Should I get a new PS?

[jpanhalt]

I generally go to bed quite early, unless I have company, and start my day at 0400 (local).  This morning, I woke a little after 0130 and as usual turned on my PC.  I love to code for a few hours early in the AM.  Total darkness.  Fans turn on, for a second or two, and then turned off.  Checked out PS with a much older PC, and it seemed to work.  So, I focused on the CPU, for which I had a spare.  That worked.  Five to 6 hours of terror later, I was up and running again.  I am Win7 with an i7-4770K CPU and Gigabyte MB.  That CPU was purchased used and worked fine for 2 years.  The original new one in 2015 gave me 7 years of service.

I am beginning to think there is a problem with the PS that kills my CPU.  PS is a Thermaltake 750W unit.  Advice?

John

[wraper]

PSU cannot just kill a CPU and motherboard remaining OK at the same time.

[james_s]

It's hard to say, but I would suspect the motherboard in this case before the PSU, unless the PSU is misbehaving in a way that is causing the regulators on the motherboard to act up. The core voltage of modern CPUs is very low at tens of amps, it probably wouldn't take too much of a spike to zap one.

[jpanhalt]

@james_ s
That was my thought exactly.  Is Thermaltake known for doing that? Is there a more reliable brand?

[james_s]

I'm so far out of the loop these days I don't even know, I thought Thermaltake was reasonable though. It's worth looking closely at the capacitors on the motherboard, and also you might try bench testing the power supply under load and looking at the outputs with a scope. Then again, power supplies are fairly cheap, so you could just replace it and see if any more CPUs fail, if the problem keeps happening then you'll have a spare PSU.

[golden_labels]

If there is no persistent symptoms,⁽¹⁾ there is no way to tell the cause from that single event. It could be literally anything, including a cosmic ray. Without effects remaining after the event or failures re-occuring, you will simply never know, what killed your CPU.

Just to start with the processor itself: there is an incomprehensible number of transistors in your CPU,⁽²⁾ accompanied by even larger traces count. A production defect could be present or a damage could appear at any moment of its life. It would develop over some time and fail under higher current during powering CPU on or seeing a spurious voltage spike. And many other singular, untraceable events could happen outside of the processor.

Make that experience productive. If six hours of terror was due to the fear of losing your data, this should be a lesson regarding having backups.

⁽¹⁾ For example for a PSU that would be wrong or suspiciously unstable voltages.
⁽²⁾ Over a billion, to be more specific, though to almost all human brains this is meaningless.

[jpanhalt]

It wasn't fear of losing my data.  I have raided drives, and separate backups.  It was fear of not being able to get a new(er) CPU online when one doesn't have online access.  (I do not have a smart phone.)  MicroCenter is a small Midwest chain of computer stores.  There is one in Cleveland.  I go there when I really need something, but it has no telephone service for consumers.  Of course, for an antiquated chip, everything is online.

[golden_labels]

Ouch. That would be indeed a bigger problem. And it’s indeed hard to protect oneself from that kind of trouble. :/

[rdl]

You really should have a backup computer. The one you have now would be a good choice. Go look for an off-lease business pc. You can find tiny style KP/Lenovo/Dell machines for under $200 that would be better than what you have now. Of the three, I prefer HP.

https://youtu.be/ZPX3A7YS-6Q

[jpanhalt]

Actually, I am actively considering a new laptop for stuff like taxes, banking, etc. It's just a matter of when and what to get. 

For stuff I enjoy doing, I like my old Win7  desktop with a full keyboard.  All of the design stuff I have is Win7 or before.  That is, I got them ridiculously cheap when I was eligible for an education discount as faculty, e.g., SolidWorks.  Every program is owned. 

Now my problem is why an Intel i7-4770K died after just 2 years.  Albeit, it was used to begin with, but the previous chip was only about 5 years old.

[golden_labels]

Now my problem is why an Intel i7-4770K died after just 2 years.

As I explained above, there is no answer to this question. Described exceptions apply, but you did not post any information suggesting this is the case.

“Just 2 years” suggests you find the time unexpected. But there is as good as any other time. It’s a random event. A CPU may fail at any time and there is no magic, which prevents it from dying just because it was “just 2 years old.” While the probability distribution is indeed not uniform,⁽¹⁾ the difference is not big enough to put that into “that should not happen” region. And, if anything, a well-kept CPU is more likely to die being 2 years old than 5 years old.⁽²⁾

Comparing merely two specimens also makes no sense. It’s meaningless from statistical perspective, but you do not need to know anything about statistics to notice the problem on a much more basic level. Observe, that there is only one option possible. The length of their life determines their labeling. The one, which died at earlier age, becomes the younger one. The one, which survives, becomes the older one. Inevitably, the older CPU always survives longer. The event delivers no information. None at all.

⁽¹⁾ It follows a bathtub curve.
⁽²⁾ If non-linear perception of “oldness” is taken into account, this becomes even more pronounced.

[jpanhalt]

 I agree, odd events can happen.  But sequentially?  Is 5 years the average MTBF for an Intel CPU?   I hope not.

[golden_labels]

Letter “M” in MTTF stands for “mean”. Mean Time To Failure.

Time to failure is a probabilistic distribution. From many numbers describing its shape (called moments) MTTF indicates its overall location. But the time to failure itself is not a single number: it remains a probabilistic distribution. For example a MTTF of 1 year does not mean, that the single, particular device will certainly fail after reaching one year, or that it will survive at least one year, or even that it is very likely it will fail around that time. It means, that after testing infinitely many devices and writing down their time to failure, the “center of mass” of these values will be at 1 year. But a single device may fail even after one second or after 100 years.

A distribution has a particular shape and a single location. Therefore there is no “average MTTF”. It’s like pouring an immovable, permanent pile of sand and saying “average location of this pile is 5 m from the tree”.

Finally, you have only two datapoints. This value can’t be reliably estimated from sample of that size. However, ignoring this, I do not see how MTTF of 5 years would sound bad for a CPU. As explained above, MTTF does not indicate your CPU must die after 5 years.

Note: in this post I assumed you meant “MTTF”, not “MTBF”. “MTBF” stands for “Mean Time Between Failures”. A CPU dying is not a repetitive failure, therefore there is no MTBF for this type of a failure.

[nctnico]

Actually, I am actively considering a new laptop for stuff like taxes, banking, etc. It's just a matter of when and what to get. 

For stuff I enjoy doing, I like my old Win7  desktop with a full keyboard.  All of the design stuff I have is Win7 or before.  That is, I got them ridiculously cheap when I was eligible for an education discount as faculty, e.g., SolidWorks.  Every program is owned. 

Now my problem is why an Intel i7-4770K died after just 2 years.  Albeit, it was used to begin with, but the previous chip was only about 5 years old.

I suspect your motherboard is broken. Intel CPUs have thermal protection so it basically can't kill itself. With 2 CPUs going bad within a small amount of time, the motherboard is the only common denominator. Normally a CPU should be able to work problem free for 10+ years.

[jpanhalt]

NEW FAILURE

My CPU failed* again today (about 1 month after install).   In the meantime, I bought a Gigabyte "re-certified" MoBo. I would be more comfortable changing the MoBo if I understood better how it can kill a CPU after running perfectly well of 1 month.

At his point. I am a bit panicked and am working on a Gateway laptop circa 2003, Windows XP.

Desperate,

John
 
*Same symptoms as before.   Tries to start, fans run for about 1 second, then it goes dark.

[MK14]

If the 'real' problem is a desire for windows 7 use.  There ARE ways of installing Windows 7 on modern computers (CPUs).

E.g. The following video (which starts off explaining about USB compatibility issues, and various ways of getting around it).



You can also play around with Linux, some of which are partly (or more) similar to windows (e.g. Linux Mint) and with things such as (Wine, 'which is a sort of compatibility emulator for windows on Linux'), may be able to run the software you want to, which worked on Windows 7.

You can even try them, without actually installing them (via CD/DVD or USB flash pens, via LIVE versions).

[Ed.Kloonk]

I love to code for a few hours early in the AM.

Who doesn't?
Never trust anyone who churns out code during regular waking hours.

Edit: There is a far side comic that I cannot find where it's dark save for a light on in one window of a building and the computer programmer in there yells "Oh, wow!". First one to find it for me wins a prize.

[golden_labels]

Three CPUs dying in the same machine in a similar fashion is harder to take as a coincidence.

If that’s indeed caused by a voltage spike, it still will be hard to debug: events are rare, reproduction is destructive and majority of cuases are on a multi-layer, tightly packed PCB. :/

If you are interested in fixing it, not debugging, I would just say: replace both the motherboard and the PSU. With my suspicion being rather towards the motherboard than power supply.

Wait for opinions from others, but my reasoning here is as follows. A motherboard contains its own voltage regulators. These are primarily SMPS. Pretty good at adjusting to changes in input the voltage and there is the inductor and output filters in the path of the current towards CPU. One does not simply pass a small voltage spike through that. Not terribly hard for a computer regulators, as they are meant to pass high currents and charge quickly, but in my mind still not on the likely side. At the same time only CPU is affected and always in the samme manner. Certainly the most fragile part, but a PSU causing massive enough voltage spikes and not upsetting anything else?

[AndyBeez]

Do you monitor your CPU temperature? So your fan is working without issues; the heatsink paste is making a thermal circuit with the heatsink, the vanes of the heatsink are dust free and, if you overclock, the fan runs at full speed? I might suspect overheating if no other parts are breaking. Do you ever smell the smell of hot metal emanating from the case? The CPU fan could fail without you hearing. Is there a temperature alarm on your OS/Bios?

[jpanhalt]

4:02 PM EDT Update

Early this morning, I installed a "refurbished" MoBo and new CPU.  Boot froze at "15" on the digital monitor with beeping  (debug LED code 15 = "pre-memory North-Bridge initialization started").  Wiggled and reset the DDR3 sticks without effect.  Took a break until about 3PM cogitating about it and doing errands.  Removed one of the DDR3 sticks and thing booted correctly (what luck -- see next).  Took care of some bios details, then moved the good DDR3 stick to the other slot.  It froze again.  Tried the original DDR3 stick in the working slot (no stick in the assumed bad slot) and still working normally.  There are 4 DDR3 slots, two grey and 2 black.  Next step will be to install both DDR3 sticks in the two black slots.  If that fails, I can live with just one DDR3 stack for the time being.

As for Andy's questions:
Yes, I run a CPU temperature utility and always wait for it to start cooling down (usually to 38 or 39°C maximum depending on ambient) before turning off.  (Home is not air conditioned.)

I originally used the boxed CPU cooler from Intel, but recently bought an enhanced cooler (BXTS15A), which is about 50% taller.  This rebuild has the newer cooler.

I do periodic vacuuming for dust and allow a little to accumulate before cleaning.  I also use the CPU temperature as a gauge for when it needs to be done.

I do not overclock.  I never smell hot metal or "hot electronics.". (The side of my case is open, and I visually check the three fans while operating.)

[wraper]

Early this morning, I installed a "refurbished" MoBo and new CPU.  Boot froze at "15" on the digital monitor with beeping  (debug LED code 15 = "pre-memory North-Bridge initialization started").  Wiggled and reset the DDR3 sticks without effect.  Took a break until about 3PM cogitating about it and doing errands.  Removed one of the DDR3 sticks and thing booted correctly (what luck -- see next).  Took care of some bios details, then moved the good DDR3 stick to the other slot.  It froze again.  Tried the original DDR3 stick in the working slot (no stick in the assumed bad slot) and still working normally.  There are 4 DDR3 slots, two grey and 2 black.  Next step will be to install both DDR3 sticks in the two black slots.  If that fails, I can live with just one DDR3 stack for the time being.

You probably have a bent CPU socket pin.

[rdl]

Life is too short to be fooling around with a problem like this. I still say you should just buy a new computer. The little HP Elite Mini I bought back in April is now my most used computer. Best $200 I've spent.

https://www.eevblog.com/forum/chat/what-did-you-buy-today-post-your-latest-purchase!/msg4803890/#msg4803890