Bad E5-2670?

[briandm81]

So I finally got my newest E5-2670's up and running...or not. When I fire up the ESXi installer it fails with the purple screen of death. Thoughts?




Oh no...should I send them back? Could it be something else? It seems pretty cut and dry to me.

 

[pricklypunter]

I don't think that's the latest ESXi build, is it? (Update 1a - 3073146) Maybe try downloading the latest one, could be a driver issue of some sort. Make sure your bios etc is up to date as well. I don't see any mention of TLB invalidate, PCPDU CPU lockup or anything that would immediately scream faulty CPU to me. I would treat this as I would any PC, make sure CPU/ RAM etc is well seated, heatsinks are clean and making good thermal contact, power supply good, RAM is good, then having ruled out the usual suspects, turn my attention to media and install software etc.

Despite what it says on screen, resist all temptation to PANIC!!!!

 

[briandm81]

It's probably not. I have this same board/cpu combo running that build, so I figured for now I would keep them consistent. I pulled both heatsinks and everything looks good from a contact perspective. Power supply and motherboard were brand new from Newegg (not that they couldn't still be the problem). The RAM is from Ebay, it appears to be seated well, but I can't really test it any other way. I guess I could load Windows on the box and see if it succeeds. All of the errors appear to be CPU related unless I'm just interpreting it wrong.

 

[pricklypunter]

It kind of looks like that when anything causes it to crash hard because it will core dump the processor registers as well as pertinent RAM contents etc to aid troubleshooting. Update to the latest bios, strip out everything to the bare min, download a copy of the latest ESXi and try that. Or as you say, stick windows on it, I find Win 7 is pretty quick at weeding out dodgy hardware, or if you have a eval version of server 2008/12, that should see all cores during install

 

[Patriot]

Are you trying to do vtd passthrough on a c0 or c1 chip?

 

[briandm81]

I have SK0KX stepping chips with VTd support. I'm not actually trying to pass anything through however. This system is intended to just host a bunch of VM's. I don't ever even intend to use VTd at this point, though it will be nice to have if I do want it.

 

[briandm81]

I started memtest before i headed out just to see if maybe memory is the issue. so far so good...long way to go.

 

[marv]

did you try with single cpu ?

 

[briandm81]

not yet. I'll definitely be trying that after memtest finishes this pass. I may try prime95 in windows too. hell, maybe my usb stick is bad...

 

[briandm81]

So far....

ESXi Install Fails
Memtest86 ran for 23 hours with no Errors
Windows Server 2012 R2 Fails
Windows 7 install starts but on the first reboot it won't come up.

Then I pulled a CPU and half the memory:
Windows 7 still won't boot, but hey maybe the install is jacked up from failing with both CPUs.
ESXi Install fails again...


Pulled the CPU and swapped it for the other CPU...again purple screen on ESXi install.



Arg! I guess I'll borrow some RAM from another system later tonight and see if it could be some odd compatibility issue. If that doesn't work, I'll borrow a CPU from another system and see if THAT works. If all else fails...it must be the motherboard.

If I'm missing anything obvious or stupid...let me know!

 

[Patriot]

What are you installing to?

 

[Stereodude]

Have you tried booting from an Ubuntu (or your favorite flavor of Linux) LiveUSB and seeing what happens?

 

[briandm81]

For ESXi, I have a 16GB thumb drive that I'm installing to and from.

For Windows, I'm using IPMI to mount an ISO image and installing to a SanDisk Ultra II connected to one of the two C602 SATAIII port.

I haven't tried booting directly to Linux yet (I'm a barely functional in linux).

Could this be a memory compatibility issue, or would that just leave me not able to boot? Memtest seems to think the memory is fine...

I can't imagine that both processors are bad...which would leave me with a bad motherboard? I'll report back later tonight once I swap memory.

 

[Stereodude]

I haven't tried booting directly to Linux yet (I'm a barely functional in linux).

It's super easy. Download the 64-bit 15.10 ISO for Ubuntu from here and then follow these instructions.

If Ubuntu boots and gets to the "desktop" the memory, motherboard, and processors are most likely fine.

 

[briandm81]

Giving that a try before I start swapping more hardware around. I've downloaded the ISO and mounted it with IPMI. So we'll see what happens.

 

[Patriot]

You ISO's could be bad, verify the checksums.
Your network could be dropping packets and messing up the ipmi mounts...

grasping at straws

 

[briandm81]

So Ubuntu seems to be working via the "Try Ubuntu" option when I mounted the ISO. It's very slow...but then again I'm running it over the network...

So now what? I guess I can try making it a USB stick and seeing how it performs, but honestly, if I can't install Windows 2012, Windows 7, or ESXi...something is bad wrong! I guess I'll start by taking all the RAM out and swapping and borrowing a few sticks from my other server. It has the same motherboard and processors, so I know it should be a safe bet.

 

[briandm81]

So I just swapped some RAM from my other box...installation has already gotten much further. Just have one CPU right now and four modules. Might be either selling the RAM or attempting a return on Ebay (not sure how that might go). But, we'll wait and see if this finishes or not...

 

[Stereodude]

So Ubuntu seems to be working via the "Try Ubuntu" option when I mounted the ISO. It's very slow...but then again I'm running it over the network...

So now what? I guess I can try making it a USB stick and seeing how it performs, but honestly, if I can't install Windows 2012, Windows 7, or ESXi...something is bad wrong!

Bad SATA/SAS cable? Flaky SATA/SAS port on the motherboard? Bad drive?

Your system passes memtest86+ (not foolproof) and runs another OS fine that bypasses your SATA/SAS subsystem, I wouldn't be focusing on the CPU or RAM at this point.

 

[briandm81]

I would have thought so too...but ESXi just finished installing with different RAM. Not that this is for sure the problem, but its sure seems to be memory related.