recent months: NXP and Freescale, Intel and Altera, Avago and Broadcom, and On Semiconductor and Fairchild. The semiconductor industry has cash and wants to spend it. What this means for the Atmel product line is left to be seen. The most popular micros probably won’t be discontinued, but if you’re using unpopular Atmel micros such as the ATtiny10 you might want to grab a reel or two before they’re EOL’d.


183 thoughts on “Microchip To Acquire Atmel For $3.56 Billion

          1. Atmel Studio actually works very well with its supported parts, once they got the initial bugs worked out when they made the AVR studio to Atmel Studio transition. Its nice to be able to use one familiar environment for both 8 bit and ARM projects.

          2. Yeah. I spotted a bug in Atmel Studio’s includes of one of their less common AVR8 which prevented the use of some of the IO. You could still use it if you knew the port addresses though.

            I hit the AVRFreaks forums with a copy of the include file to override the defaults. By the next version, my include started complaining and it turns out Atmel implemented my changes!

    1. MPLABX works fine on amd64/x86 Ubuntu (debian), but I’m still annoyed they charge a license fee for the full version of legacy pic18/pic16 C compilers.

      If one has plenty of flash memory, than the gcc unoptimized output on Atmel chips will work for most platforms.
      Also note, the newer PIC32 series also has the gcc as it’s compiler (again, they charge for the full version), but both don’t compare to the performance of most modern $5 ARM SoC.

      I don’t think the technology will improve for either product line with the merger, as Microchip has been dragging its feet for years while transitioning toward the gcc standardized compiler suite.

      1. But the Microchip PIC compilers are crippleware. The CEO of microchip has openly stated in the past that they’ll never offer a high performence C/C++ compiler for free.

        I hope Microchip doesn’t kill off the AVR, as that will push everybody onto an ARM solution.

        1. I would dare say that Atmel’s attitude helped make Arduino the success that it is. If Microchip changes that attitude, then most probably we’ll see the next Arduino built around an LPC1114. :)

          1. I think Microchip has chosen the right time. Arduino is on it’s death knell now. I don’t know if you have used it recently.

            After the Adruino vs Arduino fiasco the latest version of the IDE has everything locked up. Installing libraries is via URL and if you try to install *YOU OWN* libraries just by adding them to the libraries folder then the IDE will tell you that *YOUR* libraries are not approved and can’t be used. So now I have totally thrown it out. What’s the point of an IDE if you can’t write / use your own libraries.

            The same goes for different board configs. I used to use it for almost anything ATmega. Favorites were the ATmega644 and ATmega1284. In the new IDE I can’t make the changes I need so that I can use these chips because they’re not approved by heil Arduino!

            Well at least I learnt a bit about the GCC toolchain but otherwise Arduino is dead to me.

          2. To RÖB:
            I guess you are not using the arduino.cc, original ide since i’ve never seen it say anything about “not approved libraries”. I know maybe 1.6.2 said about that about the boards, but that was removed in the next version and as i understood it you could still use them anway. I currently have many libraries i have installed without the library manager and there’s no trouble with them. Currently have 1.6.7.

            And about adding boards/chips, i have not done that myself, but somehow people manage to add 3rd party boards anyhow, so i guess it’s doable.

          3. @[J] I used the original IDE (1.0.x) forever lol it was great.

            The new versions want’s you to only use approved boards so the software wont allow the necessary changes for custom setups and other chips.

            As far as I am concerned it no longer deserves to be called open source when it deliberately is configured for one specific retailer and won’t allow customization.

            Of course I could just use version 1.0.x as I did before but if the writing is on the wall that Arduino is going to do a reprap then I’m out of here before the whole thing collapses.

          4. To RÖB:
            Where are the limitations of only approved boards/libraries? Like i said i don’t see any limitations in the newer IDEs, so please give a specific example of this ” software wont allow the necessary changes for custom setups and other chips”. The new IDE supports ESP8266, Chipkit boards with PIC32 chips and so on and so on. Atleast before the Chipkit was found from the board manager, i installed it manually and there was no BS about not allowing unofficial boards.

            Here, have a list unofficial 3rd party boards for the board manager: https://github.com/arduino/Arduino/wiki/Unofficial-list-of-3rd-party-boards-support-urls#list-of-3rd-party-boards-support-urls

          5. @[J] The laptop I installed 1.6.x on died so I can’t be sure of the version it was. I had a look at the change log and it seems that I was probably using the previous version as there have been changes to how libraries are handled.

            In any case this what happened from their –
            I went to sign up to the forums – I had trouble with the captcha which is normal as I have an uncommon form of dyslexia. The form would drop some fields each time I got the captcha wrong and cause it to fail the next time as I didn’t notice that the fields lost the information.

            Eventually the form called me an idiot and permanently banned me from the forum before I could even sign up.

            Makes me wonder what they would have done if I was in a wheel chair? Call me pathetic and slice my arms off perhaps. OK, I will say it – this is disgusting and blatant discrimination against a person with a disability!!!!

            Now back to the install. It suggested that I copy the old version to a new folder if i wanted to keep it. Why??? I don’t know – there is no reason that they can’t both run on the same system!

            Copying didn’t work – some registry key prevented me from doing it that way and I had to do the uninstall of the older version before installing the new.

            Then as I said the libraries didn’t work. My old install was rooted by the new install and I just found it easier after the insulting response from their forum registration … easier to delete the many, many years of work that I had done with the Arduino platform and through the lot out permanently.

            Now that due to their shitty form and my dyslexia, I am banned permanently from their forums where they can edit / moderate / delete my posts and I will forevermore post my anguish at their shitty product in places that they can’t edit / delete / moderate my posts.

            And that leaves only one question …. who’s the idiot now!!!

          6. To RÖB:
            Well, ok, they are not intentionally trying to stop you from using “unofficial” libraries. It’s just that there have been changes in the environment and some libraries that worked with 1.0 versions do not work with >1.5. There could be updates to those libraries or atleast alternatives.

            I have a project that uses a LCD menu system, which does only compiles on the 1.0.x versions. I keep a 1.0.6 for that. Newer stuff i do with the latest version. Although i’m planning on changing the menu library on that project too.

            So you can have 2 versions at the same time, maybe don’t use the installer version, if you are using windows.

            Can’t say much about that forum thing. Maybe they have an email or something where you could ask, if you feel like it.

          7. @[J]

            They have written and included software to reject unapproved libraries and boards.

            It doesn’t change anything that they have unchecked the box. They wrote it for a reason so the writing is on the wall.

            As for the forum.

            Well perhaps I should write to them. I should start my email with an apology for having the genetic per-disposition that has lead to my dyslexia … you know like how black people should start their job interviews with an apology for being black.

            No offense to darker skinned people, just want to make a point and I am not prejudiced.

          8. @[J]

            I wasn’t trying to use an older 1.x.x library with 1.6.x

            If I wanted to use a 1.x.x library then I would have stuck the the 1.x.x IDE that was already installed.

            I was trying to install Trinket (Attiny85) USB library to use with a DigiSpark which is also based on the ATtiny85.

            And guess what – It seems that because Auduino doesn’t make the DigiSpark or the Trinket then it’s no go for the competition!!!! So how open is that Arduino IDE when this happens or is in the planning????

        2. @[2kas]
          I didn’t throw away years of work because I couldn’t solve a captcha. I already said that it’s extremely common that I can’t do a captcha. I never get a captcha in less then six goes and most places lock you out after 3 tries. Think about how many times you do a captcha – think about how you go through so many of them when you loose a laptop or a PC crashes and then tell me again that *I* need and anger management course. Like doing them all 6 to 12+ times wouldn’t bother you at all!

          I *did* through out years of work because they are gearing up (with the software) to lockout any boards except ones that they approve (READ: provide).

          That means there is no forward path for what I use the IDE for so it is useless to me now and I will just go back to starting with Studio and not bother to use the IDE to sort out protocols first.

          As far as I am concerned – when the heads at Arduino are dicking with the software to lock out unapproved boards (on what was previously an open platform) then it’s time to jump ship because in doing so they will be locking out the boards/libraries that I use it for ie – my very *own* libraries and board configs will be locked out!!

          1. It definitely sounds like you need to learn anger management. The only thing you might regret is not doing it sooner. Think about it, you could be using your energy constructively solving your issues right now instead of throwing hissy fits in a comment section. What is a couple of sessions compared to having an improved quality of the rest of your life? I haven’t used Arduino for a while but there is no way I believe they are locking user hardware or libraries out. They might have broken something though. And your problems with captcha, couldn´t a friend help you when you need to solve many of them when recovering from from a crash? It can’t be that often, is it? IM a screenshot of the captcha window to someone when you need to solve one quickly? Nowadays when I run into an obstacle I only see solutions, but you won’t believe how many wake up clocks and stuff I smashed when I was younger. And doors, always the doors. What is it with doors that make you always punish them for your frustrations? I am much, much happier today and get so much more done.
            Honest and sincerely

          2. @[2kas]

            You really need to read your own post in context.

            I made a rational decision to through the lot out as it is obvious to me that any time invested in that platform is going to be wasted time when they switch of support for unofficial products. If that’s not there intention then why develop the software to exactly that?

            And yes the endless captchas do get to me but get [2kas] – you telling me to get some anger management because captcha’s annoy because of my dyslexia is akin to saying to some one in a wheelchair that they just need to get out and run a few laps and get over it.

            Seriously, I didn’t read past that point because your showing some seriously shady character with that line of thinking.

          3. You seemed to be be in a rage. I sincerely tried to be helpful. The wheelchair analogy is completely wrong. I don’t say you should do something you cannot do. Just ask someone to pick an item from the top shelf if you can’t reach yourself.

      2. If Microchip are using GCC as the base for their compilers, they are legally bound by the GPL to release the source code to that (meaning unless they are mixing GCC and some other tool in a way that is permitted by the GPL and the secret sauce is in the other tool, anyone should be able to take the GCC source and compile it themselves to get the same thing without the price tag)

          1. The reason there’s little freely supported compilers for PIC8 is because the architecture is completely unsuited to compilation. Amongst the many deficiencies:

            1. No data stack, this makes local variables and recursion a virtual impossibility.
            2. Limited call stack and limited access to its contents even on ‘high-end’ CPUs.
            3. No proper pointer registers, instead I/O locations are used as pointer registers – a chronically inefficient system, reminiscent of the pdp-8 from 1965.
            4. An accumulator-based architecture, which means lots of shuffling in and out of w. It’s the opposite of what a compiler needs.
            5. 1960s style skip/goto for conditional jumps (except on some PIC18s).

            etc. etc.

            It’s just dire, dire, garbage – a perfect example of everything you shouldn’t have in an architecture.

        1. They do release the source code for their fork of gcc. Unfortunately they’ve done a nice little end-run around the GPL by adding a clause to the license of their C library and peripheral libraries that forbids you from using them with anything but the official, crippled, license-locked compiler binaries from Microchip.

    2. Well PICs are slow (4 clock cycles = one instruction), there are no good dev tools for Linux (and in general nothing really working except for MPLAB(X) which is one of the shittiest pieces of software I ever used) and their support/help forum is the worst place on earth.
      I really hope that whatever they do gets supported by gcc (like avrs) and that there will be some good IDEs that can be used…
      (and yes I know how to code for both in C and Asm. we use PICs at school and at home I use AVRs).

      1. Not true for 16-bit line (PIC24, dsPIC) and PIC32. I have to deal with different microcontrollers at my job (8 to 32 bits, bare metal or with an OS). In general, I prefer Microchip for its 16-bit product line, Atmel for its 8-bit line and ARM-32bits when I need an OS.

        1. Dont forget the PLLs on the 18Fs, 48MHz Is good enough for me…. div/4 clocks i believe where originally put in place to allow the use of cheap ceramic resonators instead of crystals.

  1. “..From phones to computers to toasters, there’s a very high probablitiy you’re going to find something produced by either Atmel or Microchip somewhere within 15 feet of your person right now….”

    That was a hard guess on a blog which is read by many hackers :)

    I don’t think anything will be EOL’d as Microchip still produces (new) stock of old models. AFAIK Atmel has less models then Microchip has, so chances are they will continue them.

          1. Dig deeper.


            Now its replaced by a 16F54.

            No internal clock.

            No internal ADC or any analog peripherals.

            Was the cheapest microcontroller for many, many years as an OTP part. Windowed parts covered in gold had to be erased in a UV light table and this could only be done a limited number of times.

  2. Source for the acquisition that the post is missing:
    >In agreeing to the acquisition by Microchip, Atmel ended its agreement with Dialog Semiconductor, paying a termination fee of $137.3 million.
    >Stockholders of Atmel will receive $7 per share in cash and $1.15 per share of Microchip stock in the deal announced on Tuesday. The Dialog agreement announced in September encompassed $4.65 in cash and 0.112 of Dialog’s American depositary shares for each share of Atmel.

    Looks like Microchip has been busy and they are not just interested in microcontrollers.

    >Chip makers have been combining to cut costs and build scale for their customers. Microchip agreed to acquire the 37-year-old chipmaker Micrel in May for about $839 million. Microchip purchased Supertex for $394 million in 2014, and the year before that, bought the closely held Brussels-based EqcoLogic for an undisclosed amount.

      1. Minecraft is huge, and Atmel is tiny, in terms of market share. While wildly popular among the hobbyist maker community, I don’t really see their MCUs in real-world products. Most products I see use 8051, PIC, S08, or H8 MCUs.

        Atmel MCUs are just too expensive for what you get. Their dev tools aren’t bad, but when you’re building 100k units annually, spending hundreds (or even thousands) of dollars for a Keil C51 compiler license is a drop in the bucket, especially when 8051 MCUs are 15 cents each, and Atmel parts are 45 cents each.

          1. Not entirely surprising – I know everybody and their dog is doing integrated on-chip capacitive touch sensing these days, but that wasn’t always so: the first kids on the block with a popular discrete touch-sensing IC were Qprox (1-2-8 key sensor ICs etc.) and guess who bought them shortly after? Atmel…

        1. You and I must be buying different electronics. I see the Atmel logo in every other thing I pull apart, be it expensive short run items, mass produced, or cheap-arse chinese crap. I’ve seen their stuff in consumer garbage and in industrial settings. I don’t think I’ve seen much of their high end 8bit stuff, I’ve never seen an ATXMEGA out in the wild, but plenty of Tinys MEGAs and ARMs.

          Also don’t under-estimate how lucrative it is to have an end-to-end automotive hardened solution. You likely sit next to many Atmel products when you get in your car not just microcontrollers but CAN devices, radio controllers, maybe even their touch controllers. Producers love it when they can reduce the number of vendors they need to vet, and providing a large range of products really helps them here.

  3. I hope that whatever happends, that GCC still will be supported. Realisticly, if I can program my atmega48 or atsamd21 in mplabx with some ICE device and still have GCC, then OK.

    So I guess after the renaming of AVR Studio to Atmel Studio it will now become MPLABX Studio?

      1. You could always ask them for the source code and remove whatever license-checking wrapper they’ve introduced. building gcc & binutils isn’t exactly rocket science (though it can be tedious the first time).

          1. Even better, Atmel uses stock arm gcc (which is maintained by ARM and everyone else). If microchip tries to sell you optimizations for an ARM core, RUN THE FUCK AWAY FAST.

        1. You could, but the license agreement for their C library and possibly also their peripheral libraries forbids you from using them with anything other than the official, Microchip-compiled, DRM-locked rebranding of gcc.

          1. Maybe someone needs to create a FOSS-licensed clone of the Microchip stuff then, just like glibc started out as a GPL clone of the proprietary Unix C library…

      2. “The best I can do for free is -O1”
        Unfortunately that isn’t even a particularly bad thing… we had the old mcc18 bought and licensed so we could have used all the “O’s” we wanted – and ended up explicitly disabling most of them in every one of our projects after stumbling upon compiler screw-ups of atomic facepalm level at the end of many hunts for some inexplicable bug…

    1. While the question is a good one for the AVR target of GCC in the long term, the SAMD21 use a ARM Cortex-M0 core, not an AVR core. I highly doubt that GCC will start to lost the ARM Cortex-M0 target before several decades.

    1. If the DoJ would do their damn job then this acquisition would be blocked. Atmel is the only company keeping Microchip honest. At this point: the Atmega328 is an educational asset and shitting all over that with Microchip’s greed is a travesty. Microchip is a dinosaur that should be acquired by Atmel if anything.

      1. Both ATMEL and Microchip are losing ground to ARM. The Atmega328 will be around for as long as people are purchasing it, just as the PIC16x84 is still available.

        As for the educational market, don’t kid yourself. 10 years ago most students were well versed in Microchip due to the cheap and easily programmable PIC16x84 which overtook the Basic Stamp (which was a PIC underneath), Arduino changed that not ATMEL. Plus as big as the original Arduino is, it’ll be a generic ARM that leads the next 10 years of the student market.

        1. Atmel is making a lot of ARM Cortex chips since a couple of year: http://www.atmel.com/products/microcontrollers/arm/default.aspx

          I for example use very successfully the SAMA5D35 (Cortex-A5) on Linux. A peripheral SAMD21J18A (Cortex-M0) is programmed with OpenOCD from the main A5 using the 2 signals SWD bus and communicate with it using full speed USB. A very powerful combinaison. Last revision of the chips have very few anomaly and none of them are making any problem with the targeted application. I found no corresponding solution with Microchip

        2. Maybe where you live that was the case. Certainly 15 years ago when I went through my degree all the educational institutions were pushing Atmel products. Cheap and easily programmable got crapped on by pushing lots of STK500s at a major discount to educational institutions, not to mention a completely free IDE.

          Most of us learnt AVRs long before the Arduio showed up, and most of us laughed at how poorly implemented much of the Arduino’s functions were. I don’t miss the days of being forced to write assembly for an AVR, but man did it make us learn the ins and outs of those damn things.

          1. We used to learn ASM and then C on the 8051 architecture. That was around the year 2000. I guess it really depends on what the teachers were familiar with. I have to say tho that i’ve never used any 8051 chips later on at work, only once i had a project where we had to replace an 8051 chip with something else because the part got cancelled, we then moved over to an M16C from Renesas. Today, basically everything is done with different application tailored Cortex M (from M0 to M4) chips, with a few PIC applications for the very low cost, low power, high volume segment and a PowerPC chip for the more high end Linux-Kernel based stuff.

      1. You can sell anything if the buyer is willing to spend that much.
        Who cares is Minecraft sold for x amount compared to Atmel?

        You’re forgetting that these things are not ‘commodities’ that are manufactured and regularly sold by various vendors.

  4. Does this mean that we’ll see Atmel’s line of unique controllers gradually get phased out? AVR8 parts reaching EOL being replaced by PIC versions? etc…

    I’m not sure if this is a good thing or just sucks. There _are_ other architectures as was mentioned in passing but they kind of suck for one reason or another (have you seen how much an eZ80 costs?). I guess this is going to let ARM fill those voids when they appear. I kind of liked programming the PIC/AVR in ASM over a relaxing weekend.

    Interesting thought…Since PIC had OOPic’s “BASIC”, and AVR had Arduino’s “C”, what’s next? GreenArray’s Forth?

    1. Unlikely they are still selling updated 8051 for legacy designs as does ATMEL.

      The microcontroller market is drifting towards ARM and I suspect that Microchip is heading that way as well but they will retain the legacy markets for quite a long time.

      As for tool design, like many organisations the accountants struggle with departments that don’t make money directly, even if it makes the product lines vastly more profitable. It’s the same with R&D in many companies, they much prefer to buy in new-tech with market analysis and profit forecasts. That’s not the whole reason, take power supplies for example. 20 years ago most companies would roll it’s own but theses days it is closer to a specialised skill set in many cases.

      1. For quick prototyping I recently changed Arduino (boards + the whole abstraction layer) for cheap STM32’s off of Ebay + mbed. It’s the same shit in different pack :D

        The comment was due to my often use of ATtiny and 16F chips when I need to make permanent projects for a small price. If they start to EOL either of these chips, I won’t be happy.

        STM8 looks like a nice chip, supported by SDCC and ST-Link/V2 works on Linux.

        1. STLink V1/V2 comes with STM8 eval board or $3 Chinese clone STLink. Hardware debugger can be used on the commercial C compilers demo or in open source tool chain via OpenOCD.

          Unlike the other 8-bit micros, RAM, FLASH (and I/O) are in same address space, so it can execute code in RAM.

    1. You can get cheap 10 pcs STM8S003 TSSOP parts for $3 with “free shipping” from China. They might not do single instruction cycle, but they are very price competitive for the lower end 8-bit chips. SDCC and OpenOCD support STM8!

  5. What are the the cheapest PIC and AVR MCUs for hackers (i.e. not buying 10,000 direct from mfr)?
    On Aliexpress I can find 50 ATtiny13a for $13, and 50 pic10f200T for $10.20.
    If there were a basic MCU one could buy for <10c ea, then all those "but you could do that with a 555 timer" arguments would become moot…

    1. I watched that video and since I have Pickit3 I haven’t used for months decided to try it. Connected it to USB port, started Pickit3 programmer and it offered me firmware update, clicked ‘yes’, and in few seconds my programmer was up to date. Didn’t have any of problems Dave talks about. So yeah, sometimes things can go wrong, sometimes it gets complicated, but I don’t think it’s a reson to give up some platform only because of that.

      1. I bought a cheep programmer and had the same problems so I gave up with it and then bought the PICkit3 in the hope that it would solve the problems.

        Not it seems that the PICkit3 is much the same. One annoying problem is that you can’t read the version number of the currently installed firmware unless you have the same version of software which you can’t achieve because you can’t read the firmware version. So once you have it working then write the FW version number in the programmer. The other annoying this is that you need the correct version of software to update the firmware so that you can use it with a different version of software.

        Dave’s situation as a little more complex as he hadn’t used / upgraded the firmware for a long time so he has to revert a lot further back with software versions.

      1. Yep – what more, Atmel didn’t that much of development last years in AVR line, they were focusing more on ARM offerings. This is true for NXP too, for example. They produced a lot of MCU with 8051 core, but they NRND-ed it in favour of ARM devices.
        All the companies are going where money are, not caring much about hobbyists.

          1. Exactly, there are plenty of devices that come in same packages as the AVR versions, like the SAMD20E is the same as ATMEGA328.
            Sure, not much through hole….but not difficult to solder.

          2. Buying eval boards and modules is a higher barrier to entry. That is, they’re OK if all you want to do is play with someone else’s evaluation board, but as soon as you want to produce something designed yourself then it becomes much harder to generate a production version.

            Consider FIGnition, the DIY 8-bit computer. I couldn’t have done this if all I had access to was SMT AVRs, because (a) the evaluation boards aren’t set up the way you need (e.g. an arduino was designed for 16MHz, not 20MHz), but also (b) no-one would be able to make from scratch – it wouldn’t really be a DIY computer. And if it’s not a DIY computer then the barrier to entry for me would be higher (because I’d have to have components pre-soldered onto the board) and not really a DIY kit. It just wouldn’t have happened.

        1. While 8051s may be NRND(comeon, its an old core), they still planned (before Microshit) to produce it until at least 2020 from my last discussions with business development managers.

      2. Say what you want about AVR, but they still have their place. I still have yet to see an ARM device that has the I/O closeness and ease of use you need in some situations. The lower end ARM stuff is better than it used to be, but it still cannot replace bare metal AVR in every application. The learning curve alone to use half the peripherals in an ARM part can make a project take far longer and be a whole lot more complicated than it needs to.

        1. Exactly. 8 bit uC’s still have a place. Same type of logic: let’s say you need a simple timed delay after pressing a button. You could do this with an 8 bit uC, but most people will opt for a simple 555 circult because you can just select a couple caps and resistors and be done with it. If you used a uC you need to write some code, program the chips, etc. Even if both solutions cost the same, it would make sense to use the simplest solution that works.

          Bump this up to the next level…you have a simple application that needs a microcontroller, if an 8 bit one works, why dig into a complex ARM CPU? It’s overkill.

    1. Why did you find the Atmel documentation “rubbish” ? I found it one of the top best practice in this industry. I especially like the fact that a single document contain all the informations, from specification, to ordering and mechanical, including all chip version errata + workaround, and documentation change history.

        1. Atmel has a single datasheet/programming guide for the smaller parts, like the tiny and the megas. Go further for things like XMEGA or ARM and you pretty much get a separation between the datasheet and the “user manual”.

    2. Atmel has had by far the best documentation in my experience. I have a feeling this is b/c maybe their docs are written by people used to writing academic papers (they read like papers), not someone assigned to do this that hates it, leaving important info scattered all over the place.

      1. The best part of their documentation system is they generate their peripheral library headers from the source used to generate documentation. It guarantees its always in sync.

        1. They also have what’s appearing to be some very wrong design guidelines we may reverse soon. I’m not some blinded fanboi. But for programmers, they have the best docs. Hands down.

  6. Sadly this acquisition is planned to only increase profitability and lowering the R&D. This is very clear in the page 5 of this document:

    The new Microchip expect to increase the Atmel gross margin of the business from 47.5% to a combined gross margin of 54.6% and even 59% in the long term. To archive this there expect to reduce the R&D from 18.4% to 14.3% and even 11.5% in the long term.

    From this point of view, the new bring not added value for the Atmel’s customers. The Atmel ARM Cortex chips family are very good for example. I use them in a number of projects and I wonder what there future will be. Because for sure, there will be dramatic change when someone pay a such big amount.

  7. I was surprised the article did not mention the importance of Open Source to Atmel and the Arduino project. It was Atmel’s very awesome (and strange at the time) support of the open source AVR-GCC compiler that caused AVRs to be used for Arduino. Without that, the project that became Arduino wouldn’t exist. If around 2005 there was another chipset with an open source compiler (that actually worked with modern chips and had first-party support, not like SDCC or similar), then the Arduino team likely would have chosen it.

    If Microchip supported an open source compile in the early 2000s, Arduino may have been PIC-based today.

    1. “If Microchip supported an open source compile in the early 2000s, Arduino may have been PIC-based today.”

      the 8-bit PIC is such a terrible architecture for a high-level language. That’s the primary reason why almost no open-source compilers (except for SDCC and only for some PICs) have been written for it.

  8. I rarely use AVRs or PICs anymore. But all of these mergers are not good for engineers, and other professionals, and ultimately not good for us hobbyists. Whatever happened to the word ‘anti-trust’? Does anyone remember when the government played a part in ensuring that companies didn’t get too big and broke up monopolies. I still remember how anti-trust lawsuits in the 90s made people like Bill Gates squeal.

    That was before corporations started to buy all the politicians and in the process undermine democracies and the rule of law all for the sake of the mighty dollar.

    So many Semiconductor mergers happened in 2015: http://www.allaboutcircuits.com/news/semiconductor-mergers-and-acquisitions-in-2015….so-far/ . Perhaps the biggest one is NXP’s acquisition of Freescale for $40 freakin Billion!

    Many of these companies cite the reason for acquiring other big companies is to ensure their own survival. But I highly doubt that. I think they have a need to keep growing at any cost. Well ….in a world made of finite resources, the term ‘sustainable growth’ is an oxy-moron.

    Having a small number of Mega-corporations instead of many just ‘corporations’ means less Engineering jobs in the west and more outsourcing to the east. It means more monopolistic proclivities, and ultimately less innovation. It also means that hobbyists and small business owners will have fewer and fewer places to source parts and will have less choice when it comes to purchasing parts and I.C.s.

    This whole thing sickens me.

  9. As the companies get bigger, our voice as hobbyists gets smaller. Soon these companies will not listen to any suggestions or bug reports unless it comes from a customer buying a million chip per year. Even if you know someone who works in these companies they will also have a tiny voice inside the behemoth and all creativity will be stomped under next quarter’s profits.

  10. pic is way faster anyhow cuz they can go all the way to 40mhz when avr can only do 20mhz sometimes on 5 volts. and pic is way more C optimized so it takes way less code bytes toy get the same thing done. I seen the xmega and it is the onyl avr kind of fast so maybe microchip will take it and make a small 8 pin out of it.

  11. ” Atmel chips were a little easier (and cheaper) to program, but were a little more expensive” Really depends where you are and who can sell them to you. For me, in europe atmel has always been the cheapest. I used the atmega8 a lot, and there was nothing cheaper from anyone I could get with the same specs.

  12. I you can make something with AtMega328 then you surely can make it with P18F2550 and vice versa. If one of them is too slow then probably other one is too. You switch to faster uC or even different platform (ARM). Binding yourself to one platform is quite limiting.

      1. Right, the 8 bit pics had 35? instructions in their set, while AVRs have 130. Stands to reason that there are some AVR instructions that need more than 1 pic instruction to get the equivalent, eating up the space and speed.

        1. AVR and PIC are quite different. The PIC has asymmetric bus widths between data and program. ie data can’t be stored in program flash and code can’t be executed from RAM.

          The program ROM/FLASH is 12 to 14 bit wide. An mediate register load consists of 4 to 6 bit of an instruction with 8 bits of data or one word.

          PIC’s were a very early stage RISC.

          Some AVR’s have self write (to Program FLASH) and execution.

          1. AVRs are the same – program memory is 16-bits wide, data memory is 8 bits wide, and they live in separate address spaces with no ability to execute code from RAM and special instructions required to read and write program memory. I believe there are a few more exotic AVR parts that map both address spaces onto the same memory but the normal AVR chips in Arduinos and the like don’t. Also, PIC definitely isn’t RISC-like, its arithmetic instructions are classic CISC-style accumulator-memory ones. AVR on the other hand is a very RISC design with register-register arithmetic on a reasonably large bank of mostly identical registers, most of which executes in a single clock cycle thanks to pipelining.

          2. Before RISC we only had CISC so is it surprising that one of the very earliest RISC CPU’s looks more *CISC like* to you or what you have come to expect of RISC.

            Sure, most of todays RISC CPU’s are different but the early PIC era is where all that started.

            The early PIC’s had very little by way of addressing modes. On a CISC CPU’s you have an almost overwhelming amount of fancy addressing like CALL (HL+IX) or call the location that is indicated by the contents of memory at a location equal to some register plus some other register.

  13. Avrs are cheap to program??? Well it depends if you are going to stick only to a couple of low-end modules. If that is not the case (which is frequently happening with me) and you need to switch between different families I prefer michochip’ programmers.

  14. I honestly dont get where all these supposed “problems” with PIC come from.
    Im not arguing on whose better… i just Dont see how or why people have such a hard time with PICs while I can whip up a project in 5 min.

    Speed or clk/4 problem? Get a PIC with a 96Mhz PLL. Div/4 helps to clean up the clock signal… its not a cripple, its good desing IMO.

    Pickit3? Never had an issue… across multiple PIC lines.
    ONCE i noticed a preprogramed DELL weather app interfered with the usb driver.

    Mplab? I have not yet figured out why people hate it…. never had a problem running my command line compiler…ever. (would like code version control, but thats not a problem)…i use V8.92, as mplabx at the time didnt play well with my compiler…ive never “upgraded”….

    Compiler? I agree Microchips compilers are aweful, atleast when i tried them. I paid for a 3rd party compiler and have never had an issue, and it integrates perfectly to mplab.

    I read the comments here and ive never had any of the reported issues above.

    I just dont understand why my toolchain works perfectly while others struggle so much… maybe because the tools where made for windoes and users above use the linux/iOS versions?

    1. It doesn’t divide the clock by 4 and then use it. It processes instructions in 4 cycles using the original clock. There’s no “cleaning up”. And it’s not a good design, because the higher clock requires more power, and causes more EMI.

      1. Regarding the current consumption:
        Atmega328PA (32k FLASH, 2k RAM) takes 12mA (=60mW) max at 8MHz delivering 8MIPS, while
        PIC18F26K22 (64kFLASH, 4k RAM) takes 9,5mA max at 5V (=50mW), or 8,5mA max at 3V (=25mW) while running at 64MHz (16MHz + PLL), delivering 16MIPS.

          1. Now you are comparing different generations of chips :-) PIC18 are old plain 8-bitters, with simple peripherals, Xmega AVR are somehow different beasts.
            No matter what, my point was that PLL isn’t significant power consumer, when running at the same speed with or without PLL, difference is in range 50-200uA, according to my measurements.

        1. The point is that the PIC would be using even less power if it where running a single cycle core at 16 MHz. You can’t invalidate that point by comparing it to a totally different device. And it’s not just the power consumption of the PLL, but also the core that’s clocked at 64 MHz.

  15. For those crying that Microchip will destroy their favorite architecture: STFU- this is about the only way that your favorites will remain around! Microchip is a peer, and has no history of acquisition to wipe out competitive parts. Hell, they work REALLY hard to not EOL parts – it is stated policy! http://www.microchip.com/mymicrochip/filehandler.aspx?ddocname=en544375 Keep buying the parts, and they will keep making them.

    What *NOBODY* here has mentioned is that Atmel is a business. It has to make money for its owners. As a business ATML has been on shaky ground for a long time. Their profits have been up and down, trending down. If you look at their stock- the P/E ratio (Price/earnings) is over 200- MCHP is about 25.

    Fundamentally, Atmel is in serious trouble. They have some good technology and good people (which I’m sure Microchip will keep). This is why they have been an acquisition target. They need somebody to fix what is broken in the business. Without this, they will fail.

        1. Bully is abusing programmers heads for decades with page switching, changing register names btw chips for no reason, needing 4 clock cycles for 1 instruction, making HW programmers obsolete, and making closed source monstrosities. If Atmel products go down, people will look elsewhere, and Microchip itself will be worse off in the future. Someone will fill the void for Atmel clones, if Microchip as much as blinks.

          1. You might not like the news, but you have no clue what bluuy is.

            It is not like Microchip come in beat up the employees with baseball bat and tell them they have to pay protection money or they don’t get to go home.

            Can’t say for Atmel, but a lot of tech companies offer golden handcuffs to their key employees so that they have a reason to stay a period of time after a merger or acquisition.


    1. what’s the difference if it a good or bad wolf is eating your sheep? I still remember that I had to rework all ifs comparison in my code because shitty pic compiler couldn’t deal with a constant on the left. It has produces overhead by loading the constant into a register first to compare it with another resister. But as far as both companies are murican companies I don’t press a shift if they exist or not

    2. >Microchip is a peer, and has no history of acquisition to wipe out competitive parts.

      They have a history of destroying support, opposing open source( and using it in ways only favorable to them and refusing to give back upstream). While the Atmel parts will remain, the community and tools will go to utter shit.

  16. a sad day indeed.
    Microchip wants everyone to have to buy $$$ programmers to be able to program their chips, and they want to be the only manufacturer or them, they want people to pay for the IDE and compilers to program their chips.
    They cant design a processor worth its weight in dirt.
    As pointed out, their community support sucks.
    I guess I’d better start buying up all the avrs I’m gonna ever need in my lifetime, unless I’m gonna start using ARM processors to flash leds.

  17. I am shocked how little discussion there is of this selling out. Sure many post here but very few forums and blogs covering this. Even here most the replies are only about the studio and programming, few of the fact that Atmel is no more.I just find it very strange. As if everyone is looking the other way.

Leave a 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.