Simulation Platform Handling Calculation from Day to Hour

Chip Simulation is an increasingly important part of semiconductor manufacturing. It is especially true to the global leading semiconductor chip manufacturer, GlobalFoundries who needs to perform chip design verification and tests, which require a large amount of storage and computing resources. GlobalFoundries found that the existing chip simulation platform handling a typical calculation task could not meet the service requirements and affected the chip R&D, testing, and delivery.

The New Products Trend poses Great Operational Pressure

With the consumption of electronic products trending towards mobile devices, the demands for small-size and low-power-consumption chips are rising, which calls for higher chip-making technologies. In response, major global chip foundry giants start to develop 16 nm and 14 nm fabrication processes. However, the huge investment on advanced fabrication plants (for example, ten billion dollars for a new plant) poses great operational pressure on chip foundry enterprises.

As a global leading semiconductor chip manufacturer, GlobalFoundries has always invested heavily in advanced technologies. It acquired the IBM chip manufacturing business unit in 2014 to strengthen its expertise. Its 12-inch and 8-inch wafer fabrication plants in Singapore are the most important R&D and manufacturing bases of GlobalFoundries. Facing tough challenges from new fabrication processes and techniques, the Singapore plants were grappling with improving chip simulation efficiency.

Existing Simulation Calculation Platform cannot address the services requirement

Each time after receiving a big contract, the chip foundry enterprise need to perform chip design verification and tests, which require a large amount of storage and computing resources. The existing simulation platform, however, cannot meet the service requirements. The simulation platform has a number of drawbacks. For example, it takes about one day to complete a typical simulation calculation task on the simulation platform. This seriously affects the chip R&D, testing, and delivery. Secondly, the key components, such as processors and memory, of the existing servers have limited scalability, which restricts the expansion of computing and storage capability. The aged server architecture cannot meet service requirements of the simulation platform. Moreover, the chip simulation software provides poor concurrent capability, and its performance and reliability cannot be improved through distributed deployment. To prevent the simulation process from being interrupted, each server must deliver high performance without breakdown.

How the simulation system performance increases by over 10 times?

To improve the performance of the existing simulation platform, GlobalFoundries seek the solution of high computing platform from Huawei. In this solution, Huawei new-generation 8P RH8100 V3 servers and high-performance ES300 V2 PCIe SSD cards are used to replace the original servers on the simulation platform. Under the new computing platform, it provides optimal computing capability and scalability. It makes it ideal for large-scale key applications and server consolidation and virtualization. The server also supports 192 DIMMs and provides up to 12 TB memory capacity and 1.2 TB high-performance of the PCIe SSD cards. It eliminates data exchange between processors and hard disks in the calculation process and loads simulation calculation data to the memory at one time and a little process data to SSD cards, reducing read and write operations on hard disks. The server improves the simulation system performance by over ten times.

Address the future capacity expansion

To ensure IT capacity meets current and future business requirements in a cost-effective manner, the RH8100 V3 supports 192 DIMMs, twenty-four 2.5-inch SAS HDDs, SATA HDDs, or SSDs, 16 PCIe expansion slots, and 40 GE and 100 GE networks. It offers the needed flexibility for future expansion. The RH8100 V3 also supports online maintenance and replacement of the DIMM, PCIe card, hard disk, fan module, PSU, and DVD-ROM drive, preventing service interruption during capacity expansion.

Redundant features ensuring stable system operating

Most fault-tolerant computer systems are designed to handle possible failures, including hardware-related faults, the RH8100 V3 provides 60 RAS features and fault tolerance of chips, such as automatic recovery for processors and links, ensuring stable system operating. It also provides reliability technologies, such as memory mirroring, memory backup, and hot-swapping of PCIe cards. The server is the only server that allows hot-swap of DIMMs, preventing system breakdown caused by memory hardware faults. It also supports hot-swapping of key components, such as PSUs, fan modules, and hard disks, without opening the chassis cover. This enables quick replacement of faulty components during normal system operation.

Aging Testing: 15% lower failure rate than the industry average

Aging test is used to help determine the long-term effects of expected levels of stress within a shorter time. Huawei performs full disk read and write tests in addition to common tests for the hard disk aging test. The standard of Huawei’s full disk read and write tests is far higher than that of random sector read and write tests in the industry. With high standards of the entire process, Huawei servers offer 15% lower failure rate than the industry average, ensuring reliable running of simulation services for Singapore GlobalFoundries.

Overall Benefits found by customers

Through strict POC tests, Singapore GlobalFoundries started to use Huawei RH8100 V3 servers and ES3000 V2 PCIe SSD cards for simulation services from 2014. Huawei servers build an efficient chip simulation platform for GlobalFoundries. The new chip simulation platform reduces the time for handling a typical calculation task from one day to one hour. The new server supports 192 DIMMs, twenty-four 2.5-inch SAS HDDs, SATA HDDs, or SSDs, 16 PCIe expansion slots, and 40 GE and 100 GE networks. It offers the needed flexibility for future expansion. Huawei servers offer 15% lower failure rate than the industry average, ensuring reliable running of simulation services.