NEWS: AMD CPU Hardware bug can lead to stack corruption -not related to our soft

Started by Jeremy Collake, March 07, 2012, 08:27:23 AM

Previous topic - Next topic

Jeremy Collake

Software Engineer. Bitsum LLC.

Hotrod


edkiefer

the way I read it (I don't understand all terminology either) is this guy found a bug in Hardware under a certain condition the out data gets corrupted (at memory address ? ) .
He then mad a more simple test to show AMD what it takes to reproduce it fast .

AFAIK , CPU and its micro code sometimes does have erratic errors but many times there work around to them or there in trouble  :)

What was interesting is this happened in a simple single threaded run , I am not sure what app he was running or if he was compiling code when this happened originally so not sure how rare this is, probably pretty rare .
Bitsum QA Engineer

Jeremy Collake

Yes, it is an extremely rare condition. None-the-less, in these conditions, corruption of the stack pointer can occur, causing (most often) a crash of the immediate process. If it occurred in a kernel mode component, it would be a BSOD. Fortunately, the frequency of these occurrences is surely so rare that none of us have ever seen it.
Software Engineer. Bitsum LLC.

Jeremy Collake

The funny thing is that tech media on the web is too busy writing for their advertisers to report on real news these days, lol.
Software Engineer. Bitsum LLC.

Hotrod

Quote[â€"]DasWood 7 points 1 day agoThe tested and affected CPUs are
AMD Opteron(tm) Processor 6168 (1900.01-MHz K8-class CPU)
AMD Phenom(tm) II X4 820 Processor (2799.92-MHz K8-class CPU)
 


I have one of the Sempron 3100+ , hope it doesn't go back that far.  :(

Jeremy Collake

Dunno. It goes back a long way, but nobody knows how far yet I think. It would be interesting to get the code sample together. For accuracy, it may require a USB key and micro-OS to replicate reliably though, as so many variables are involved.
Software Engineer. Bitsum LLC.

edkiefer

http://en.wikipedia.org/wiki/List_of_AMD_Sempron_microprocessors

Sempron 3100+ was released 2005 and is 754 socket which is like 3 -4 sockets back from K8 series so it could be in it but that is a lot different cpu .

http://products.amd.com/en-us/DesktopCPUResult.aspx

As Jeremy said it would need testing to make sure .
Bitsum QA Engineer