Multiprocessor BSOD ? That is a weird one

jibbajabba

One Server with 1x5410 Xeon (Quad Core) ..

Installed Windows Server 2008 Web x64 just fine .. Shutdown - put in a 2nd CPU - boot up :

Bluescreen :



Yes, both CPUs are identical .. I am wondering .. As the CPU is a quadcore - it should have installed the multi processor kernel anyway .. so why the heck throws it a blue screen when inserting a second CPU

I will now try to install it with both CPUs - which will probably work .. but this is just weird weird weird ...

astorrs

Weird, I would pull one CPU, boot it up and check which HAL it's using and force it to use the SMP one.

jibbajabba

Now tried installing the OS using both CPUs and same problem ..

Will have to check with CPUz whether both CPUs maybe do have a difference I don't see .. for crying out loud

jibbajabba

Aww man, this is painful ..

We have a bucket full of 5410 CPUs .. now check this out :



See the stepping

Same batch .. different stepping .. no wonder it goes boom .. Now the fun part : finding two identical CPUs

astorrs

Stupid, otherwise identical. Sigh. Good luck.

dynamik

Wow.. that's crazy

tiersten

It isn't complaining about the stepping. MXCSR on the SSE unit on the two CPUs differ.

jibbajabba

Interesting enough, both CPUs from the above server have a different block of serials ..

On Xeon CPUs you normally have two lines .. first line is always identical (I guess some sort of batch number) but in this case it wasn't ...

Now the problem with stuff like this : If you every want to upgrade a server to a dual processor in the future (it will use the correct HAL anyway since Xeons are multicores) you have to hope its the same batch .. if not .. hope you have another matched pair from one identical batch and swap both ...

Considering we have hundreds of server this could be fun

I haven't seen this on Server 2003 yet - or maybe we were just lucky ... more testing needed I suppose ...

jibbajabba

tiersten wrote:

It isn't complaining about the stepping. MXCSR on the SSE unit on the two CPUs differ.

I see I forgot to add the "interesting" column

http://i33.tinypic.com/rtg3zt.jpg

So Core Stepping is indeed different (not SSE ?)

tiersten

Gomjaba wrote:

tiersten wrote:

It isn't complaining about the stepping. MXCSR on the SSE unit on the two CPUs differ.

I see I forgot to add the "interesting" column

http://i33.tinypic.com/rtg3zt.jpg

So Core Stepping is indeed different (not SSE ?)

I know that the stepping is different. I'm just saying that MXCSR is what the bugcheck is complaining about. You can run Windows SMP with different steppings of a CPU. Decoding bugcheck 0x0000003E.

Intel Xeon 5400 Datasheet wrote:

Mixing processors of different steppings but the same model (as per CPUID instruction) is supported.

Datasheet link

[edit]Added datasheet info[/edit]

tiersten

Decoding the values given don't make sense if it is a MXCSR register however. Check with whoever manufactures your motherboard to see if there are any firmware updates and also ask Microsoft support.

We just order the correct parts from IBM or HP. If it doesn't work then we just shout at them until they fix it.