Loading...
Please wait, while we are loading the content...
Similar Documents
XC 40 TM System Diagnosability : Functionality , Performance , and Lessons Learned
| Content Provider | Semantic Scholar |
|---|---|
| Author | Schutkoske, Jeffrey J. |
| Copyright Year | 2017 |
| Abstract | The Intel Xeon Phi CPU 7250 processor presents new opportunities for diagnosing the nodes based on this processor in the Cray XC system. This processor supports a new high-bandwidth on-package MCDRAM memory and related interfaces. It also provides the ability to support different NonUniform Memory Access (NUMA) configurations. The new Cray Processor Daughter Card (PDC) also supports an optional PCIe SSD card. This processor requires new BIOS, administrative commands, power and thermal limits, as well as new diagnostics to validate functionality and performance. This paper describes the diagnostic tool chain changes required to support this processor and PCIe SSD card. It describes the functionality, performance, and lessons learned from diagnosing problems at scale. It also provides detailed examples on how to diagnose node faults within the Cray XC40 system. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://cug.org/proceedings/cug2017_proceedings/includes/files/pap140s2-file1.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |