1. 18
  1. 3

    There needs to be a point where we have to realize that the only way to prevent this is using ECC-RAM everywhere it matters (and I have a hard time coming up with contexts where it doesn’t).

    Intel destroyed the possibility of using ECC-RAM in below-Xeon personal computers with their horrible artificial market segmentation. Their rowhammer-mitigations are just a dire attempt at keeping up with this toxic business practice.

    AMD does this better, given every non-APU Ryzen-chip supports ECC memory (unofficially), however, it often depends on the mainboard-manufacturers to enable this feature in the BIOS and many people don’t know about this. Many assume ECC requires powerful processors or something, which is patently false.

    If you want for things to improve, be able to rely on your hardware in this regard and make ECC more popular again, you can do three things:

    • Buy AMD and not Intel, and let Intel know about your reasons
    • Contact AMD and mainboard-manufacturers and express your interest in official ECC-support
    • Buy ECC-supporting mainboards and ECC-memory
    1. 6

      ECC has never protected against rowhammer.

      Hardware side channels have always been present. Cache timing vulnerabilities recently came to prominence, but had been theorized decades ago. I believe rowhammer had been as well. There is no secure way to run untrusted code on trusted hardware. None.

      That said ECC is great and should definitely see more adoption.

      1. 4

        Even that might not be enough, from the article:

        What if I have ECC-capable DIMMs?
        Previous work showed that due to the large number of bit flips in current DDR4 devices, ECC cannot provide complete protection against Rowhammer but makes exploitation harder.