EmbeddedRelated.com
Forums
Memfault State of IoT Report

Atmel Bought by Microchip

Started by rickman June 28, 2016
On 7/1/2016 6:14 PM, antispam@math.uni.wroc.pl wrote:
> In comp.arch.embedded Paul Rubin <no.email@nospam.invalid> wrote: >> antispam@math.uni.wroc.pl writes: >>> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >> >> $0.528 at Mouser in qty 10000. $0.89 in qty 100. Qty to get it at 0.25 >> must be awfully large. >> >> http://www.mouser.com/ProductDetail/STMicroelectronics/STM32F030F4P6/ > > $0.25 is from STM advertising. Yes, there is large difference compared > to prices that small buyer gets from established places. But the > same is with products from other manufactures. I got mine from > Aliexpress seller, $0.44 in quantity 10, $0.43 in quantity 100. > Of course there is risk of fakes, but at least there is some > hope that Chinese are reselling from a big lot bought at > manufacturer price.
I wish. I could get some parts really inexpensively, but my contract assembler won't guarantee the units if I use any of these potentially crap parts. -- Rick C
On Fri, 01 Jul 2016 13:33:07 -0700, the renowned Paul Rubin
<no.email@nospam.invalid> wrote:

>antispam@math.uni.wroc.pl writes: >> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. > >$0.528 at Mouser in qty 10000. $0.89 in qty 100. Qty to get it at 0.25 >must be awfully large. > > http://www.mouser.com/ProductDetail/STMicroelectronics/STM32F030F4P6/
STM32F103C8T6 must be < $0.70 USD, based on what I see from China, but it's $3+ in 1K from Digikey. --sp -- Best regards, Spehro Pefhany Amazon link for AoE 3rd Edition: http://tinyurl.com/ntrpwu8
In comp.arch.embedded rickman <gnuarm@gmail.com> wrote:
> On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: > > In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: > >> On 6/29/2016 1:59 PM, Les Cargill wrote: > >>> Phil Hobbs wrote: > >>>> > >>>> Some Cortex M0s are under half a buck at distributor prices, which is a > >>>> lot cheaper than an ATmega. > >>>> > >>> > >>> This is also true. I haven't had the privilege yet. > >> > >> Not sure if this is a valid comparison. CM0s under a dollar are very > >> limited devices. I can't imagine there aren't AVRs with similar > >> capabilities at similar prices. > > > > Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. > > In small quantities it is cheaper than ATmega 328. This ARM > > processor in some aspects is smaller than ATmega 328: it has only > > 20 pins, 16kB of flash. Also has no analog comparator and no > > support to connect low frequency crystal. OTOH 12-bit ADC > > running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of > > RAM for many tasks make it more capable than ATmega 328. > > > > I think that for tasks which STM32F030F4P6 fits well it is > > hard to find replacement in AVR line at comparable cost. > > And there are bigger models in the same series, still > > under 1 dollar. Also, while limited I would not call > > them "very limited": there are currently selling 8-bitters > > which much more limited. > > Pin count is a huge driver in low priced devices. In a tour of one of > Atmel's manufacturing facilities it was pointed out that testers are > very expensive machines and often the cost of a chip is driven by tester > time. So a package with fewer pins will nearly always cost less than a > package with more pins. > > Try to compare apples to apples in the package department. It can be > very important. >
1) IIRC the bigger version, STM32F030K6 which has 32 pins and 32kB of flash was still cheaper than ATmega328. 2) It seems that manufactures put more features into chips with more pins. So it may be hard to find AVR chip with 20 pins and other features comparable to the CM0 chip. Actually on average I consider the F030F6 to be more powerful chip than ATmega 328, but that is subjective -- some aspects of F030F6 are much better, few other (like number of pins) are worse. I have limited familiarity with AVR product line but supect that several features of F030F6 have no match in low cost AVR-s. -- Waldek Hebisch
On Fri, 1 Jul 2016 16:11:26 -0400, rickman <gnuarm@gmail.com> wrote:

>On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: >> In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: >>> On 6/29/2016 1:59 PM, Les Cargill wrote: >>>> Phil Hobbs wrote: >>>>> >>>>> Some Cortex M0s are under half a buck at distributor prices, which is a >>>>> lot cheaper than an ATmega. >>>>> >>>> >>>> This is also true. I haven't had the privilege yet. >>> >>> Not sure if this is a valid comparison. CM0s under a dollar are very >>> limited devices. I can't imagine there aren't AVRs with similar >>> capabilities at similar prices. >> >> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >> In small quantities it is cheaper than ATmega 328. This ARM >> processor in some aspects is smaller than ATmega 328: it has only >> 20 pins, 16kB of flash. Also has no analog comparator and no >> support to connect low frequency crystal. OTOH 12-bit ADC >> running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of >> RAM for many tasks make it more capable than ATmega 328. >> >> I think that for tasks which STM32F030F4P6 fits well it is >> hard to find replacement in AVR line at comparable cost. >> And there are bigger models in the same series, still >> under 1 dollar. Also, while limited I would not call >> them "very limited": there are currently selling 8-bitters >> which much more limited. > >Pin count is a huge driver in low priced devices. In a tour of one of >Atmel's manufacturing facilities it was pointed out that testers are >very expensive machines and often the cost of a chip is driven by tester >time. So a package with fewer pins will nearly always cost less than a >package with more pins.
Pin count is a driver more because the package is more expensive than because of test time. I/Os are usually tested in parallel. High pin count testers are more expensive but it's not the major driver.
>Try to compare apples to apples in the package department. It can be >very important.
Absolutely. One of the ARMs (M7) I'm using costs about 25% more in a 100-ball BGA than a 100-lead QFP. OTOH, the cost of another (QFN-M0) is basically $X + $Y*leads + $Z*flash (Z is really small).
On 7/1/2016 8:17 PM, krw@attt.bizz wrote:
> On Fri, 1 Jul 2016 16:11:26 -0400, rickman <gnuarm@gmail.com> wrote: > >> On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: >>> In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: >>>> On 6/29/2016 1:59 PM, Les Cargill wrote: >>>>> Phil Hobbs wrote: >>>>>> >>>>>> Some Cortex M0s are under half a buck at distributor prices, which is a >>>>>> lot cheaper than an ATmega. >>>>>> >>>>> >>>>> This is also true. I haven't had the privilege yet. >>>> >>>> Not sure if this is a valid comparison. CM0s under a dollar are very >>>> limited devices. I can't imagine there aren't AVRs with similar >>>> capabilities at similar prices. >>> >>> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >>> In small quantities it is cheaper than ATmega 328. This ARM >>> processor in some aspects is smaller than ATmega 328: it has only >>> 20 pins, 16kB of flash. Also has no analog comparator and no >>> support to connect low frequency crystal. OTOH 12-bit ADC >>> running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of >>> RAM for many tasks make it more capable than ATmega 328. >>> >>> I think that for tasks which STM32F030F4P6 fits well it is >>> hard to find replacement in AVR line at comparable cost. >>> And there are bigger models in the same series, still >>> under 1 dollar. Also, while limited I would not call >>> them "very limited": there are currently selling 8-bitters >>> which much more limited. >> >> Pin count is a huge driver in low priced devices. In a tour of one of >> Atmel's manufacturing facilities it was pointed out that testers are >> very expensive machines and often the cost of a chip is driven by tester >> time. So a package with fewer pins will nearly always cost less than a >> package with more pins. > > Pin count is a driver more because the package is more expensive than > because of test time. I/Os are usually tested in parallel. High pin > count testers are more expensive but it's not the major driver.
Not sure how you can test I/Os in parallel when one of the things you need to test is that they aren't shorted. All I know is from the guy who worked there and had the testers. I seem to recall he went through the math verbally showing that you need to test low priced parts pretty damn fast!
>> Try to compare apples to apples in the package department. It can be >> very important. > > Absolutely. One of the ARMs (M7) I'm using costs about 25% more in a > 100-ball BGA than a 100-lead QFP. OTOH, the cost of another (QFN-M0) > is basically $X + $Y*leads + $Z*flash (Z is really small). >
-- Rick C
On Sat, 2 Jul 2016 22:42:29 -0400, rickman <gnuarm@gmail.com> wrote:

>On 7/1/2016 8:17 PM, krw@attt.bizz wrote: >> On Fri, 1 Jul 2016 16:11:26 -0400, rickman <gnuarm@gmail.com> wrote: >> >>> On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: >>>> In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: >>>>> On 6/29/2016 1:59 PM, Les Cargill wrote: >>>>>> Phil Hobbs wrote: >>>>>>> >>>>>>> Some Cortex M0s are under half a buck at distributor prices, which is a >>>>>>> lot cheaper than an ATmega. >>>>>>> >>>>>> >>>>>> This is also true. I haven't had the privilege yet. >>>>> >>>>> Not sure if this is a valid comparison. CM0s under a dollar are very >>>>> limited devices. I can't imagine there aren't AVRs with similar >>>>> capabilities at similar prices. >>>> >>>> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >>>> In small quantities it is cheaper than ATmega 328. This ARM >>>> processor in some aspects is smaller than ATmega 328: it has only >>>> 20 pins, 16kB of flash. Also has no analog comparator and no >>>> support to connect low frequency crystal. OTOH 12-bit ADC >>>> running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of >>>> RAM for many tasks make it more capable than ATmega 328. >>>> >>>> I think that for tasks which STM32F030F4P6 fits well it is >>>> hard to find replacement in AVR line at comparable cost. >>>> And there are bigger models in the same series, still >>>> under 1 dollar. Also, while limited I would not call >>>> them "very limited": there are currently selling 8-bitters >>>> which much more limited. >>> >>> Pin count is a huge driver in low priced devices. In a tour of one of >>> Atmel's manufacturing facilities it was pointed out that testers are >>> very expensive machines and often the cost of a chip is driven by tester >>> time. So a package with fewer pins will nearly always cost less than a >>> package with more pins. >> >> Pin count is a driver more because the package is more expensive than >> because of test time. I/Os are usually tested in parallel. High pin >> count testers are more expensive but it's not the major driver. > >Not sure how you can test I/Os in parallel when one of the things you >need to test is that they aren't shorted.
You write checkerboards. For peripherals, operate more than one at a time.
>All I know is from the guy >who worked there and had the testers. I seem to recall he went through >the math verbally showing that you need to test low priced parts pretty >damn fast!
Sure but not the point. I'm not saying that test isn't important but it doesn't scale linearly with I/O. It's pretty trivial to check a hundred I/Os for opens and shorts.
>>> Try to compare apples to apples in the package department. It can be >>> very important. >> >> Absolutely. One of the ARMs (M7) I'm using costs about 25% more in a >> 100-ball BGA than a 100-lead QFP. OTOH, the cost of another (QFN-M0) >> is basically $X + $Y*leads + $Z*flash (Z is really small). >>
On 7/3/2016 12:46 AM, krw@attt.bizz wrote:
> On Sat, 2 Jul 2016 22:42:29 -0400, rickman <gnuarm@gmail.com> wrote: > >> On 7/1/2016 8:17 PM, krw@attt.bizz wrote: >>> On Fri, 1 Jul 2016 16:11:26 -0400, rickman <gnuarm@gmail.com> wrote: >>> >>>> On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: >>>>> In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: >>>>>> On 6/29/2016 1:59 PM, Les Cargill wrote: >>>>>>> Phil Hobbs wrote: >>>>>>>> >>>>>>>> Some Cortex M0s are under half a buck at distributor prices, which is a >>>>>>>> lot cheaper than an ATmega. >>>>>>>> >>>>>>> >>>>>>> This is also true. I haven't had the privilege yet. >>>>>> >>>>>> Not sure if this is a valid comparison. CM0s under a dollar are very >>>>>> limited devices. I can't imagine there aren't AVRs with similar >>>>>> capabilities at similar prices. >>>>> >>>>> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >>>>> In small quantities it is cheaper than ATmega 328. This ARM >>>>> processor in some aspects is smaller than ATmega 328: it has only >>>>> 20 pins, 16kB of flash. Also has no analog comparator and no >>>>> support to connect low frequency crystal. OTOH 12-bit ADC >>>>> running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of >>>>> RAM for many tasks make it more capable than ATmega 328. >>>>> >>>>> I think that for tasks which STM32F030F4P6 fits well it is >>>>> hard to find replacement in AVR line at comparable cost. >>>>> And there are bigger models in the same series, still >>>>> under 1 dollar. Also, while limited I would not call >>>>> them "very limited": there are currently selling 8-bitters >>>>> which much more limited. >>>> >>>> Pin count is a huge driver in low priced devices. In a tour of one of >>>> Atmel's manufacturing facilities it was pointed out that testers are >>>> very expensive machines and often the cost of a chip is driven by tester >>>> time. So a package with fewer pins will nearly always cost less than a >>>> package with more pins. >>> >>> Pin count is a driver more because the package is more expensive than >>> because of test time. I/Os are usually tested in parallel. High pin >>> count testers are more expensive but it's not the major driver. >> >> Not sure how you can test I/Os in parallel when one of the things you >> need to test is that they aren't shorted. > > You write checkerboards. For peripherals, operate more than one at a > time.
Do you know what is adjacent to what inside the chip? Even in a memory a checkerboard is not sufficient.
>> All I know is from the guy >> who worked there and had the testers. I seem to recall he went through >> the math verbally showing that you need to test low priced parts pretty >> damn fast! > > Sure but not the point. I'm not saying that test isn't important but > it doesn't scale linearly with I/O. It's pretty trivial to check a > hundred I/Os for opens and shorts.
No one said it was linear, just that it is monotonic. To check I/Os takes a *lot* more than checking for shorts. They need to verify the data sheet info. -- Rick C
On Sun, 3 Jul 2016 03:08:04 -0400, rickman <gnuarm@gmail.com> wrote:

>On 7/3/2016 12:46 AM, krw@attt.bizz wrote: >> On Sat, 2 Jul 2016 22:42:29 -0400, rickman <gnuarm@gmail.com> wrote: >> >>> On 7/1/2016 8:17 PM, krw@attt.bizz wrote: >>>> On Fri, 1 Jul 2016 16:11:26 -0400, rickman <gnuarm@gmail.com> wrote: >>>> >>>>> On 7/1/2016 4:03 PM, antispam@math.uni.wroc.pl wrote: >>>>>> In comp.arch.embedded rickman <gnuarm@gmail.com> wrote: >>>>>>> On 6/29/2016 1:59 PM, Les Cargill wrote: >>>>>>>> Phil Hobbs wrote: >>>>>>>>> >>>>>>>>> Some Cortex M0s are under half a buck at distributor prices, which is a >>>>>>>>> lot cheaper than an ATmega. >>>>>>>>> >>>>>>>> >>>>>>>> This is also true. I haven't had the privilege yet. >>>>>>> >>>>>>> Not sure if this is a valid comparison. CM0s under a dollar are very >>>>>>> limited devices. I can't imagine there aren't AVRs with similar >>>>>>> capabilities at similar prices. >>>>>> >>>>>> Hmm, I am using STM32F030F4P6. Advertised price in quantity is $0.25. >>>>>> In small quantities it is cheaper than ATmega 328. This ARM >>>>>> processor in some aspects is smaller than ATmega 328: it has only >>>>>> 20 pins, 16kB of flash. Also has no analog comparator and no >>>>>> support to connect low frequency crystal. OTOH 12-bit ADC >>>>>> running up to 1Mps, 5 16-bit timers, DMA chanels and 4kB of >>>>>> RAM for many tasks make it more capable than ATmega 328. >>>>>> >>>>>> I think that for tasks which STM32F030F4P6 fits well it is >>>>>> hard to find replacement in AVR line at comparable cost. >>>>>> And there are bigger models in the same series, still >>>>>> under 1 dollar. Also, while limited I would not call >>>>>> them "very limited": there are currently selling 8-bitters >>>>>> which much more limited. >>>>> >>>>> Pin count is a huge driver in low priced devices. In a tour of one of >>>>> Atmel's manufacturing facilities it was pointed out that testers are >>>>> very expensive machines and often the cost of a chip is driven by tester >>>>> time. So a package with fewer pins will nearly always cost less than a >>>>> package with more pins. >>>> >>>> Pin count is a driver more because the package is more expensive than >>>> because of test time. I/Os are usually tested in parallel. High pin >>>> count testers are more expensive but it's not the major driver. >>> >>> Not sure how you can test I/Os in parallel when one of the things you >>> need to test is that they aren't shorted. >> >> You write checkerboards. For peripherals, operate more than one at a >> time. > >Do you know what is adjacent to what inside the chip? Even in a memory >a checkerboard is not sufficient.
Oh, good grief/ The designer knows! Have you ever done VLSI test?
> > >>> All I know is from the guy >>> who worked there and had the testers. I seem to recall he went through >>> the math verbally showing that you need to test low priced parts pretty >>> damn fast! >> >> Sure but not the point. I'm not saying that test isn't important but >> it doesn't scale linearly with I/O. It's pretty trivial to check a >> hundred I/Os for opens and shorts. > >No one said it was linear, just that it is monotonic. To check I/Os >takes a *lot* more than checking for shorts. They need to verify the >data sheet info.
I/O is a trivial part of test. Think!
> ... Intel has been making noise about > getting back into the embedded market, > where they once had a huge presence with the 8051 product line
Intel dropped the ball when they gave up the xScale ARM line to focus on x86 architecture from Pentium to Atom to Quark. Intel is trying to force themselves into the embedded market with the Galileo Arduino-like platform with the Quark Chip, and the Edison SoC (system on chip). But I see nobody really adopting that despite familiarity and comfort with the x86 architecture. Intel does not seem to collaborate with others. All the other chip makers collaborate or are consortiums, particularly ARM since it's fabless: the core is licensed to others. The latest Intel show-n-tell I attended seemed rather arrogant: they want the entire vertical market from embedded portable device to router to server and everything in between. No need to play nice with others.
On 18 Jul 2016 17:02:46 -0400, jeffj@panix.com (Jeff Jonas) wrote:

>> ... Intel has been making noise about >> getting back into the embedded market, >> where they once had a huge presence with the 8051 product line > >Intel dropped the ball when they gave up the xScale ARM line >to focus on x86 architecture from Pentium to Atom to Quark.
Intel hasn't done *anything* right except X86 (and there is some argument here, too).
> >Intel is trying to force themselves into the embedded market >with the Galileo Arduino-like platform with the Quark Chip, >and the Edison SoC (system on chip). >But I see nobody really adopting that despite >familiarity and comfort with the x86 architecture. > >Intel does not seem to collaborate with others. >All the other chip makers collaborate or are consortiums, >particularly ARM since it's fabless: the core is licensed to others.
You got that right. Every time I've had a presentation from Intel it's been by the most arrogant people imaginable. Their pricing model is based on X86, as well. A model that doesn't allow their partner to make a profit.
> >The latest Intel show-n-tell I attended seemed rather arrogant: >they want the entire vertical market from embedded portable device >to router to server and everything in between. >No need to play nice with others.
Yep. That's all I've seen from them, too, and I'm not in the IT market.

Memfault State of IoT Report