What others say about soft errors
| "Soft errors have yet to become a major consideration in the design of graphics processing systems, probably because the primary market is the consumer desktop, where reliability requirements are lower." | |
| "The Folding@Home GPU client has now been in distribution for 6 months, running on approximately 500 GPUs. Over this sample set, Folding@Home has shown a failure rate of approximately 1%" | |
| "One example of a new market where GPUs are starting to make inroads is that of radiology, where GPUs are employed for medical image processing. This is a domain in which errors are often very costly, both in financial terms and liability, and potentially also in human life. Erroneous computation can lead to death." |
Source: A Hardware Redundancy and Recovery Mechanism for Reliable Scientific Computation on Graphics Processors, NVIDIA Research, 2007.
"More recently, in 2000, Sun Microsystems observed [soft errors] in their UltraSPARC-II based servers, where the error protection scheme implemented was insufficient to handle soft errors occurring in the SRAM chips in the systems." |
|
| "In 2005, Hewlett-Packard acknowledged that a large installed base of a 2048-CPU server system in Los Alamos National Laboratorylocated at about 7,000 feet above sea-levelhad been crashing frequently because of cosmic ray strikes to its parity-protected cache tag array." |
Source: Computer Glitches from Radiation: A Problem with Multiple Solutions, Microprocessor Report,5/19/08-01