Did I kill my CPU?

[Galatian]

Hey guys and gals,

I'm having a little bit of a problem at hand here. I upgraded to an i7-4790K this summer on an ASRock Z97 Extreme9. PSU is a Corsair AX1500i and I used an EVGA GTX 670 FTW2. I have delidded the CPU and was trying to achieve a good overclock and it seemed to have been finally stable at around 4,8 GHz Core and 4,0 GHz Uncore at around 1,35 V watercooled. Now I went on holiday for two weeks and made sure my computer was off. After I came back I noticed my Computer was running but I didn't think much of it.

I then noticed that I would randomly crash in games so I checked with Prime95 28.5 and soon ran into troubles (BSOD or freezes...sometimes workers dropped). So I decided to run everything on stock again. First I simply "soft reseted" the Bios in the UEFI screen. That didn't help so I used the jumpers to reset the Bios. This didn't help either...still problems. So I thought maybe the RAM went bad. I used the HCI Memtest CD to check the RAM overnight. Result was 1046% coverage without errors. I now removed the BIOS battery and essentially hard reseted the BIOS and I also downloaded the newest BIOS. I will check if this solved my problems, but I wonder if I somehow managed to make my CPU permanently unstable with my overclocks?

Any thoughts?

 

[blake0812]

What OS are you using? Are you sure you set EVERYTHING back to stock?

 

[Galatian]

What OS are you using? Are you sure you set EVERYTHING back to stock?

Windows 8.1 everything updated. Yes everything is back to stock. I have reseted the BIOS several times now and it looks like everything is back on Auto. I can even see the CPU clocking no higher then 4,2 GHz with all cores loaded, which should be the stock setting.

 

[blake0812]

Did you check your hard drive to see if that's causing problems?

 

[Galatian]

I have an Samsung 830 SSD with 256 GB. S.M.A.R.T says no errors. Any other way to check it?

 

[blake0812]

Not sure to be honest, I would say you'd have to reinstall Windows or use Windows Recovery Options to fix it.

When you get a blue screen, what message does it say?

 

[Galatian]

Cryptic messages ;-)

I used BlueScreenView to check up on them. Latest crash this morning was from ntoskrnl.exe...which kinda like crashes for all kinds of reasons...

I will check the Samsung Magician and might to a fresh install.

 

[blake0812]

I just remembered in the command prompt (you should probably do this in safe mode) you can type chkdsk /f and that scans and fixes bad files.

 

[9enesis]

might be a video card fault or psu for that matter too....try replacing to rule out one by one.

 

[Galatian]

might be a video card fault or psu for that matter too....try replacing to rule out one by one.

CHKDSK didn't find any problems on my SSD. There seems to be S.M.A.R.T. issues with my backup HDD, but I don't believe that should be causing any problems, as there is nothing relevant on it. I will still remove it from the computer.

I will run Prime95 overnight from the iGPU and see if that did anything.

 

[9enesis]

to be honest...i,d recommend you to start with the PSU if that is possible...just because it is 1.5kw unit doesn,t mean it cannot fail at some load....and if you remove GPU in the first place, you also reduce the total load on PSU ,thus might come to wrong conclusions regarding the result....

 

[Ketchup]

CHKDSK didn't find any problems on my SSD. There seems to be S.M.A.R.T. issues with my backup HDD, but I don't believe that should be causing any problems, as there is nothing relevant on it. I will still remove it from the computer...

That is a good idea. Also, I haven't heard you say anything about temperatures under load.

 

[Galatian]

Temperatures are fine...without OC not even hitting 60°C on full Prime95 load.

 

[dbcooper1]

Try booting from a linux live source to eliminate O/S and drive issues.

 

[Galatian]

Can't even get the Linux Live Distro to run. I tried both Ubuntu and Mint. Mint in save mode says

ata10: failed to resume link (SControl FFFFFFFF)
ata10: SATA link down (SStatus FFFFFFFF SControl FFFFFFFF)

Here are two pictures

And after I removed the bad harddrive and the CD out off my Blu-Ray device.

EDIT: Well I removed all SATA devices and it seems to boot now.

 

[Galatian]

Well Prime crashes in Linux as well:

Does anybody know what the error code means?

EDIT: I had a Linux guy tell me that this looks like either the Cache is broken or the timer from the main board. He told me to run Prime95 with one one core enable to check out which one it is. Any thoughts?

 

[Galatian]

Little update on this old thread: I had a back and fourth discussion with Intel. They were willing to swap out my CPU but only by not really acknowledging that their CPU is defective. Meaning I would looses my warranty on every other replace I might have. They based this on my lack of testing it with a different mainboard. I declined because the CPU is only half a year old.

Well fast forward to today and I had to swap out my Mainboard anyway and my problems are still the same.

 

[Shmee]

Sounds like the CPU...well that is too bad. Did you get the extended OC warranty? Try contacting Intel again.

 

[Idontcare]

I have delidded the CPU and was trying to achieve a good overclock and it seemed to have been finally stable at around 4,8 GHz Core and 4,0 GHz Uncore at around 1,35 V watercooled.

Little update on this old thread: I had a back and fourth discussion with Intel. They were willing to swap out my CPU but only by not really acknowledging that their CPU is defective. Meaning I would looses my warranty on every other replace I might have. They based this on my lack of testing it with a different mainboard. I declined because the CPU is only half a year old.

Well fast forward to today and I had to swap out my Mainboard anyway and my problems are still the same.

Wait...let me get this straight, you had already delidded your CPU (invalidating any warranty whatsoever, including the performance tuning plan which is expressly invalid if you mechanically alter the CPU's packaging) and yet despite this you somehow managed to convince Intel to accept the notion that they needed to replace your walking-wounded CPU...and even then you elected to not take them up on the deal because you were worried about losing your warranty on any additional future replacements?

Srsly? You weren't due the first replacement, let alone any other future ones and yet you felt it wasn't a sweet enough deal to pursue. Words cannot...

 

[Galatian]

Wait...let me get this straight, you had already delidded your CPU (invalidating any warranty whatsoever, including the performance tuning plan which is expressly invalid if you mechanically alter the CPU's packaging) and yet despite this you somehow managed to convince Intel to accept the notion that they needed to replace your walking-wounded CPU...and even then you elected to not take them up on the deal because you were worried about losing your warranty on any additional future replacements?

Srsly? You weren't due the first replacement, let alone any other future ones and yet you felt it wasn't a sweet enough deal to pursue. Words cannot...

The point I was trying to make is this:

1. Intel will only replace your CPU with a warranty if you go through a lot of hurdles. As a consumer I don't see, why it would be a problem for me to send them my chip and let them check the CPU. Quite simply this is what my reseller would do if I would have taken it up with them (and in fact I might just do that, because it would have been way easier to begin with). Instead Intel not only want's me to check every component on my expense, they also literally told me in an Email that they consider neither Linux nor Prime95 to be indicative for checking out a CPU. They couldn't even respond to the errors messages I was getting. The solution is to run their little program for 5 minutes and they think that is indicative of a stable processor.
2. The more important point I'm getting at is that I think overclocking the Ringbus can be really dangerous. I think this is what actually "destroyed" my CPU. And this should go out as a warning for everybody considering it.

Now weather I will get a replacement or not (they will check the CPU for tempering anyway) is a different question. Is it shady? Perhaps. But on the other hand I pay a lot of money more for the K CPUs and I pay more for the warranty here in Germany. Legally it is actually still up to debate wether or not you can modify something you purchased and loose your warranty over it. For example those stickers "Warranty void if removed" are actually arguments that get thrown out of court, because simply removing the sticker or even removing a screw does not brake a product. In fact it has to be shown that an actual modification resulted in a malfunction. So in a sense delidding should only void the warranty if you really destroy the CPU in the process. And in that sense me overclocking the CPU might have resulted in the errors now popping up and if that is the case I do have the warranty for overclocking. Your mileage and views might differ in the US.

 

[Ketchup]

Legally it is actually still up to debate wether or not you can modify something you purchased and loose your warranty over it. For example those stickers "Warranty void if removed" are actually arguments that get thrown out of court, because simply removing the sticker or even removing a screw does not brake a product....

You are generalizing. Every product comes with its own warranty, with its own terms. Call Intel back and take them up on their offer before they change their mind.

 

[tolis626]

The more important point I'm getting at is that I think overclocking the Ringbus can be really dangerous. I think this is what actually "destroyed" my CPU. And this should go out as a warning for everybody considering it.

Ermmm... Isn't the 4790k's cache clocked to 4.0GHz by default? I have mine overclocked to 4.3GHz (Down from 4.4 or even 4.5GHz I've used previously) using 1.22V (1.249V measured by HWMonitor) and have had no problems whatsoever.

With that said, and even though there seems to be a problem, do not use Prime95. Try something else, like OCCT, Asus' RealBench, Aida64 etc. Prime95 28.5 and later can fry your CPU for breakfast, and isn't really indicative of a realistic load. Not by a long shot. Apart from that, 1.35V is pushing it. I currently use 1.305V and max I've used is 1.317V, and that's already too much. I'm using a Corsair H110, but that's hardly the point. Keeping temperatures in check doesn't mean you're ok. Degradation can just as easily happen by too much voltage. Although I have to say that it did happen quite fast.

 

[BonzaiDuck]

One of your screenies that looks like a command window shows a BSOD "stop code" of 124. This is a "general hardware" failure and can mean anything, including VCORE or [VCCIO/VCSSA/ etc.] IMC voltage either too high or too low.

124 is not by itself indicative of some CPU damage or failure: I'm testing out a CPU at the moment with prelim OC settings, got a BSOD-124 and VCORE or LLC was too low. And this for modest 4.3 Ghz clock.

But the circumstances of your de-lidding make a whole bunch of shee-*t possible. Even so, the de-lidding would not have rendered the CPU damaged if you did it carefully. But -- yeah -- the "warranty pickle" -- the big green weenie. It's a completely different thing to lap a processor down to bare nickel: hardly any risk of damage, but no . . . freakin' . . . warranty. De-lidding -- another problem entirely.

 

[Denithor]

I'm guessing that even if you take them up on the offer to replace, once they receive your chip and see it has been de-lidded - they're going to just send it back to you and claim it was either physically modified or damaged and they are not responsible anymore.

 

[Galatian]

Hey seriously I don't have a problem purchasing a new one. I always plan on that possibility. It the price you pay when you want to overclock.

The i7-4790K was priming without problems on stock when I purchased it and it was also doing it after delidding it and on 4,8 GHz Core. It was only when I started to push the Uncore that I received issues. Even after I reset the BIOS Ty default the problems still persist. The Linux error issues indicate that either the Cache or the Timer from the Mainboard is faulty, but since I exchanged the mainboard and checked every other component it pretty save to assume that the CPU degraded. Now is is because I pushed close to 1,35 V? Perhaps but that was Core Voltage and other people seem to be running fine on it.

Prime 95 28.5 is an extremely heavily load. But that's the point if it is stable under these conditions voltage and temperature wise it's save to assume any other load will require less voltage and produce less temperature.

I believe Ringbus overclocking is just really not useful at all, because you only get small increases in synthetic benchmarks. But not only that, I seem like it can be really dangerous and it seems to be more sensitive to high currents. Oh and I was trying to get more core and Uncore in sync meaning 4,8 GHz for the Ringbus.

4,6 GHz was fine for both off them, but well I got greedy...so I'll have to pay the price.