KILL your i7 980X AT ONCE (and RMA it)...

[Deux]

Just fired up my worst chip with one core on both a UD9 and E760, chip still runs fine afterward. Will keep running it with one core for the next few days and will report back.

[FurballRZ]

if you whrer my friend and did that, id have (f)ucking killed you) still wierd that something like this happened
two interesting facts/ Qs from this
1. if some 980x frome one batc are bad ocers and othhers are good, wouldnt this mean the entier batch stuff is a hoax
"There are a few 3005F's doing good, but the vast majority of that batch sucks b@lls.
Here's one of the good 3005F's"
2. could i be it be that the chip dies because you haqve the full 1+v on one core, and one thread so it wouldd be like running the cpu on 12v (obvioussly not quite because of the cache ect, but you get my drift)

[BenchZowner]

1. There analogy of good 3005F's against the analogy of good 3003B's is what we should use as a "measure", and we have lots of good 3003B's while we have a very limited few of good 3005F's.

2. Doesn't matter how many cores you have enabled, each core gets the same amount of voltage each time ( they all get the same voltage, the one you set the BIOS option at )

[Sam_oslo]

I7-batch differences is real, it is not a hoax. It has always been some variation in transition-batch#, the transition from good to bad in the (reverse) evolution, and the 3005 seams to be the missing link.

This 1C/1T issue may have something to do with Intel's brand new Turbo Boost on the Core i7-980X?. On single-threaded applications, one Core get 2 speed bins (266MHz) while on multi-threaded applications, it will just be a slight 133MHz increase at stock. This 2-bins bump will translate to even bigger frequency increase in high-BCLK benching (EDIT: and maybe this gives a big voltage bump to that single core too?). I don't know how this 1C/1T degradation/dead happens, but this 2-bins Turbo Boost is a brand new future and may somehow be the cause?.
EDIT2: In case, and in some other possible cases too, the same issue would raise whenever you run a single-treated applications that sets a 100% load on one core, even if you are running 6C/12T, I would claim.

EDIT: some edits about voltage bump after reading M.Beier's comments below VV

[M.Beier]

If 1C/1T benching can kill/degrade the chip, then I'm wondering if applications that use only one Core can have a similar effect even if you are running 6C/12T?

EDIT:
I had seen this thread before, and got scared when Everest Memory Bench had 100% load on one Core while all other Cores were absolutely stuck to 0%.

Hmm, was thinking maybe in spikes 1 core is exposed to the power for all 6 cores? Would make sense to me...
But Im not expert

[[XC] Oj101]

That's a very expensive lesson to learn, George. If I were your friend we would no longer be friends

[unclewebb]

The Core i7-920 can also use 2 bins of turbo boost when running a single core with the other 3 disabled or in the C3/C6 sleep state and they seem OK. The Core i7-720QM mobile chips can use up to 9 bins of turbo boost so I don't think the turbo boost that the i7-980X is using is anything unusual for Intel.

[xpatar]

If 1C/1T benching can kill/degrade the chip, then I'm wondering if applications that use only one Core can have a similar effect even if you are running 6C/12T?

EDIT:
I had seen this thread before, and got scared when Everest Memory Bench had 100% load on one Core while all other Cores were absolutely stuck to 0%.

I've been using Furmark for hours testing the stability of my vga card and I didn't notice any degradation in performance of the chip. Furmark uses 1C/1T and my cpu is running at stock at the time with 6C/12T enabled.
My cpu has a batch number 3003B320. What a pity those chips are dying
Until this problem is resolved I won't even dare using 1C/1T through system bios.

[Rage]

I didn't think there was anything wrong with using a single or only two cores w/o HT...surprising...

Unfortunate about your friends 3003B....I think 3003Bs are dying off the shelves and are being replaced with 3005Fxxx

I must have gotten lucky then.. I orded a 980x last week and got it a couple days ago and it's a 3003B batch.

[charlie]

in all these years I think this is literally the first BIOS-kill method I can remember that wasn't related to voltage.

[[XC] Oj101]

back chief, I haven't seen you for a good six months

[prznar1]

This issue apply only for 980x of other i7s too? I'm running 4 cores HT OFF underclocked, webrowsing, office work. I'm sure some people are using 2 cores when downloading files or something.

Thank you George for the info. Hope you have no problems with RMA.

i would not be bothering my self with 45nm cpu. but the rest of 32nm cpu range, are needed to be checked.

[charlie]

back chief, I haven't seen you for a good six months

been a while... I'm researching new toys to play with

[prznar1]

been a while... I'm researching new toys to play with

like those on your avatar?

[charlie]

like those on your avatar?

darn nice silicon(e)

[prznar1]

darn nice silicon(e)

it will never be that good as it looks like

[[XC] Oj101]

been a while... I'm researching new toys to play with

Whatchagot?

[Kinc]

Me and elmor tested this during the weekend. CPU is still alive and well

[xpatar]

Still playing with 1C/1T through windows with hyper pi having 6C/12T enabled. At the moment not an issue.
This is insane. Could this be a design problem of the processor?
Or a bios fix could solve the problem? Not just deactivating the option of using one or more cores through system bios.
Intel should respond immediately, we're talking about a processor that costs a 1.000 euros. That kind of products shouldn't be affected that easy.

[QuadDamage]

I couldn't get my 980X ES to post with just one or even two cores enabled on Asus P6T DLX V1 and 2104 BIOS and always had to reset the CMOS after that. I guess i'm lucky i didn't kill the chip... I'd prolly spend the next few days wondering what the hell happened.

[Leeghoofd]

Is it worthwhile to put up a list to gather info :

Mobo used and bios version
CPU settings : Bclock x Muliplier, cores set... HT on/off
Vcore, QPI voltages

tests run : 3Dmark, Vantage, Pifast,...

[xpatar]

Even though at first I wouldn't even try it, I decided to give it a try.

What the ........ If anything goes wrong I'm going to rma

Currently on air and everything @ stock speed.

Bios v. 1003, batch number 3003B320.

Bios settings: 1 core HT off,



CPU still alive


Bios settings: 2 cores HT off,



CPU still alive


Bios settings: 6 cores HT off,



CPU still alive and kicking

I rebooted, set into bios 6C/12T, restarted and booted into windows.

Everything works as it should.

It seems not every Gulf is affected.

[FatAlbert]

maybe try longer time with some cores off ? one week with 3 cores + ht ? or maybe quad core gulfy without ht?

[K404]

I'm pretty sure i've ran 1C/1T for 3D05 and my chip still works fine... I run 2C/2T VERY frequently and... no difference. I'm still finding new (higher) bench points as I get used to the chip

[Splave]

tryed 2c2t and 1c1t for spi 6.2ghz 1.875v on my a0 with RIIIE still works daring arent I?