Why Reliability is Important
Today all servers include a range of features designed to make them reliable
ClearSpeed’s HPC accelerators include the same features and are as reliable as the host servers
ClearSpeed's processors are designed to provide accurate and
reliable acceleration:
- ECC on all internal and external memory
- Detects, reports and/or fixes soft memory errors
- Low clock speed
- Low power
- Passive cooling
- Temperature and power monitoring
- Dynamic clock control
GPGPUs don't have these reliability features and consequently they will:
- Silently generate incorrect answers
- Crash more often
- Reduce the reliability and stability of any system
Read what Nvidia say about soft errors...