Hardware information: ETH special mining machines are coming, and a large wave of mining cards are coming back to the market?

    Starting in June 2017, speculators took a fancy to the skyrocketing digital currency and purchased mining machines one after another. Among them, the one that has the greatest impact on players is the ETH mining machine. Because there are no dedicated mining machines for ETH and ETC, GPU mining has become the first choice for miners. That is to say, from this time on, graphics cards above GTX 1060/RX 470 and above were crazily swept away, and the price of graphics cards remained high.



    Now the good news is that this situation is about to be alleviated by the ETH mining machine that Bitmain is going to launch. According to foreign media reports such as CNBC, Christopher Rolland from Susquehanna Financial issued a report on the 26th, saying that they visited Bitmain last week and confirmed that Bitmain has developed a chip called "Application Special Integrated Circuit (ASIC)", which is specifically used for Ethereum mining and will be shipped in the second quarter of this year.

    Previously, some media said that the new mining machine will be named "F3". It has three computing boards. Each computing board uses 6 ASIC chips developed for the ETH algorithm. Each ASIC chip is equipped with 32 128MB DDR3 flash memory particles. A total of 72GB of cache is available for one mining machine. The predetermined computing power will reach 1GH/s. (The computing power of GTX 1060 6G graphics card is about 20MH/s)

    Many media have analyzed that if F3 goes public, it will have an impact on both AMD and NVIDIA. Currently, digital currency mining has brought huge profits to these two companies. NVIDA has even launched a GF series of graphics cards specifically for cryptocurrency. The impact on AMD may be much greater than that of NVIDIA. The main reason is that NVIDIA's main focus is not only mining-related businesses, but also the support of e-sports-related industries. AMD's graphics cards have significantly worse market performance than NVIDIA in this regard.

    There is a new participant in the "Huashan Sword Debate" in the field of SSD. Samsung Z-SSD has launched a model with M.2 interface. Z-SSD is the most advanced enterprise product developed by Samsung to compete with Intel Optane storage. It also focuses on ultra-high read and write and low-latency features. In January 2018, it announced the SZ985 in the form of PCIe interface and expansion card.



    According to Anandtech, at the Samsung Open Compute Project (OCP) summit, Samsung showed off the SZ985 SSD with M.2 interface on the platform.



    Interestingly, the outside world has speculated that Samsung will pile up a series of black technologies from the main control to flash memory like Intel. However, after careful identification, the main control is actually a Phoenix 8-channel product. Like PM981, it is likely to be used in the future successor of 960 Pro/EVO.

    Performance is still impeccable, supporting the NVMe protocol, using PCIe x4 channels, with capacities of 240G and 480G, and maximum continuous read and write speeds of 3200MB/s and 2800MB/s respectively.

    In terms of reliability, the hard disk writes per day (DWPD) is 30. Taking 240GB as an example, this means that during the warranty period (usually 5 years), the daily write volume can reach 30 times that of 240GB (30 times for the entire disk), which is 7200GB... It is really terrifying.

    The global memory market is mainly monopolized by three companies: Samsung, SK Hynix and Micron. Their combined share is as high as 95%. It is not easy for China to break through the blockade in the memory market. Although there are companies such as Ziguang, Changxin, and GigaDevice involved in the memory chip market, in terms of existing DDR memory technology, the technology and production capacity of Chinese companies are still far behind.

    In addition, memory is not limited to traditional DDR technology. There are many emerging fields that are developing, such as PCM phase change storage. Compared with traditional memory, it has many advantages, such as 1,000 times faster speed and 1,000 times more durability than traditional memory. Jiangsu Huai'an Times Memory Company invested 13 billion yuan to build the largest phase change storage base in China. It started official operation last week, with an estimated annual output value of 4.5 billion yuan.

    News from China Jiangsu Network stated that on March 22, the on-site promotion of Huaian's integrated circuit industry and the completion and operation launch ceremony of the phase change memory factory of Jiangsu Times Core Semiconductor Co., Ltd. were successfully held in the Huai'an High-tech Industrial Development Zone. This marked the beginning of the operational phase of the phase change memory project with an investment of 13 billion yuan. This project completed the capping of the factory building and the purchase of equipment in November last year. It took less than nine months from the start of construction to the capping, and it started operations at the end of March this year.

    According to previous news, the total investment of this project in Huaian, Jiangsu is 13 billion yuan, with a first-phase investment of 4.3 billion yuan. After completion, it will produce 100,000 12-inch PCM phase change memory chips annually. Phase change memory chips have become the memory chip standard of the 21st century. Compared with traditional memory chips (such as memory), it is a non-volatile memory chip, but its speed is 1,000 times that of traditional memory chips, and its reliability is 1,000 times that of the latter. Of course, these indicators are theoretical. The current mass-produced PCM memory chips are not so strong. Intel's 3D XPoint flash memory is said to be a PCM memory chip, but it has never been confirmed. Its advertised characteristics are that its performance is 1,000 times that of flash memory, its reliability is also 1,000 times that of flash memory, and its capacity density is 10 times that of flash memory.