there’s 1MB (1024kB) of flash instead of 512kB !!! THIS… is a nice surprise, the ST-Link Utility insists about 512kB but with stm32flash.exe it not only reads 1024kB but i managed to upload a 950kB jpg image, verified during upload, downloaded, renamed from .bin to .jpg and voila… the same image (i just didn’t tried to run it on the MCU
)
stm32flash also states 128kB of RAM but they say that 64kB are mumbo jumboed somehow (cache?), so at the moment no pain
also, some sites like RS Components declare the VET6 “1024kB”, and a chinese store, but, anyways..
The F103C8 is known to have 128kB flash when the spec says there should only be 64k
I know some other people have tested the larger F103 MCU’s and they all have more flash than in the spec, but I can’t recall if anyone else noticed the F103 having more RAM
I don’t think the F103Cx has more RAM than in the spec, but the actual amount of RAM or Flash seems related to what STM call the “Density” of the product.
It looks like MCU’s in the same series with the same “Density”, tend to have the Flash and RAM that is available in the best MCU in that group ( Series and Density)
It may be worth searching the forum to see if anyone has done RAM testing on other STM32 MCU’s and not just testing the flash
BTW. I edited your title to make it easier when people are searching for a specific MCU type
[RogerClark – Fri Aug 25, 2017 1:35 am] –
The F103C8 is known to have 128kB flash when the spec says there should only be 64k
both my blue pills have 128k, they look the same but bought 1-2 years apart
also an ARM MCU with 128kB of flash IS worth buying, in my opinion
[RogerClark – Fri Aug 25, 2017 1:35 am] –
more RAM
i shoulded state that the f407vet is supposed to have 192kB of RAM, it displays actually less, but according to the datasheet 64 kb of it is managed differently, so this is likely the reason why it displays less
[RogerClark – Fri Aug 25, 2017 1:35 am] –
It looks like MCU’s in the same series with the same “Density”, tend to have the Flash and RAM that is available in the best MCU in that group ( Series and Density)
that’s a pretty good theory, but it’s strange that the part of the MCU that costs more (i think) is “given for free”, this may be related to the country of origin of these chips (i’m “familiar” with fake components also)
[RogerClark – Fri Aug 25, 2017 1:35 am] –
BTW. I edited your title to make it easier when people are searching for a specific MCU type
never mind, my fault, i thougth that the text of the topic was enough for searching
AFAICT, TI was the first to reuse the same mask for all the chips with the same core and not telling about it int MSP430, although you can tell by the datasheet, as power consumptions are the same for all chips!
IT looks like ST is playing the same game, and the STM32F407xE are actually the sme as STM32F407xG having 1MB of Flash memory and 192 KB of SRAM (64 KB of it is Core-Coupled Memory, only accessible through the cache).
ST may (or not) qualify the chips for their announced capacity, so you take your chances when using this extra memory.
I asked ST about fake STM32’s and they were not aware of any, so I think these are ST made components, its just not worth the hassle of zapping the RAM on boards which have more RAM etc than they are supposed to.
This is really only of benefit to hobbiests etc and also to companies who sell cheap products e.g. the STLink clones from China which use a F101 which is not supposed to have USB but does have USB.
Anyone using their MCU’s commercially, on a large scale, could not really take the risk of using a lower spec product only to suddenly find in a new bach, that they no longer had the additional RAM or Flash
But its great that we can make use of it as 1Mb would be great for camera video frame grabbing etc

About the flash I am almost certain I had checked at least 1 of my VET boards and couldn’t write past 512KB, but I will recheck again.
I wonder if that “VG” marking on the top of the chip is any telling of what mask they used to manufacture it…
[Squonk42 – Fri Aug 25, 2017 5:22 am] –
Making a mask for achip is expensive, thus manufacturers only use as few as possible. OTOH, marketing is pushing for having an orthognal matrix with pin counts and memory size as axis, in order to create a range that they can charge more or less tot he customer…
sounds pretty logic
[RogerClark – Fri Aug 25, 2017 5:42 am] –
F101 which is not supposed to have USB but does have USB
USB can be faked AFAIK…. see the USBasp programmer (that i have, it uses non signed drivers and also sometimes isn’t recongnised by USB )
[ag123 – Fri Aug 25, 2017 4:12 pm] –
chips that do not pass the qa for 1m are marked as ve
logic also… not only this, the ST-Link Utility INSISTS about the blue pill having 64kB and the VET having 512kB i really don’t know where this identifier is apart the “stm32f4xx\stm32f40x” identifier, but the program CAN read the entire memory, not programming it beyond “official” size (the ST-Link Utility has a lot of bugs BTW)
[victor_pv – Fri Aug 25, 2017 4:27 pm] –
I wonder if that “VG” marking on the top of the chip is any telling of what mask they used to manufacture it…
[BulbChangeExpert – Fri Aug 25, 2017 4:35 pm] –[victor_pv – Fri Aug 25, 2017 4:27 pm] –
I wonder if that “VG” marking on the top of the chip is any telling of what mask they used to manufacture it…
![]()
there are VG and VQ … from both Korea and Philippines (the PHL on mine or KOR on the internet (or China (CHN) and Malaysia (MYS) on my blue pills)), that’s what google images displays
[BulbChangeExpert – Fri Aug 25, 2017 4:55 pm] –
there are VG and VQ … from both Korea and Philippines (the PHL on mine or KOR on the internet (or China (CHN) and Malaysia (MYS) on my blue pills)), that’s what google images displays
I have at least 3 or 4 F4 by now, I think I checked the flash size in one of them, by trying to read and write past 512KB. In the F1 stlink will write past the reported address if the mcu actually has more, but perhaps that fails in the F4 and the loader is able to read and write to it?
it apparently can calculate, read and write the real memory without taking into account IDs and stuff, you can even burn past a bootloader (specifying a start address)
but the ST Link Utility (the actual GUI program) can read the whole memory, it can even (un)protect all the sectors, if you set “Size” at 0x100000 it will read the whole, you can’t modify the bonus part but if you try to input something like 0x100100 it wil not read past 0x100000

Not sure how I checked the other time, but perhaps I did something wrong.
Nice to know there is a whole 512KB extra in those.
I also checked CCM RAM and SRAM, and could only read 64KB and 128KB respectively on both boards.
The costs of chip’s cutting/bonding/assembly/testing/packaging are much higher than the production cost of a Silicon wafer (the chip foundries charge usually flat fee for a wafer) therefore messing with many chip types is much costly than to mess with a few one. The single chip (die) on a 300mm wafer could cost $0.25-$0.5 (based on the STM32 chip size), while you get 3k-6k chip dies (working) off the wafer.
The Masks (or “the chip tapeouts”, the masks are used for the UV lithography process, you need 20-36 of them today in order to produce an MCU/CPU) are the most expensive production cost item – they have to be refreshed from time to time and their production costs are high (a set of masks for 90nm node – ie STM32F407ZG – may cost about $0.5-1mil).