3 ms·
Can someone tell me how many errors over the past 10 years or so would have been fixed by ECC memory. I've no idea at all how much better it would be and what
by andy_ppp 10y ago
Can someone tell me how many errors over the past 10 years or so would have been fixed by ECC memory. I've no idea at all how much better it would be and what the current error rate it prevents is? It sounds more nice to have than vital? Is there a performance cost to it at all?
- 8draco8 10y agohttps://blog.codinghorror.com/to-ecc-or-not-to-ecc/ https://blog.codinghorror.com/to-ecc-or-not-to-ecc/
- acoye 10y agoWould prevent Rowhammer attacks.
- detaro 10y agoNot completely, since those can produce multi-bit flips.
- justin66 10y agoWith SECDED ECC, in the event that two bits are flipped, the system can detect this and halt.
- anoother 10y agoDDR4 is innately much more resilient to rowhammer than previous standards.
- chithanh 10y agoMore resilient maybe, but not Rowhammer-proof: https://arstechnica.com/security/2016/03/once-thought-safe-ddr4-memory-shown-to-be-vulnerable-to-rowhammer/ https://arstechnica.com/security/2016/03/once-thought-safe-d...
- justin66 10y agoArstechnica writeup of a Google study: https://arstechnica.com/business/2009/10/dram-study-turns-assumptions-about-errors-upside-down/ https://arstechnica.com/business/2009/10/dram-study-turns-as...