Method and System for Determining Proportion of Bit-Flips that Convert to Computing Errors
Publication Date: 2010-Jul-19
The IP.com Prior Art Database
A method and system for determining proportion of bit-flips that convert to computing errors is disclosed.
Method and System for Determining Proportion of Bit -Flips that Convert to Computing Errors
Disclosed is a method and system for determining proportion of bit-flips that convert to computing errors.
Computing devices are more and more being constructed from components and circuits that are individually error prone. This is of great concern in computing systems, especially those that rely upon correct dataset and continuous operation. If all bit errors manifested into a system failure, system failure rates would be alarmingly high. Systems occurring in nature are resilient to errors, and computing systems are also naturally resilient to errors. That is, a single bit flip does not have unit probability of inducing a system failure, but is derated
The method and system disclosed herein performs multiple tests on the computing device to determine the proportion of bit-flips that convert to computing errors in the computing device. The multiple tests include an accelerated test as illustrated in the figure.
The accelerated test is performed on a computing device by irradiating the computing device with a particle beam to induce soft errors. The particle beam may be a neutron beam, a proton beam or an alpha particle beam. A particle source such as a hot underfill may also be used. Accordingly, particles impinge the computing device to generate a considerable number of errors in a reasonable amount of time. The errors