Defective processor core?

[Lenne]

Hi friends!

The question is on the title. I simply can't have my system stable with both cores enabled. And since this is a locked CPU, it hasn't been OC'd. With 1c/2t everything is working stable. What I have tried:

-Two different motherboards (Asus Z170 Pro Gaming & Asus Prime Z270-P) with both having the newest BIOS
-Two different kind of DDR4s (2x4GB Corsair Vengeance LPX 2400 & 4GB Crucial Sport LT 2400) with and without XMP, and of course just one stick on all the slots without any help
-Three different graphics cards (MSI GTX780Ti, MSI HD7770, Gigabyte GTX560Ti 448)
-Also with the CPU's integrated GPU
-With and without my PCI-E soundcard
-Countless fresh Win 10 Pro installs
-M.2 and SATA SSDs
-Two PSU:s (Corsair HX620, Antec HGC 620W)
-Power savings on and off

Just resetted the bios via removing the battery and every important setting relating to CPU is on default, except like I said, it's running on 1c/2t. Better than nothing, I guess.

But thanks for tips & tricks already!

 

[qubit]

Try running with one memory stick and if that doesn't work, swap the stick and memory slot.

Try underclocking it a bit if possible.

See if these two things make it stable. We can then think of the next step to take.

 

[basco]

do ya mean with hyperthreading on\off?
or 1 core with ht?
does more volt help?

 

[Bill_Bright]

Try running with one memory stick and if that doesn't work, swap the stick and memory slot.

The OP already stated he tried,

Two different kind of DDR4s...
...and of course just one stick on all the slots without any help

He also tried two different motherboards, different drives, two different PSUs, and reinstalling the OS.

So IMO, basco, you have tried everything you can and through a process of elimination, have narrowed it down to the CPU. The only thing we don't know is the model number of the CPU. You need to check the QVLs (qualified vendors lists) for each motherboard (found on the motherboards' websites) to ensure that CPU is listed to ensure it is compatible with those boards.

Bottom line, you should not have to change any BIOS settings from the defaults to have a stable system with all cores, assuming the CPU is compatible and is properly cooled with a correctly mounted CPU cooler and adequate case cooling.

 

[Vya Domus]

What do you mean by unstable ? Reboots , shutdowns , BSOD ? Having a defective core sounds unlikely , and to be honest I haven't even heard of such thing. If you have the option to RMA , do it as you have already tried many things that could have been the possible cause of this.

The only thing that I can think of at this stage is overheating. However you are using something that should be more than enough. Maybe this is the worst case of poor IHS thermal paste from Intel ?

 

[thebluebumblebee]

The one potential point of commonality that I see is the W10 install source. You need to try another download, and/or switch out thumb drives.

 

[Caring1]

Try this.
https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool-64-bit-

 

[Bill_Bright]

Having a defective core sounds unlikely , and to be honest I haven't even heard of such thing.

It happens. It is not that uncommon, though typically they are found before the processor leaves the factory. In fact, AMD is known to rebrand some of their quads that failed testing as triple cores. This may sound like a deception but actually it is a sound business practice, and not uncommon either. It avoids a total loss for AMD, and consumers still get a decent CPU at a very decent price.

I note RAM and disk makers have been doing similar things for decades. I first learned about it with floppy disks. They were manufactured to be double sided and double density, for example. But if testing showed one side was bad, it would be sold as single sided. Or if density tests failed, sold as single density.

Good idea to use the Intel Diagnostic Tool.

 

[Solaris17]

iv ran into bad CPUs before. The last one I legit had an issue with was a 6500k. The other 2 CPUs had bad integrated GPUs but worked perfectly with it disabled. It is possible.

 

[Lenne]

do ya mean with hyperthreading on\off?
or 1 core with ht?
does more volt help?

Exactly, 1 core with HT on. Tried up to 1.4 volts and also bumped other voltages a bit.

And yeah, BSODs with almost every time with a different error message.

I guess I could try to make a new install media to my USB drive, but all these shitty problems began right after I changed from G4400 to this G4560. IIRC this worked two days without problems, but I remember talking on phone with my friend when I started to get BSODs.

And it's kinda hard to try that Intel diagnostic tool since this won't even boot to Windows anymore with two cores activated ( ), but since I kinda don't have anything better to do, I could try a fresh install with a new media.


So much thanks for the tips, let's download the Windows Media Creator or whatever that was..

I'll report what happens next. If happens.

 

[thebluebumblebee]

but all these shitty problems began right after I changed from G4400 to this G4560

Then ignore what I said and RMA the CPU.

 

[cdawall]

I would RMA the CPU and move on.

 

[Lenne]

I guess so, I've already RMA'd the Corsair 2x4 kit and Asus Z170 and have to pay for those "non-needed RMAs". I guess I'll get a new one, put this on RMA and sell it right after it comes from RMA.

Anyway, thanks guys.

 

[Vya Domus]

It happens. It is not that uncommon, though typically they are found before the processor leaves the factory. In fact, AMD is known to rebrand some of their quads that failed testing as triple cores. This may sound like a deception but actually it is a sound business practice, and not uncommon either. It avoids a total loss for AMD, and consumers still get a decent CPU at a very decent price.
Good idea to use the Intel Diagnostic Tool.

I am aware of this , but it's not the same thing. Those aren't really defective CPUs that are being sold, they are just cut down dies with portions of them disabled regardless of whether or not they were defective , they are designed to work like this. What we are talking here about is a CPU that that didn't go through this processes. The thing with these chips and x86 architecture is that if there is any kind defect it will result in complete fail at boot , also considering how effective the binning process is a half-broken CPU is something very strange indeed. This might just be some very strange microcode/BIOS bug related issue instead.

 

[sneekypeet]

Just on the off chance, have you verified in Device Manager that you do not have both CPUs showing? I seem the vaguely recall something like this happening to me when i was swapping Conroe CPUs a lot. Shot in the dark, but may be worth a quick look.

 

[Lenne]

Just on the off chance, have you verified in Device Manager that you do not have both CPUs showing? I seem the vaguely recall something like this happening to me when i was swapping Conroe CPUs a lot. Shot in the dark, but may be worth a quick look.

Like I said, I've reinstalled W10 Pro at least five times when finding out what's the problem, so.. :/

Also got this Prime Z270-P a week ago and of course a fresh install.

 

[Bill_Bright]

I am aware of this , but it's not the same thing. Those aren't really defective CPUs that are being sold, they are just cut down dies with portions of them disabled regardless of whether or not they were defective , they are designed to work like this.

I was replying to your comment where you said you have never heard of a defective "core". Never hearing of something does not mean it does not happen - the fact is with processors, it does happen. Not often, but it does.

And it is the same thing. AMD does not design and manufacture quad cores for the purpose of selling every one of them as triples. That would be a HUGE waste of very precious and expensive die material - and a big loss of potential profits. So rather than tossing the processor in the trash for a total loss, they just disable the defective core and sell it as a triple - as long as the remaining cores work fine.

As I said before, this (dummying-down of specs) is a very common practice across all sorts of industries. This is even taught in industrial business classes.

RAM modules are often manufactured to run at the fastest speed the manufacturing process and raw materials allow. Then they are tested and those that cannot run at those maximum speeds are packaged and sold as slower speed modules - at a lower cost, but still at a profit.

Other components are manufactured to the tightest tolerances, but those that can't meet those specs are sold as cheaper versions.

@9700 Pro - thanks for sharing this CPU is a G4560. According to the QVLs, it is compatible with both boards. So it is not that.

Do you still have the G4400 you can put back in to see if the problem goes away?

 

[Derek12]

It happens. It is not that uncommon, though typically they are found before the processor leaves the factory. In fact, AMD is known to rebrand some of their quads that failed testing as triple cores. This may sound like a deception but actually it is a sound business practice, and not uncommon either. It avoids a total loss for AMD, and consumers still get a decent CPU at a very decent price..

Similar to Intel Core Solo



Obviously CPUS can go bad... They do not use different technologies than GPUs and most other ICs. If they fail, why CPUs can't?

 

[Vya Domus]

I was replying to your comment where you said you have never heard of a defective "core". Never hearing of something does not mean it does not happen - the fact is with processors, it does happen. Not often, but it does.

And it is the same thing. AMD does not design and manufacture quad cores for the purpose of selling every one of them as triples. That would be a HUGE waste of very precious and expensive die material - and a big loss of potential profits. So rather than tossing the processor in the trash for a total loss, they just disable the defective core and sell it as a triple - as long as the remaining cores work fine.

As I said before, this (dummying-down of specs) is a very common practice across all sorts of industries. This is even taught in industrial business classes.

RAM modules are often manufactured to run at the fastest speed the manufacturing process and raw materials allow. Then they are tested and those that cannot run at those maximum speeds are packaged and sold as slower speed modules - at a lower cost, but still at a profit.

Other components are manufactured to the tightest tolerances, but those that can't meet those specs are sold as cheaper versions.

@9700 Pro - thanks for sharing this CPU is a G4560. According to the QVLs, it is compatible with both boards. So it is not that.

Do you still have the G4400 you can put back in to see if the problem goes away?

I didn't say it isn't possible , but it's certainly very unusual. The complexity of these chips has reached a point where a flawless design is impossible , OP's CPU might be bad , but I still think this "defect" isn't a hardware thing but rather a very rare low level software bug.

Also the reason why AMD is following this practice with their binning process relies on them to manufacture a high core count chip just so that they can disable some of them even if they have 100% yields. Believe or not this is way more cost effective than having separate designs. Think about it , Ryzen's 8 core variant will not be their best selling chip , those spots will go for the 4/6 core variants which come from the same die which in most cases is probably still 100% functional . Same thing with the previous FX line. A complete waste would have been if they only relied on defective chips for their lower end product stack.
So I still believe is not the same thing compared to what Intel dose.

But at this point I'm getting rather intrigued by your firm belief that CPU failure is not unusual since I have never came across one and never heard of a concrete case either ( obviously excluding OC failures) . Maybe you have? I'm really curious.

 

[Bill_Bright]

Believe or not this is way more cost effective than having separate designs.

I do believe it. In fact, it enforces my point.

The complexity of these chips has reached a point where a flawless design is impossible

It's always been impossible. Man cannot create perfection 100% of the time. And while manufacturing techniques have vastly improved, they are still jamming 100s of millions, even billions of transistor in less than 200mm². And that is only going to get denser and denser.

As I noted, I first encountered this practice with floppy disks when I toured a BASF floppy disk factory in Germany way back in the mid 80s. Every floppy was manufactured as DSDD (double sided, double density) then tested. If testing showed problems, they would be then be tested as single sided and/or single density. I then learned about how widespread that practice was across many industries when I went back to school.

This could be a software "bug" but typically bugs affect many, or even all. I think there is a defect in this particular CPU. Perhaps it came that way from the factory, or it was damaged via ESD or some other abuse.

 

[TheoneandonlyMrK]

Esd is an under rated bitch.

 

[Bill_Bright]

Esd is an under rated bitch.

Yeah it is. The problem is for some, if they can't see, hear or feel something, they don't believe it exists! So they don't take the necessary precautions.

But that's why ESD is such a PITA. An ESD can be so tiny we (as humans) cannot see, feel, or hear any discharge, yet the arc of voltage can be of such high potentials (voltage), it can torch a Grand Canyon sized trench (microscopically speaking) through 1000s, or even millions of transistor gates in processors, memory chips and other high-density ICs. It then takes a powerful electron microscope to see the damage.

 

[basco]

i had a 980x ES version and after a bench session i could only boot with 1 core+HT for the rest of its life.
my only cpu that did that.

 

[Lenne]

@9700 Pro - thanks for sharing this CPU is a G4560. According to the QVLs, it is compatible with both boards. So it is not that.

Do you still have the G4400 you can put back in to see if the problem goes away?

You've just could check my system specs

Nah, I sold it just before I bought this and shipped it at the same time I got this from the post office. Going to get a new G4560 tomorrow, I'll report what happens.

I would have got one already but didn't have time to go to the city before the local PC stores closed. Got money after 16.00.

 

[Bill_Bright]

You've just could check my system specs

I never assume the listed specs are the ones under discussion unless the OP states it. Many posters have more than one computer or help friends and relatives with their computer problems.

Going to get a new G4560 tomorrow, I'll report what happens.

Let's hope that sheds some light on the problem.