Project Background
BYD was founded in 1995, initially focusing on battery production before evolving into a diversified high-tech enterprise spanning the automotive, electronics, and new energy sectors. Within the automotive industry, the company stands as a pioneer in new energy vehicles, having launched multiple electric and plug-in hybrid models sold globally. Beyond vehicle manufacturing, BYD also provides electronic manufacturing services (EMS), producing components for numerous renowned international brands.
Addressing Pain Points
High-intensity computational tasks
Requires execution of high-intensity numerical computations, machine learning tasks, and other demanding operations, necessitating equipment capable of rapidly processing complex calculations and rendering to meet the needs of data-intensive parallel processing tasks.
Memory storage bottleneck
Complex scenes and model editing/rendering require smooth operation, demanding at least 192GB of high-speed RAM and 2x 4TB SSDs to ensure fast read/write speeds.
Excessive operating noise
The operating environment is an office setting where workstation noise sensitivity is high. Under full load conditions, noise levels must not exceed 50 decibels. Both heat dissipation and quiet operation must be prioritized.
Solution
192 multi-core workstations
Select tower workstations supporting two AMD EPYC™ 9004 series processors to deliver multi-core, high-frequency computing power, meeting the demands of efficiently handling complex computational and rendering tasks.
High-Speed Memory Storage Solution
Equipped with 192GB+ high-speed RAM and 2x 4TB solid-state drives, ensuring smooth rendering of complex models and rapid response for software launches and data read/write operations.
Fully Liquid-Cooled Thermal Solution
Featuring a fully liquid-cooled design capable of handling a 1000W thermal design power, it balances effective heat dissipation during high-load operation with low noise levels (≤50 dB at full load).
Project Revenue
Through deployment of the Shuju Hongxin HW7425 workstation, the R&D department has achieved a significant boost in computing power. Training of A1 models and validation of deep neural networks have accelerated substantially, drastically shortening the R&D cycle compared to the original solution. This effectively accelerates technology iteration: 192GB high-speed memory and 2x4TB solid-state storage ensure smooth processing of complex models, with data sketching efficiency exceeding expectations: At full load, the device operates at 55 decibels, ensuring high-load stability and office environment compatibility. Long-term operational costs and labor losses are reduced, significantly improving the return on R&D investment.
Customer Reviews
By adopting the Shuju Hongxin workstation, we have made significant progress in our battery simulation project, not only shortening the R&D cycle but also strengthening our technical support capabilities.