Linus Torvalds blames Intel for failing to release ECC memories
Here are his words:
The use of ECC RAM is absolutely important, so their availability, and Intel has been instrumental in killing the entire industry with its horrible market segmentation. Try going out and looking for DIMMs ECC - they are really hard to find. Yes - probably thanks entirely to AMD - the situation may have improved slightly lately, but that confirms my observations. Intel has been detrimental to the entire industry and to users due to its wrong and misleading policies towards ECC memory. And if you don't believe me, look at the various generations of Rowhammer, where each time Intel and memory makers talked about how to fix the problem.
Torvalds' post continues in very colorful language, blaming Intel for the lack of widespread adoption of ECC in the consumer market. Torvalds says this is due to the Santa Clara company's complete blocking of ECC support on its consumer chipsets and processors, arguing that this alone killed any incentive for memory makers to create desktop ECC RAM.
Linus further states that Rowhammer's problems could easily be solved with ECC memories. DRAM memory cells can disperse their charges into other memory cells. Usually, it's just a flaw in the system RAM that could cause memory errors, but Rowhammer attacks use this system as a mechanism for gaining high-level privileges.
Torvalds stated that standard memories are a nightmare when developing code for an operating system's kernel. Linus talks about how incredibly difficult it is to find out where an unexplained kernel error has occurred, arguing that this could often be the result of a hardware, not a code, problem that could have been fixed with ECC.
Torvalds also praised AMD for unofficial support for ECC on consumer Ryzen platforms, giving users the ability to use ECC without paying ridiculous amounts of money for server-class hardware. Remember that the quality of the memories can be fundamental for the stability of the computer, since even the slightest number of errors can cause crashes or data loss. Unfortunately, non-ECC RAM is always at risk of errors and is never 100% stable.
