The need for high-performance computing (HPC)
In today's data-driven world, increasingly businesses across industries are embracing high-performance computing (HPC) to fulfill the demands for processing highly complex tasks and large volumes of data in real-time to gain deep insights and drive precise and fast decisions. HPC not only can process complex inspections; in addition, HPC is deployed to run engineering simulations to replace kinds of physical testing across different industries like automotive and medication to accelerate the development and time to market.
The HPC requires high-end CPUs, multiple PCIe slots for accelerator cards to offload the processing, and excellent heat dissipation and stability for processing capability much more than ever. Along with extremely high computing speed to rapidly process and analyze massive amounts of data, an effective HPC also needs highly reliable storage that can keep pace with the computing power. Axiomtek provides three industrial HPC, the GPU workstation iHPC300, the premium double-processor EATX server board IMB760 and the ATX server board IMB700.
When a large volume of complex data needs to be processed
Automated Optical Inspection (AOI) has become one of the most deployed applications in AIoT. In average cases, an edge system that features Intel Core processor plus consumer-grade discrete GPU, frame grabbers and IP cameras can do the work for inspecting steel or PCB defects. But if the user utilizes the same system and configuration for a much more sophisticated workload like wafer inspection, the speed and precision may be far from the standard. In wafer inspection, the tiny and multiple kinds of defects and probes in detail generate a much larger volume of data. Therefore, it is how HPC showcases its value. High-end GPUs, FPGAs or AI accelerator cards are required for the system to boost its computing capability; the system even needs to reach load balancing. As parallel computing includes many GPUs and CPUs to gain efficacy, the user will need the HPC with a superior CPU with better processor base frequency and much more PCIe lanes and bandwidth to incorporate the accelerator cards to fully unleash the computing capability. All in all, upgrading to HPC is a must to reach the best precision and speed at the same time.
Axiomtek’s iHPC300, a 4U rackmount HPC, is ideal for this type of application and demands. The iHPC300 is powered by the 3rd Generation Intel® Xeon® Scalable processors and has three PCIe x16 and three PCIe x8. Since the system is for high computing workloads, it has six DDR4-3200 non-ECC/ECC un-buffered R-DIMM slots with a total capacity of 384GB. To keep the big amount of data safe and sound, the six SATA3 with RAID 0/1/5/10. But the devil is in the details. The engineers separated the PCIe x16 slots with PCIe x8 slots in between. As the three PCIe x16 slots do not line up side by side, the user can install three high-end dual-slot GPU or FPGA cards for the best possible performance to process the most volume of data. Meanwhile, it will not cover and leave any other PCIe x16 slots unused.
Furthermore, the system integration of iHPC300 is excellent; the heat dissipation and power supply are reliable. It is a trustworthy industrial HPC with good after-sales service and compatibility. Axiomtek also provides an ATX motherboard, IMB700, which has the same spec as the iHPC300. It is literally the iHPC300 without a chassis. As a motherboard, it gives more flexibility for minor changes of design for the end user with special demands.
Solution Advantages:
- Compatible with high-end dual slot GPUs or FPGAs for high performance
- Stable system with excellent heat dissipation
- Highly flexible ATX motherboard for customization
High-Performance computing enables complex simulation
The HPC has maximized its value in enabling simulations that are too costly or time-consuming to be done physically. It has been turning physical testing into simulations on computers across industries and fields, such as medicine and automotive. With its deployment, simulation run by HPC partially replaces the sophisticated and expensive physical testing and tremendously reduces costs, time to results, and warranty and maintenance costs. While high-level testing becomes virtual and more accessible, it creates groundbreaking innovations.
A global biopharmaceutical company has been simulating the body’s response to the drug when developing new medicine. The number of patients involved in human clinical trials and the need for blood samples has all been reduced; as a result, HPC has saved time and money for the company while lessening the burden on patients.
Similarly, to optimize high-end sports car designs, the computational fluid dynamics (CFD) technology environment was moved from on-premises to HPC to perform more complicated simulations and lessen the time to resolve the turbulent wake that slows down the top speed. When the physical testing turns to simulation, it has more scalability, and the simulation cycle can be reduced.
When developing self-driving cars, the engineering simulations can switch between several environments like poor weather or rushing pedestrians and test if all the sensors and the car control function as expected in different scenarios.
The previously mentioned cases all required complicated processing to simulate scenarios based on real-world physics principles. To get peak performance from the computer cluster, we need parallel computing across several CPUs and GPUs.
For this extent of workloads and speed requirements, Axiomtek presents the flagship platform, IMB760; the EATX motherboard can provide the highest performance among all the products from Axiomtek. The design fully utilizes the space, so in addition to the dual 3rd gen Intel® Xeon® Scalable processors (Ice Lake-SP), it also features 16 288-pin DDR4-3200 RDIMM for up to 1TB of memory, four PCIe x16 and two PCIe x8 slots to off load the workloads to be “high demanding” ready. In the near future, Axiomtek will also develop the system based on the IMB760 to satisfy the demands for sophisticated applications.
Solution Advantages:
- Excellent AI performance with multiple accelerator cards
- Top data transmission speed with two 10GbE ports
- Great customization potential with EATX motherboard
