Metis Insight #3

“What is Chipkill?”

Chipkill, also known as Chipkill-correct, is a memory error correction technology designed to enhance the reliability of DRAM. It provides a higher level of protection compared to traditional ECC by correcting multi-bit errors within a single memory chip or across multiple memory chips.

In a DRAM chip, a wordline failure can occur. This type of failure affects an entire row of memory cells, potentially leading to a die failure where significant portions of data become corrupted.
A wordline failure often results in burst errors. Burst errors are sequences of corrupted data bits that occur in a rapid succession, typically affecting multiple bits within a word or across multiple memory cells. This makes it challenging for traditional ECC to correct, as they are designed primarily for single-bit errors.
To handle the complexities and severity of burst errors resulting from die failures, a more robust error correction mechanism is needed. This is where Chipkill technology comes into play, providing advanced error correction capabilities that can handle multiple bit errors and even entire chip failures.

This is how Chipkill Works ,
- Distributing Data and Parity Bits: Data is striped across multiple memory chips, with additional parity or ECC bits distributed in a way that allows for error correction across chips. If bits are corrupted in different chips, the system uses the parity or ECC information to reconstruct the correct data.
- Advanced Algorithms: Implementing Chipkill requires support from both the memory controller and the memory modules, with additional complexity to handle advanced error correction algorithms, such as BCH (Bose-Chaudhuri-Hocquenghem) Codes and RS(Reed-Solomon) Codes. These algorithms decode the ECC/parity information and correct errors across multiple chips.

MetisX’s computational memory solution supports the double-die Chipkill correction, which can correct data from failures of two memory dies.
This robust RAS feature offers reliable and efficient memory performance and ensures increased fault tolerance, enhanced data integrity.

Stay tuned for more technical insights through Metis Insights. We'll bring you more in-depth knowledge next time.