Blockchain: The nature of many cryptocurrencies is that blocks must be discovered by running a hash algorithm; the more blocks are discovered, the more difficult it becomes to discover them. Increasing difficulty has led to an "arms race" in computing power, often resulting in ASICs outperforming CPUs and GPUs. For example, Bitcoin was originally mined on CPUs and GPUs, but around 2013, the first Bitcoin ASICs were produced, which ran the SHA-256 hashing algorithm used by Bitcoin faster than general-purpose chips. much more efficient, thus rendering CPU and GPU unsuitable for this function. Today, Bitmain is the global leader in blockchain ASIC design, production and hardware deployment, and may surpass Nvidia in revenue in 2017. The market has become so hot that even Samsung, the world's largest chip supplier, is producing ASIC chips specifically for cryptocurrency mining. Bitmain doesn't just design and produce hardware, though. The company operates some of the largest data centers in the world, all of which are stuffed with its own ASICs, which it uses to mine cryptocurrency before selling the ASICs to resellers and other miners. Bitmain is now putting its ASIC expertise into the field of artificial intelligence and is preparing to enter the machine learning as a service (MLaaS) market to compete with solutions from giants such as AWS and Google.
ASIC miner refer to mining machines that use ASIC chips as the core of computing power. ASIC is the abbreviation of Application Specific Integrated Circuit, which is an electronic circuit (chip) specially designed for a specific purpose. There are mining machine factories that have designed ASIC chips specifically for calculating Bitcoin’s SHA256 algorithm, and the mining machines that use them are ASIC mining machines. Because ASIC chips are only built for specific calculations, their efficiency can be much higher than general-purpose computing chips such as CPUs. For example, the current mainstream Antminer S9 is an ASIC mining machine, using 189 ASIC chips, with a computing power of 13.5TH/s and a power consumption of only 1350W. For comparison, the computing power of the current flagship GTX1080Ti computer graphics card for mining Bitcoin will basically not exceed 60MH/s even if it is well optimized.
The large-scale application of ASIC mining can increase the cost of 51% attacks by 2,000 times. At the Unitize conference on July 10, Rod Garratt of the University of California, Santa Barbara presented research he co-authored with Maarten van Oordt of the Bank of Canada. The report investigates the different costs of a 51% attack based on the type of equipment used to protect the Bitcoin network. Research shows that security can be greatly improved simply by switching the network to 100% ASIC mining. The main reason is that ASIC mining machines have little use and value outside of Bitcoin mining, and attackers cannot get much return from selling the equipment used in attacks. Therefore, in order to conduct a profitable attack, they would need to spend twice as much Bitcoin, making it much more expensive and difficult to upgrade. The study estimates that if ASIC mining machines are 100% in place, 157,000 to 530,000 Bitcoins will be needed to attack and profit after the next halving.
An ASIC is a silicon chip designed for a very specific purpose and is intended to perform a repetitive function very efficiently; in contrast, a general-purpose chip can perform a wide variety of functions but with less efficiency (such as a GPU or CPU). Today ASICs are used in private data centers, public clouds and networking equipment around the world.
The following examples illustrate how ASICs are supporting the future of the IT industry today:
Machine Learning: Google’s Tensor Processing Unit (TPU) is an ASIC designed to run key deep learning algorithms as part of the TensorFlow machine learning framework. Google initially used GPUs and CPUs to train machine learning models, but has since developed a new generation of TPUs designed to both train and run models. TensorFlow is an open source machine learning library developed by Google that runs best on TPUs, but can also run on CPUs and GPUs.
IoT “edge” devices: Driving the digital revolution are the circuits embedded in smart devices. IoT devices themselves often use custom ASICs to reduce physical space on the chip and run with low energy consumption. In addition, some IoT devices are connected to cloud platforms such as AWS IoT Core, TensorFlow or Google Cloud, which themselves can run ASICs. In this way, IoT devices use ASICs to collect data from sensors, push this data into existing algorithm models running on the cloud-based ASIC, and send alert information or other results from the models back to the end user, or simply Feed the model to more accurately predict future outcomes.
Multi-cloud: Enterprise IT supports a variety of applications from social media to sports events to automated teller machines (ATMs). Enterprise IT needs to be viewed holistically as a multi-cloud environment. Today's digital enterprises rely on a combination of public cloud, private cloud and on-premises hardware. As part of this environment, the ASIC can reside on-premises or in a cloud environment. ASICs are already present in multi-cloud through MLaaS, and many enterprises are already using this technology.