Dead CPU? Help Please

[crazymonkeyzero]

Essentially I think I have a defective Intel cpu (what are the odds?) and want to confirm whether or not this is true based on forum input before I contact Intel.

The cpu is an Intel Xeon E3 1275 V2 which is in a workstation to run a computational chemistry program Gaussian09.The other day when I was running Gaussian, the computer screen suddenly went black and then the compute prompt said that the computer has recovered from a serious error, boot normally? I was like w/e and ran the same job again, and the exact same crash occurred around the same intensive step of the calculation. I was suspicious of system stability so I ran Prime 95. Here's the kicker, it crashed within seconds of running small ffts (max ftu l2 cache tested.) However when I run large ffts, everything works fine for 1+hr. So is this a cpu problem? The psu is working fine, I already tested it. Ram is also functioning correctly. I actually just replaced the mobo via Asus rma, as it would not boot last month giving a "diag cpu led". I dismissed it as a mobo problem and the new board did boot but now I'm not sure anymore. The computer functions normally until I apply stress to the system, which is probably why it crashed in Gaussian.

Could some people please give some input on this situation.

Thanks for the help!


Edit: forgot to mention. This is not a temperature related problem. Using a corsair h60, temps typically range from 28-59C always.

 

[ShadowVVL]

list full system specs?

could be ram, cpu or maybe psu.

I know you checked the ram but if you have a spare set maybe try them.

if same issue then yeah it might be cpu.It's not impossible to get a bad chip just unlikely.

 

[ViRGE]

list full system specs?

could be ram, cpu or maybe psu.

I know you checked the ram but if you have a spare set maybe try them.

if same issue then yeah it might be cpu.It's not impossible to get a bad chip just unlikely.

Agreed. And a bad Xeon in particular would be extremely unlikely given the extra testing Intel does, but it's still not impossible.

You're going to have to go through the usual troubleshooting steps to isolate the CPU. With that error it's probably the CPU, but all the same you need to eliminate the other components to be sure.

 

[crazymonkeyzero]

Here are the specs

Intel Xeon E3 1275 V2
Asus P8C WS 1155
Seasonic X 620w Gold PSU
Kingston ddr3 1600 Unbuffered ECC memory.
using integrated intel p4000 graphics on ivyb.
1tb seagate hdd 7200rpm
corsair h60 cooler

I want to re emphasize that I have already run most these tests. The trouble started in February when my computer (which I have had for the past 8 months) suddenly shut down in the middle of a job and would not. Here is a thread I created then to put everything in perspective.


http://forums.anandtech.com/showthread.php?t=2303593&highlight=



Essential between now and then I have RMA'ed the board and the new board boot, but as the aforementioned problems during the stress tests...Ram was tested on an individual basis...I can try this again. Psu should not be a problem because prime95 runs fine on large ffts, which actually draws the most power to the cpu. However, small ffts fail, (by fail I mean when I run it, one second later my monitor goes back and computer crashes, asking to restart. From what I read, small ffts test cpu.

 

[crashtech]

CPU failures are rare. Have you been able to sub a known good CPU into the existing system?

 

[ShadowVVL]

Run memtest86 for 2-4 hours then run intel burntest for 2hours.

I would test the psu with a multi just to be safe.

I guess the easiest way is you just rma the cpu and if the new one does it then you know its not the cpu.

 

[Idontcare]

Here are the specs

Intel Xeon E3 1275 V2
Asus P8C WS 1155
Seasonic X 620w Gold PSU
Kingston ddr3 1600 Unbuffered ECC memory.
using integrated intel p4000 graphics on ivyb.
1tb seagate hdd 7200rpm
corsair h60 cooler

I want to re emphasize that I have already run most these tests. The trouble started in February when my computer (which I have had for the past 8 months) suddenly shut down in the middle of a job and would not. Here is a thread I created then to put everything in perspective.


http://forums.anandtech.com/showthread.php?t=2303593&highlight=



Essential between now and then I have RMA'ed the board and the new board boot, but as the aforementioned problems during the stress tests...Ram was tested on an individual basis...I can try this again. Psu should not be a problem because prime95 runs fine on large ffts, which actually draws the most power to the cpu. However, small ffts fail, (by fail I mean when I run it, one second later my monitor goes back and computer crashes, asking to restart. From what I read, small ffts test cpu.

Try running HCI Memtest (link is in the OC'ing sticky). I thouht I had a bad 2600k CPU because LinX would fail. Memtest86+ said the ram was fine.

I troubleshot that for a while, decided to give HCI Memtest a try and discovered that my memory was bad.

Apparently Memtest86+ is not good enough to catch some critical errors that can occur on modern systems.

So give HCI memtest a try before you rule out this being a memory issue.

That said, I've RMA'ed an Intel CPU before and it was silly easy to do. So after you've ruled out the ram (if it turns out to be fine per HCI Memtest) you should RMA the CPU just to have it taken care of.

Now if you want my gut feeling here though, I bet the CPU and ram are fine. I bet what you are going to find is that the Vdroop is too large for the motherboard when it starts pulling the amps that Gaussian and small FFT prime95 require.

If you run CPUz while running Gaussian or Prime95, how much of a voltage drop do you observe? Can you tweak LLC or Vcc in the BIOS to boost the Vcore and counter-act the Vdroop?

 

[IntelEnthusiast]

If you need to give us a call you can call our technical support at 916-377-7000; they will be happy to help you out.

 

[Zap]

Have you been able to sub a known good CPU into the existing system?

:thumbsup: Best way to test for an iffy part is to swap it out.

Problem goes away? You found your culprit.

Problem stays? Might be another part causing it.

 

[crazymonkeyzero]

If you run CPUz while running Gaussian or Prime95, how much of a voltage drop do you observe? Can you tweak LLC or Vcc in the BIOS to boost the Vcore and counter-act the Vdroop?

This has now become impossible to measure while running gaussian or prime. The second I start the job, my screen goes black and computer restarts... Now large ffts also crash thee system.

I am currently running HCI memory test, will report back...so far its like 15%....and no errors. Also ran LinX....passed 20 cycles.

But now prime 95 small AND large ffts crash my computer...

But I think it might be the vdroop thing you were talking about. Would vdroop be the fault of the motherboard, or the psu?


I think I might buy an i3 or celeron to test out?

 

[Idontcare]

But I think it might be the vdroop thing you were talking about. Would vdroop be the fault of the motherboard, or the psu?

Vdroop would be the mobo, but it doesn't necessarily mean the mobo is bad. Could just mean that it is configured poorly at the moment and the LLC needs to be tweaked or optimized.

Could also be something systemic in the BIOS, like a poorly implemented power-state config. For example it could be inadvertantly calling for the low-power state voltage when transitioning between fully loaded and partially loaded when the software is loading up the next matrix test.

You might try manually disabling the power saving settings (C1E, C3, C6, EIST, etc) in the BIOS and see if that makes a difference. If it doesn't then you can always re-enable them again.

 

[nightspydk]

Just a bit of food for thought. I had a system crash at post because of a keyboard. Yes really, so the first thing you du is disconnect everything. Some of the gurus will have you boot the board outside the case. I say just everything of even your mouse and see if it posts.

Step 1.

 

[crashtech]

I think I might buy an i3 or celeron to test out?

The only problem with that is it should be a CPU that can draw a similar amount of current, because installing a CPU with lower current draw could mask a possibe problem with another component that is causing the higher powered CPU to crash. Personally I suspect the PSU, but it's just a hunch. How about an extra PSU, do you have one?

 

[crazymonkeyzero]

The only problem with that is it should be a CPU that can draw a similar amount of current, because installing a CPU with lower current draw could mask a possibe problem with another component that is causing the higher powered CPU to crash. Personally I suspect the PSU, but it's just a hunch. How about an extra PSU, do you have one?

I don't think It's the psu. Tried a sparem working 450w antec, same problems.

 

[crazymonkeyzero]

memtest passed with no errors. So next I'll mess with the bios cstates to see if it fixes anything...

 

[zaydq]

My bet is the mother board Asus gave you is also faulty. I wouldn't put it past their Quality Control.

 

[crazymonkeyzero]

Turns out, it wasn't the motherboard or the cpu, but rather a software/corruption issue which was causing the computer to crash under load. On a whim, I just ran a system restore yesterday to January, before my computer was giving me problems. Now, my computer doesn't crash anymore. However, I was wondering, if anyone can point out what program could be corrupted enough to cause my computer to crash? Windows??? DO you guys think if I let the computer run, it will eventually revert back to that unusable state? I was thinking of doing a fresh install of windows and a new bios update, would you guys recommend that?

Thanks

 

[crashtech]

Fresh Windows is never a bad idea, if you have the time. BIOS update, not so much, unless there is a specific issue that you're trying to fix, or a wanted feature has been added. Most of the time BIOS updates merely allow the recognition of newer CPUs along with a bugfix or two.

Thanks for giving us the update, it would be good for all of us to remember that data corruption can pretty effectively simulate a CPU defect. Don't think I've ever run into that before.

 

[Redoitall]

Try running HCI Memtest (link is in the OC'ing sticky). I thouht I had a bad 2600k CPU because LinX would fail. Memtest86+ said the ram was fine.

I troubleshot that for a while, decided to give HCI Memtest a try and discovered that my memory was bad.

Apparently Memtest86+ is not good enough to catch some critical errors that can occur on modern systems.

So give HCI memtest a try before you rule out this being a memory issue.

That said, I've RMA'ed an Intel CPU before and it was silly easy to do. So after you've ruled out the ram (if it turns out to be fine per HCI Memtest) you should RMA the CPU just to have it taken care of.

Now if you want my gut feeling here though, I bet the CPU and ram are fine. I bet what you are going to find is that the Vdroop is too large for the motherboard when it starts pulling the amps that Gaussian and small FFT prime95 require.

If you run CPUz while running Gaussian or Prime95, how much of a voltage drop do you observe? Can you tweak LLC or Vcc in the BIOS to boost the Vcore and counter-act the Vdroop?

I had similar problem with 3770k I thought it was the ram, the CPU , the Mb but the problem was the Vdroop being too large . After days of frustrated trials , I adjusted the Llc and wham problem solved . And think I could not even boot while having the issue .

 

[crazymonkeyzero]

I had similar problem with 3770k I thought it was the ram, the CPU , the Mb but the problem was the Vdroop being too large . After days of frustrated trials , I adjusted the Llc and wham problem solved . And think I could not even boot while having the issue .

Spoke too soon. Computer crashed 2 hours into the job. Darn! Can vdroop all of a sudden be a problem if your computer was working fine for the past 6 months? I am currently reinstalling windows and seeing if that will help the issue. If not, then I will check/ try to adjust vdroop.

 

[crazymonkeyzero]

By the way,, what are safe LLC or Vcc adjustments. Like a safe range where I can guarantee won't fry the chip.