Question 4790k dead, but only kinda?

[Ketchup]

Kids computer room is running a 4790k. Old, but still plenty of years, or so I thought. Yesterday afternoon it failed to start Windows. It would load to the login screen, then immediately restart, like someone hit the reset button. The Windows install had been from an older Intel CPU, 2500k, so at first I thought perhaps an issue with that swap had finally surfaced, or one of the SSDs was going, as they are getting older. Did a fresh install of Windows just fine with only one drive plugged in, but as soon as it starts installing drivers from Windows update, the issue resurfaces. Swapped power supply and video, same deal. Moved to a different room to connect to different devices and used different USB ports, same thing. I was starting to scratch my head a bit when I decided to run OCCT and the Intel stress test. Those both caused a hard power down/restart in less than a second after I started them (and this was just on integrated video). I checked the settings for the CPU fan and it is set to 100% by 70 degrees and the bios does recognize the fan is spinning. The failure on the CPU tests make me believe the CPU is failing.

I have only had one CPU fail on me personally (over 20 years ago), and only a couple on CPUs on PCs I have worked on. Since it runs ok without driver "enhancements", but fails with a stress test, I am wondering if some cache has gone bad, but have never heard of that. I was curious to see if anyone here has ever experienced something like that.

 

[In2Photos]

I would reinstall the CPU cooler for starters and reseat all memory, maybe even use a pencil eraser to clean the contacts. Check the CPU socket while you're at it.

 

[aigomorla]

but as soon as it starts installing drivers from Windows update, the issue resurfaces.

This sort of leads me to think something else in your system is causing the fault.
Might be a bad board, or bad ram.... more leaning on bad ram as DDR3 was definitely not my favorite generation, however they did greatly improve over DDR2 in failing.

Those both caused a hard power down/restart in less than a second after I started them (and this was just on integrated video)

It could be the ram again.
Can you go see in windows Event Viewer what the fault was?
Or will it not boot into windows anymore after trying to install update?

I am wondering if some cache has gone bad, but have never heard of that. I was curious to see if anyone here has ever experienced something like that.

Devils Canyon ran stupidly HOT, because it was the first generation Intel went HAHA~ *nelson laugh* and used non solder TIM on the IHS.
This really ate at thermals, but Intel said it saved the die from micro factors from indium solder applying micro stress on the die.
A lot of us felt this was BS, and delided them along with using Liquid Metal to seal it back up.

But that cpu is old.
Launched in 2017, its given you a good 6 yrs.
Fortunately they are not too bad on the used market in EBAY if you need a new cpu.


But i am thinking in order..
RAM -> Board

Memtest the Ram.
If it fails memtest, your problem is either the board or Ram.
If you can't memtest, try 1 stick only and swap them out if you see problems.
If you only get problems on 1 ram stick... its the RAM.
If its repeatable on both ram sticks single, then you need to do more diagnostics.

Also i might unseat the CPU and inspect the pins.
Then reseat it gently, as maybe the cpu shifted somehow, and a pin is touching something it shouldn't.

 

[VirtualLarry]

Got some budget HP G3 SFF i5-6500 w/4GB 350(X) LP GPUs, Win10 Pro (tattoo'd from Factory), 256GB NVMe + SATA SSD for game storage, 2x8GB dual-channel DDR4. LMK if interested. Cheap!

 

[coercitiv]

I was starting to scratch my head a bit when I decided to run OCCT and the Intel stress test. Those both caused a hard power down/restart in less than a second after I started them (and this was just on integrated video). I checked the settings for the CPU fan and it is set to 100% by 70 degrees and the bios does recognize the fan is spinning. The failure on the CPU tests make me believe the CPU is failing.

Or the board. You can visually inspect the board for any sign of failure in the VRM area (inflated caps), or you can add a Vcore offset of 20-50mV and see if that temporarily improves stability (like failing after 30-60 seconds instead of 1-2 seconds). If you don't like the idea of pushing more voltage, you can limit max clocks instead.

 

[Ketchup]

Some good ideas here, thanks all.I don't think it's the RAM, mainly because I have seen plenty of RAM crashes over the years and there is always at least one blue screen. This one goes straight to power reset. Which would point to the board, power supply, or CPU -but I have never seen a bad CPU load windows, then crash, but I don't know what all is involved in loading Windows - maybe it's a single-core operation. Ruled power supply out, so that leaves the parts that it's time to replace. Although it was a neat idea, I wish Asus had not decided to cover everything on that board in plastic. If I were to visually inspect everything, that means taking out a LOT of screws, and honestly I don't know if I want to mess with it. Many good years out of that build.

 

[Ketchup]

Well, figured it out. Core 3 is dead. Asus lets you run 1-4 cores, but only in order, so they will have to settle for a dual-core 4790k for now.

 

[VirtualLarry]

Well, figured it out. Core 3 is dead.

Wowzers. Cores a-dyin'. Who would have figured. Good work.

 

[Markfw]

Well, figured it out. Core 3 is dead. Asus lets you run 1-4 cores, but only in order, so they will have to settle for a dual-core 4790k for now.

Time to upgrade to a 5800x (cheapest and most affordable power CPU IMO)

 

[A///]

First time I've heard of an individual core die. Usually there's other parts of the Intel processor that are bound to die first, but often only when exposed to clocks or voltages that are unnatural to the processor. Haswell was the first tri gate platform for modern intels wasn't it? Not bad to see 10 years before something failed imo.

 

[VirtualLarry]

FIVR, too. Maybe it hot-spotted itself to death, slowly over time.

 

[Ketchup]

FIVR, too. Maybe it hot-spotted itself to death, slowly over time.

This is what I am leaning towards. Perhaps the heat spreader wasn't quite tight/close enough out of the factory. Realizing this is a hot CPU anyway, it always did seem to run a bit hotter than it should, based on what others reported back in the day. I had an EVO cooler on it for several years, re-attached a couple times, but it would always throttle at the full 4.7 GHz under full load (such as OCCT). I got a better Thermalright cooler for it a few years ago, but by then of course, it would have been too later for any damage that was already starting. The only clue was that, over the years, this same phenomenon (hard reset) would happen every few months while the kids were gaming. At the time, I thought it was the aging GTX 660, but it continued after replacing that card, albeit a bit less often, so that makes a little more sense now

 

[mikeymikec]

Can you update the thread again in a couple of weeks' time to say how it's going? Overall this seems to me like a situation of "I think I've fixed it! It's all going well! It's.... oh. I've still got a problem".

I'm also curious to know how you figured out which core has died.

 

[Ketchup]

Sure. Unless the boys have bought replacement parts before then, in which case I will update before I remove it.

 

[A///]

FIVR, too. Maybe it hot-spotted itself to death, slowly over time.

It is possibility... I'd had a 5930K that had similar issues about a year after ownership that I sold it off at a huge loss and bought 1st gen skylake. idr of the idea of a failed cpu crossed my mind then but even if it had i wouldn't have put much thought into it because we were well past the time period where mobos and cpus would fail @ random. your psu has a higher chance of failing today compared to anything else.

 

[DAPUNISHER]

I have no difficulty accepting a 9yr old overclocked CPU used regularly, has become faulty.

 

[A///]

I have no difficulty accepting a 9yr old overclocked CPU used regularly, has become faulty.

I find it remarkable it comes down to that. I've ran high clocks on air systems way back before aio's became a thing on the market and they still run the clocks today. they struggle running windows 10 of course because of the processor lacking certain instruction sets where deprecation takes a stand and makes it slow but still usable in 99% of cases. I'm not saying its not degraded, I fully expect it to be since 4th gen was the first with a tri gate. I believe my old haswell -e used trigate tech too and as said that was giving me problems so I updated to skylake. was a nice gut punch of spending that.

 

[DAPUNISHER]

I am no IEEE guy. Hell, I don't even armchair that stuff. My practical experience is that a CPU can fail, degrade, or become flaky/unstable. I don't know the why of it. heat damage? single event upset? gremlins?

 

[mikeymikec]

I have no difficulty accepting a 9yr old overclocked CPU used regularly, has become faulty.

I just re-read ketchup's posts and see nothing to suggest it was ever overclocked...?

 

[DAPUNISHER]

I just re-read ketchup's posts and see nothing to suggest it was ever overclocked...?

but it would always throttle at the full 4.7 GHz under full load (such as OCCT).

Max turbo on the 4790k is 4.4GHz, so either he has a typo or it was overclocked.

 

[Ketchup]

July and still running (semi) strong on 2 good cores.

 

[Shmee]

I would grab a 4790k off of ebay for cheap, as a replacement. This does sound like a CPU issue, and honestly I have a friend @faxon who had an issue with a 4790k he had. We thought it was the board, but turned out to be more of a CPU issue, but maybe also partly board as well?

If you can't fix it with a CPU drop in, I would go ahead and do a full system upgrade most likely, probably using a B550 and 5700X.