Page 1 of 1

How the chip works

Posted: Sat Mar 26, 2016 3:33 pm
by Pito
Hi, an interesting info on how the GD duo-die-chip works. That info also explains how they achieve a zero flash wait-states..
http://zeptobars.com/en/read/GD32F103CB ... ga-Devices

Re: How the chip works

Posted: Sat Mar 26, 2016 5:22 pm
by martinayotte
Interesting !

Thanks for the link ...

Re: How the chip works

Posted: Mon Mar 28, 2016 12:09 am
by ahull
Die has 110 pads, 9 of which are used by a flash die. GD32F103CBT6 is in TQFP48 package - which again suggests that this die is universal and also used in higher pin count models.
So the GD device may well have additional features on the die... very interesting. :D

Re: How the chip works

Posted: Fri Sep 30, 2016 1:59 am
by Sylvan_YZY
so they prepared 128kib extra sram for code that barely can reach this size in most projects, what a simple clean tidy waste solution i like them :(

Re: How the chip works

Posted: Fri Sep 30, 2016 3:11 am
by RogerClark
Its a shame there does not seem to be a way for the code to access the program space as RAM - so if the program was small e.g 32k you could use the remaining flash / RAM as RAM.


I know a few people were investigating whether there was a way to do this, but I don't think anyone found a way to do it

Re: How the chip works

Posted: Fri Sep 30, 2016 11:11 am
by Pito
I bet there is a magic sequence you can access the ram.
Maybe we have to ask the vendor directly on that :idea:
The flash is a standard 8pin serial SPI flash the company produces afaik, so it seems they just added a small chunk of code into the perpetual bootloader area which loads the ram from the flash upon hw reset. They also spent a spare SPI most probably for that (but bitbanging from unofficial hidden pins might be used too - see the picture - the flash pins are bonded to the pads so with a better photo you may simply fingerpoint the pin numbers).
Also writing to flash is redirected to the flash.
Maybe messing with a SPI_X not "officially" available for the specific chip may return some code?? A SPI_N+1 maybe :)
Or bitbang some hidden pins may show some activity??
Can we somehow read out the perpetual bootloader code ??
You can also sniff the Vcc line during the boot and decode the flash->ram boot sequences :)

Re: How the chip works

Posted: Fri Sep 30, 2016 11:58 am
by RogerClark
It may be worth contacting Greg (sorry I can't remember his username), as he has contacts in GD.

Re: How the chip works

Posted: Fri Sep 30, 2016 12:42 pm
by Pito
100% sure they have chips without the SPI flash mounted in, but with the flash pins exposed/active. So they develop and play with standard packages. It would be great to get such GD32 chip handy for experiments..
Imagine Fram/Mram instead of the flash. Unbelievable combo :ugeek: