Can software damage memory?

[Shrek]

KB5019275 Windows 10 19042.2546, 19044.2546, and 19045.2546 - Jan. 19 Windows Update - Windows 10 Forums (tenforums.com)

"
This update addresses an issue that damages memory. The issue occurs when you use certain HD audio controller hardware.
"

 

[umeng2002]

Poorly worded notes, most likely. It sounds like it might over write some NAND portion of a controller or something like that. Or maybe they are talking about data corruption in the software stack, not really hardware damage.

 

[LabRat 891]

Poorly worded notes, most likely. It sounds like it might over write some NAND portion of a controller or something like that. Or maybe they are talking about data corruption in the software stack, not really hardware damage.

Gonna need a better/more-official clairification.
Unless things have changed, Msft's Update Teams are typically US-based, and the changelogs/KB articles are in clear-English (although often lacking in detail).
In other words,I wouldn't be so fast to excuse it as poor grammer. Yes, it does look (potentially) like a typical '-inglish' error. '-inglish' as in cheap Chinese gadgets' instructions, or Indian news articles that haven't been reviewed by an editor fluent in English.

I have a friend w/ a B450 board recently seem to 'kill' RAM pulled from another working machine. I assumed it's board-related, but this is certainly recent-enough to overlap with his problem(s).

 

[Solaris17]

Poorly worded notes, most likely. It sounds like it might over write some NAND portion of a controller or something like that. Or maybe they are talking about data corruption in the software stack, not really hardware damage.

Yeah I totally agree, this is likely an addressing issue that is damaging some memory heap. Its not like windows is changing your ram voltage lmao.

EDIT:

Yeah context is key. QoL improvement list full of other software bug fixes. These arent marked critical or hotfix.

 

[LabRat 891]

Yeah I totally agree, this is likely an addressing issue that is damaging some memory heap. Its not like windows is changing your ram voltage lmao.

EDIT:

Yeah context is key. QoL improvement list full of other software bug fixes. These arent marked critical or hotfix.

View attachment 279975

I still am hesitant to make that assumption.
AFAIK *all* modern attack vectors take advantage of 'otherwise legal' commands and actions at the HW level. For instance: (IIRC) a 'row hammer' attack is a spamming of commands, such that the physical characteristics of the RAM are exploited to read/write where otherwise there is no permission(s). Being able to cause physical damage with 'otherwise legal' commands, seems plausible. (oversimplified: I'm thinking like of a typewriter 'crashing')

 

[Solaris17]

I still am hesitant to make that assumption.
AFAIK *all* modern attack vectors take advantage of 'otherwise legal' commands and actions at the HW level. For instance: (IIRC) a 'row hammer' attack is a spamming of commands, such that the physical characteristics of the RAM are exploited to read/write where otherwise there is no permission(s). Being able to cause physical damage with 'otherwise legal' commands, seems plausible. (oversimplified: I'm thinking like of a typewriter 'crashing')

uh huh. So maybe we are talking about different "row hammer" attacks. I seem to remember a different one that has absolutely nothing to do with physically damaging cells.

Row hammer - Wikipedia

en.wikipedia.org

Anyway choose to believe what you will. I'm going to open netflix without my windows update tonight and watch star trek until my ram fails.

 

[Bill_Bright]

Being able to cause physical damage with 'otherwise legal' commands, seems plausible.

I don't agree at all. What seems plausible is the data stored in memory can be damaged (corrupted). But for software to physically damage hardware - nope.

What might also be considered plausible is malicious code corrupting the data stored in firmware - rendering the hardware inoperable. But that still is not "physical" damage.

 

[umeng2002]

If there is some voltage controller connected to a bus, and some weird audio driver behavior somehow addresses to that device causing it to blasts the RAM with too much voltage... That's all I can think of.

 

[Shrek]

Overlockers are able to change voltages with software, so I hope this cannot be done as an attack.

 

[Bill_Bright]

Even if voltages are somehow changed to some unsupported value, the typical result is instability and shutdowns - not physical (as in irreversible) damage to the memory sticks.

 

[lexluthermiester]

Can software damage memory?

By "memory" we can presume you mean RAM? Short and simple answer: No. RAM is not like a CPU, GPU or any other kind of processor. RAM will operate at it's rated stable design unless damaged by an external PHYSICAL force. No software can push it to overheat like a CPU or GPU. Now, Overclocking/overvolting can damage RAM, but that qualifies as an external physical force.

Overlockers are able to change voltages with software, so I hope this cannot be done as an attack.

It sure can, but requires direct on-site physical access.

Being able to cause physical damage with 'otherwise legal' commands, seems plausible.

That is in fact completely false. Let me say this once more for clarity:

Software can NOT damage RAM! Full stop, end of discussion!

 

[ThrashZone]

Hi,
Guess you aren't talking about a dodge ram


Think the board would likely get killed/ dimm slot before the memory would be killed but never say never when overclocking errors do happen with finger spasms
Wonder how many sticks and boards have been killed trying to break world records

 

[qubit]

Nah, can't physically damage memory, like everyone else is saying.

There are only very few ways that software could ever damage hardware. One way that comes to mind is the software that ran on 1980s computers to make the floppy drive play a tune by moving the head back and forth. This causes damage through premature wear and tear and perhaps heat to some extent.

Another way to possibly damage 40 track floppy drives was to try and drive the head past the end of their travel by sending a command for tracks past the end of travel. If the drive didn't reject the invalid command and the head repeatedly hit the backstop, then the drive might get damaged eventually.

 

[Shrek]

It sure can, but requires direct on-site physical access.

Good to know

 

[Assimilator]

More context is definitely needed. I wonder if Raymond Chen knows...