前段时间坛友撸了一大堆pos机，其中最多就是gd32系列单片机（
利用一元米刷pos机内部单片机），然后今天看到了一个外国人对这芯片进行了拆解,看起来很有趣，然后就转过来给大家看看，顺手翻译了，渣翻译，大家将就看
源地址
http://zeptobars.com/en/read/GD32F103CBT6-mcm-serial-flash-Giga-DevicesGD32F103CBT6
Giga Devices was a serial flash manufacturer for quite some time. Whenthey launched their ARM Cortex M3 lineup (with some level of binarycompatibility to STM32) - instead of going conventional route of makingnumerous dies with different flash and SRAM sizes they went forSRAM&logic die and separate serial flash die. How this could workfast enough? Keep reading :-) At least
ESP8266 already taught us that executing code from serial flash and reaching acceptable speed is not impossible.
一段时间以来，兆易创新是一个串行flash生产商。当他们开始发布他们的ARM Cortex M3系列产品（某种程度上与stm32兼容）不同于传统理念，在一个晶元上集成有不同容量大小的flash 和 SRAM ，他们吧SRAM和逻辑部分做在一个晶元上，又做了一个独立的串行flash晶元。然而它能工作的足够快吗？继续阅读。至少在
ESP8266芯片上已经证明在外挂储存上实现预期速度不可能
Useof serial flash allows Giga Devices to increase maximum flash size intheir microcontrollers quite a bit (currently they have up to 3MiB) andto save quite a bit on ARM licensing fees (if they are paying "per diedesign").
使用串行flash储存允许兆易创新增加了微控制器的最大flash容量（他们甚至做到了3MB容量）同时还省了给arm公司交授权费
主芯片大局
Die has 110 pads, 9 of which are used by a flash die. GD32F103CBT6 is inTQFP48 package - which again suggests that this die is universal andalso used in higher pin count models. Die size 2889x3039 µm.
晶元有110个引脚，其中9个是用于连接flash晶元。GD32F103CBT6使用TQFP48封装。这意味着同一片晶元可能被用于更多引脚的封装上
晶元尺寸 2889x3039 µm.
Logo:
北京
ADC capacitor bank:
ADC电容列阵
After etching to poly level we clearly see that there is no flash on the die:
腐蚀聚脂层后我们看到晶元上没有设计flash
SRAM sizes are 32KiB in each largest block (128 KiB total) - storescode, which means first 128KiB could be accessed faster than typicalflash. GD32 chips with 20Kb of SRAM or less have no more than 128KiB offlash, so all flash content is served from SRAM. This might also meanthat startup time is slower than one would expect. With this SRAMmirroring it is not surprising that GD32 is beating STM32 in performanceeven on the same frequency and loosing in idle & sleep powerconsumption. Consumption at full load is lower than STM32 due to better(smaller) manufacturing technology.
每个最大的单元的SRAM是32KiB(总共128 KiB)-储存代码？？？？意味着第一个128kb的flash能比正常水平快很多。gd32的片上20kb的SRAM 或者不超过128KiB的flash储存是直接从SRAM 中提取的，这意味着他启动时间比一个周期慢。因为这种SRAM的镜像，这就不足为奇的知道在同等工作频率下的掉电模式或者休眠模式下，GD32比 STM32 的功耗更高。即使更先进的制造技术比stm32更节能（这句不懂对不对。。求大神带。。）
2 smaller blocks are 10KiB each and are likely to be user-accessible SRAM.
两个小模块每个10KiB，像是使用者SRAM
4 smallest blocks closest to the synthesized logic are 512B each.
最靠近逻辑单元的4个最小的模块每个512B
SRAM has cell size 2.04 µm², which is ~110nm. Scale 1px = 57nm:
SRAM尺寸是 2.04 µm²,是 ~110nm，比例尺1px = 57nm:
Standard cells:
标准逻辑单元
Low power standard cells:
低电压保持单元
Flash die:
flash晶元（外挂）
尺寸1565x1378 µm.
Flash die size: 1565x1378 µm.
PS. Thanks for the chips go to dongs from irc.
话说这大神设备挺牛逼的，照片养眼
渣英语不容易，大家包容，有m币尽管砸
