SET provides computing power solutions
for AI and blockchain
SET's MAC computing network is seamlessly integrated with our EC2 instances, helping us reduce computing latency by 15% during peak AI reasoning tasks. This sustainable computing power source is a perfect fit for AWS's green cloud computing initiative.
When training large-scale language models, computing power requirements grow exponentially. SET's distributed MAC computing power pool allows us to flexibly schedule additional computing resources at a cost 22% lower than traditional GPU clusters. We plan to integrate their computing nodes on Google Kubernetes Engine (GKE).
Low-latency computing power is crucial in RLHF experiments. SET's global MAC node network enables us to quickly deploy small inference clusters in different regions, significantly reducing data round-trip latency. SET computing power model may become an important part of future AI training infrastructure.
Traditional ASIC mining machines face high energy consumption and heat dissipation issues, and SET's MAC computing network provides an environmentally friendly alternative. After testing, we found that SET's equipment is 18% more energy efficient than traditional mining farms on the SHA-256 algorithm, and there is no need for additional cooling costs. This is very attractive for mining pools.
Outside of CUDA optimization scenarios, many parallel computing tasks can be efficiently processed by distributed CPU clusters. SET's MAC computing network proves this. Their equipment performs well in some matrix operations and may form a complementary computing ecosystem with NVIDIA's Grace Hopper super chip in the future.
When deploying private AI models for customers, sudden computing power demands often lead to budget overruns. SET's on-demand leasing model allows enterprises to temporarily call on thousands of MAC devices to avoid long-term investment in fixed hardware. This model has been verified in the financial and medical industries, saving an average of 34% in computing power costs.
Blockchain confirmation + machine learning load monitoring, patented heat dissipation solution extends equipment life by 30%
SET's patented heat dissipation solution extends the life of the equipment by 30%, and its energy conversion rate reduces the power consumption per T of computing power by 3.2 times compared to traditional mines.
AI training and crypto mining face a growing computing power gap. SET can provide flexible computing power gap supplementation.
Leading the design of Google's distributed computing platform architecture, managing 2000+ node clusters. PhD in Computer Science from Stanford University, focusing on edge computing resource scheduling Responsible for overall strategy and partner ecosystem at SET.