Someone finally published an application that injects ECC errors!

[Ericloewe]

Well, it seems that finally someone decided to take advantage of the ECC error injection capabilities of modern processors:

http://www.memtest86.com/features.htm

Of course, there's a catch. Several, actually:

• They want you to fork over 40 bucks per simultaneous instance you plan on running.
• Those 40 bucks only get you a year's worth of updates (almost guaranteeing that you'll need to pay again when the next platform arrives)
• It's unbelievably bloated. That thing has a GUI. A fscking GUI. In a program that's meant to test memory. At best, that implies crazy shuffling of the whole program around physical RAM. At worst, it implies a larger portion than necessary of RAM that is not tested.
• They're one of those companies that pick up open-source stuff and proceed to close-source it.
• If you read the release notes, they say that the big feature, ECC error injection, is "untested" on Xeon E3 v1, v2, v3 and v4. And they don't explicitly mention support for this on Xeon E3 v5.

 

[HoneyBadger]

"Your system 'may support' testing."

Boy, there's a vote of confidence.

Also would have liked to see injection support for Pentium/i3 processors so we can settle that debacle once and for all regarding what does and doesn't work.

 

[Ericloewe]

"Your system 'may support' testing."

Boy, there's a vote of confidence.

Also would have liked to see injection support for Pentium/i3 processors so we can settle that debacle once and for all regarding what does and doesn't work.

I'm assuming that they work identically to the Xeons, meaning that their injection operation is identical.

Come to think of it, I never did check the Core ix datasheet.

 

[Bidule0hm]

Well, it seems that finally someone decided to take advantage of the ECC error injection capabilities of modern processors:

http://www.memtest86.com/features.htm

Of course, there's a catch. Several, actually:

• They want you to fork over 40 bucks per simultaneous instance you plan on running.
• Those 40 bucks only get you a year's worth of updates (almost guaranteeing that you'll need to pay again when the next platform arrives)
• It's unbelievably bloated. That thing has a GUI. A fscking GUI. In a program that's meant to test memory. At best, that implies crazy shuffling of the whole program around physical RAM. At worst, it implies a larger portion than necessary of RAM that is not tested.
• They're one of those companies that pick up open-source stuff and proceed to close-source it.
• If you read the release notes, they say that the big feature, ECC error injection, is "untested" on Xeon E3 v1, v2, v3 and v4. And they don't explicitly mention support for this on Xeon E3 v5.

In others words: it's a useless paying piece of crap... :D

 

[Starpulkka]

edit: on second thought its cristmas and all,,
you forgot to mention injection needs uefi.

Merry christmas- Hyvää Joulua ja onnellista uutta vuotta 2016.
edit:2 theres so much intel crap so its not recommend to buy that software only for ecc testing for intel.
Jetty Thingsings?

 

[Bidule0hm]

Code:

       __,_,_,___)          _______
      (--| | |             (--/    ),_)       ,_)
         | | |  _ ,_,_        |     |_ ,_' , _|_,_,_, _  ,
       __| | | (/_| | (_|     |     | || |/_)_| | | |(_|/_)___,
      (      |___,   ,__|     \____)  |__,          |__,
 
                              |                         _...._
                           \  _  /                    .::o:::::.
                            (\o/)                    .:::'''':o:.
                        ---  / \  ---                :o:_    _:::
                             >*<                     `:}_>()<_{:'
                            >0<@<                 @    `'//\\'`    @
                           >>>@<<*              @ #     //  \\     # @
                          >@>*<0<<<           __#_#____/'____'\____#_#__
                         >*>>@<<<@<<         [__________________________]
                        >@>>0<<<*<<@<         |=_- .-/\ /\ /\ /\--. =_-|
                       >*>>0<<@<<<@<<<        |-_= | \ \\ \\ \\ \ |-_=-|
                      >@>>*<<@<>*<<0<*<       |_=-=| / // // // / |_=-_|
        \*/          >0>>*<<@<>0><<*<@<<      |=_- |`-'`-'`-'`-'  |=_=-|
    ___\\U//___     >*>>@><0<<*>>@><*<0<<     | =_-| o          o |_==_|
    |\\ | | \\|    >@>>0<*<<0>>@<<0<<<*<@<    |=_- | !     (    ! |=-_=|
    | \\| | _(UU)_ >((*))_>0><*<0><@<<<0<*<  _|-,-=| !    ).    ! |-_-=|_
    |\ \| || / //||.*.*.*.|>>@<<*<<@>><0<<@</=-((=_| ! __(:')__ ! |=_==_-\
    |\\_|_|&&_// ||*.*.*.*|_\\db//__     (\_/)-=))-|/^\=^=^^=^=/^\| _=-_-_\
    """"|'.'.'.|~~|.*.*.*|     ____|_   =('.')=//   ,------------.
    jgs |'.'.'.|   ^^^^^^|____|>>>>>>|  ( ~~~ )/   (((((((())))))))
        ~~~~~~~~         '""""`------'  `w---w`     `------------'

 

[joeschmuck]

I like the idea of the new MemTest86 except for the price, but there is an option for a site license so if we purchase one site license for the forums, we could all chip in $5.00 and have a copy of it. Oh, damn, site license doesn't mean that? Where is that written:D

 

[cyberjock]

Okay, so if its in a paid product, there's a possibility it'll be in a free product. Maybe we'll get lucky and someone in the forums will find a way to implement something that works via something like a simple python script!

Anyway, the fact that it says its "untested" for E3 v1, v2, v3 and v4 makes me think that they're saying "give us your money, and this may work.. but no guarantees of course!" and they have reason to believe it won't work, but are still happy to take your money thinking you may be stupid enough to think it will work.

 

[Ericloewe]

nyway, the fact that it says its "untested" for E3 v1, v2, v3 and v4 makes me think that they're saying "give us your money, and this may work.. but no guarantees of course!" and they have reason to believe it won't work, but are still happy to take your money thinking you may be stupid enough to think it will work.

That's more or less my line of thought.

If memtest86+ had a modern version that worked with UEFI and in long (64-bit) mode, it'd be a prime candidate to have this feature tacked-on.

I still think this is something anyone with a bit of experience writing UEFI executables could accomplish quickly, since Intel's datasheets are readily available.

 

[Bidule0hm]

Unless you can access CPU registers in Python I don't think you'll be able to use it. ASM or C are pretty much mandatory for this kind of things. I'm very good with C but I just don't have the time to read Intel datasheets for the most used CPUs in our servers. However if someone can tell me which value to put in which register for the CPUs you want to be able to test I'm ok to dev the program ;)