P55-UD6 socket burn

[Leeghoofd]

And what about Gigabyte/Foxconn ? The main issue seems to be there ? Would we be sure this would not happen at a mild OC of eg 4ghz on air ? Only time will tell

[Raja@ASUS]

As soon as you OC (frequency) you're out of warranty, the warranty VID is based upon stock operating frequency of the processor in question. So, no, we're not covered or eligible for a return unless someone is having a good day at the RMA lounge.

[massman]

No, I don't believe so....It's not "their style"...

[saaya]

hmmm makes sense george... pins are not flexible enough to keep the pressure against the cpu pads constant, lower pressure = worse contact = higher resistance = higher temperature = pins lose even more flexibility and force to push up against the pads... and the more pressure you put on them the easier it happens...

but it always happens at one spot, so i think the socket distributes the pressure unevenly as well, which doesnt help...
so while the socket clamp down mechanism works fine in normal situations, here it seems to be a part of the problem, otherwise vcore pins would fail at random locations...

and the socket is foxconn... at least on the earlier versions of the ud6 it was?
http://media.photobucket.com/image/g...6bp1S-back.jpg

francois, if anybody can tell you something about the properties of a contact pin/pad by looking at its shade and color, its probably george

and i dont think overclocking as in increasing the clockspeed voids any warranties, increasing voltages and temperatures is what damages logic, not increasing its clockspeed... if your cpu runs within specced temp and voltages, i dont think the warranty is voided and intel will refuse an rma...

[Raja@ASUS]

I trend to agree that over a certain number of VREG, there are not much difference, never the less, in cornet case , adding voltage source and current source in parrallel (as explain in Norton theorem ( http://en.wikipedia.org/wiki/Norton's_theorem or Thévenin for the hell of it ) , there is a benefit when the work load is heavier ... In the case of overclocking, it will cause more stable current when you are pulling hard on it.
It can cause some current rush too, so, it is an interesting trade between cost and perfection.

When you look at what overclocking is about, it is about making hundreds millions of transistors pulsing faster than their specs, having a more stable VCC is a plus ... it is hard to do the science behind this, but a variation of few mV during a massive pull of 200W can cause transistors to stop reconizing the same switching point over the dice, due to electric propagation. This is how multiphase actually help.
If you want to check the minimum Phare required, it is fairly simple, remove them one per one, and run at 4.4Ghz ... you will discover than under a certain number, the VREG warm up, start ocillating, and cause crashed ... I did this study on Skulltrail ... I havr no data for P55.

I think this is largely based upon the switching frequency of the circuit and the current handling per phase (and obviosulty including deoupling). A greater number of phases will reduce stress by having a shorter duty cycle per phase (although max current capability may still be limited if the company uses small and cheap fet's). Output impedance per phase wouldbe a point of contention I guess, although ultimately one has to come back to total current handling and bandwidth.

[Russian]

OK, board and cpu is working but will it clock and be benchable to similar speeds as before?




Mate, it is not just genuine vodka It's one of the best vodkas and my favorite
Do they sell this in Finland too?

I found it funny that someone from Finland would dare drink something made in russia. especially if it was called Ruski Standard.

it did obviously run out of its specs ... so no RMA. Fair!

I would tend to agree. I mean really, how many overclockers can successfully claim that their burn out chip as a result of LN2, etc. Can be RMA'd through normal channels while openly saying that they overclocked the chip and burnt it out. Nobody.


As for the whole number of phases, we all know that the number isnt really what matters. Nor does whether its digital or analog... its something different altogether considering we have seen analog pwms perform better than digital ones, especially those with really good switching frequencies.

[hipro5]

Shi(f)t!!....Why did they do it?...


P.S. By RMA I ment a "regular working CPU" that after some time (maybe months) in this same socket gets burned like this....

Let's see it someway different.
A reviewer changes MANY times CPUs on a mobo.
After some times the pins of such a socket looses their flexibility and one day he burns a CPU AND a mobo he has bought.
CAN he claim an RMA for both CPU AND Mobo?

[K404]

If the rig was ran at stock and within warranty settings... why should it be refused?

Warranty doesnt state that RMA will only be honoured if the rig is not rebuilt more than X no. times.

As long as this doesnt start popping up at stock, I dont see what the big problem is.

We OC, we push voltages and temps... we run the risk, we take it like men when things fall apart....although it is slightly 2-faced when things that are advertised by companies are warranty-voiding if we try them out

[kiwi]

I don't think you can. Obviously this socket that wears out does not meet Intel spec anymore. You can claim mobo RMA though because they made a y socket. But then again, this applies only if stock cooling was used. No definite answer here.

If my PSU runs out of spec and burns all my system I can't claim RMA, can I? Only PSU rma but not the whole system. Same here.

I found it funny that someone from Finland would dare drink something made in russia. especially if it was called Ruski Standard.

Well, you remember an old joke about Russia. They make only 2 good things: weapons and vodka

[Aerou]

so I'll put up a 24/7 WCG machine with an i7 860 and wait and see, if it goes into smoke, right?

[youngpro]

hey vivi someone said VASELINE in this thread... just thought youd want a heaps up man!

[saaya]

cirthix, yes, but not every vodka and weapon from russia is good
what they really do well in russia and eastern europe are... women :p
ahum, enough ot

i thought we already ruled out that its a gb thing cause other boards are affected too?

[Eeky NoX]

Umh! I'd like this issue to get fixed (or better say resolved) before I purchase the mobo

Hurry!! Just one month now ! If it was proven that's the socket fault...

...and than manufacturers would create a new revision or we'd have to play "russian roullette" so ?

[BikerKS]

It seems like @no advantage for 24 phases on LGA1156, if the pins are the bottle neck for higher amps . I meen Gigabyte makes good work, Intel/Foxcon engeneering(socket) .

[Drwho?]

and i dont think overclocking as in increasing the clockspeed voids any warranties, increasing voltages and temperatures is what damages logic, not increasing its clockspeed... if your cpu runs within specced temp and voltages, i dont think the warranty is voided and intel will refuse an rma...

I recommand you go and read the license agreement of Intel OR AMD processors before you speak.

Francois

[massman]

it did obviously run out of its specs ... so no RMA. Fair!

Obviously?

But, fair point: this chip should never be RMA'd. Wouldn't try it anyway/

[Raja@ASUS]

I recommand you go and read the license agreement of Intel OR AMD processors before you speak.

Francois

I think he overlooked operating frequency bearing direct relationship to power when he made reference to 'logic damage'..lol. I'd imagine this is the major contributing factor to needing warranty stipulations outlining the warranty VID range at stock operating frequency.

[xt0m]

I recommand you go and read the license agreement of Intel OR AMD processors before you speak.

Just did so for the heck of it

Excerpt from http://download.intel.com/support/pr...r_warranty.pdf

EXTENT OF LIMITED WARRANTY
Intel does not warrant that the Product will be free from design defects or errors known as “errata.” Current characterized errata are available upon request. Further, this Limited Warranty does NOT cover:

Any costs associated with the repair or replacement of the Product including labor, installation or other costs incurred by you, and in particular, any costs relating to the removal or replacement of any Product that is soldered or otherwise permanently affixed to any printed circuit board; OR damage to the Product due to external causes, including accident, problems with electrical power, abnormal electrical, mechanical or environmental conditions, usage not in accordance with product instructions, misuse, neglect, alteration, repair, improper installation, or improper testing; OR any Product which has been modified or operated outside of Intel’s publicly available specifications or where the original identification markings (trademark or serial number) has been removed, altered or obliterated from the Product.

Now, zapping to i7-800 series product instructions (installation and usage section) indeed voltage settings are mentioned, but absent is changed clock speed, so I'd think either "alteration" or "improper testing" could cover overclocking without electrical changes. Things is it isn't clearly outlined in the license agreement.

[Praz]

but absent is changed clock speed, so I'd think either "alteration" or "improper testing" could cover overclocking without electrical changes. Things is it isn't clearly outlined in the license agreement.

The following from the link you provided pretty much covers overclocking of any sort.

• any Product which has been modified or operated outside of Intel’s publicly available specifications or where the original identification markings (trademark or serial number) has been removed, altered or obliterated from the Product.

[xt0m]

The following from the link you provided pretty much covers overclocking of any sort.

Blherrghh.. hadnt seen, lol - my bad

[saaya]

I recommand you go and read the license agreement of Intel OR AMD processors before you speak.
Francois

so changing any clockspeed above what intel officially supports/specifies voids the warranty?

Core i7 1366 doc:

Although these units are factory-configured for 1333 MHz integrated memory controller frequency, Intel does not support operation beyond 1066 MHz

so wait, running an i7 above ddr3 1066 voids warranty?
intel certifies xmp profiles that go above 2000, but running above 1066 voids the warranty? i dont even know a SINGLE person who runs his i7 at only 1066... even pre-configured systems from dell and hp etc run their memory above 1066...

so intel basically made the official warranty so tight, about every single end user who ever uses an i7 voids its warranty and intel doesnt actually have to rma the cpu if they want to...

i think you guys are overdoing it a bit francois

back on topic, so it basically boils down to this?
1.) too low contact pressure of the socket pins
2.) unbalanced current draw through the socket/cpu package

how can we avoid or fix this:
1.) (Intel) more vcc and respective vss pins
2.) (Intel) reworked cpu package for more balanced current draw
3.) (Intel/Foxconn) stronger socket pins to improve contact pressure
4.) (Intel/Foxconn) check socket lever and make sure the pressure is distributed evenly accorss the pins when the socket is closed.
5.) (Gigabyte etc) thicker vcore power plane to balance current draw better
6.) (Gigabyte etc) more stable current supply from the pwm?
7.) (user) make sure the cpu socket and pads are clean
8.) (user) make sure pins arent bent slightly
9.) (user) make sure the cpu is seated 100% correctly and pins and pads make good contact by wiggling it around before closing the socket
10.) (user) make sure the cpu doesnt get mounted too tight with too much pressure

thats all i can think of...
1/2 wont happen, im sure... its about cost and this is a mainstream socket which means every fraction of a cent counts... and at stock it works fine...

3/4 would be cool if intel would look into this and maybe revise their guidelines to the socket mfcs, or if the socket mfcs would look into this and revise their sockets... this mostly depends on the board makers requests though...

5/6 idk how much this would help, but at least thicker and well laid out power planes cant hurt... about the pwm, they should do this regardless of this problem anyways about time we get rid of this pwm mega phase-mania...

7/8/9/10 i doubt this helps much, but thats all i can think of end users can do to at least make it less likely to happen or not make it even worse...

[Raja@ASUS]

You can omit 3 from the list, the switching frequency affects the transient response and bandwidth, it is not the root cause. To prevent the burnout from happening period on setups, you'd need to lower VID/current, as soon as you do that, you have to lower the CPU operating frequency too.

[massman]

back on topic, so it basically boils down to this?
1.) too low contact pressure of the socket pins
2.) unbalanced current draw through the socket/cpu package
3.) possibly pwm switching related?

This is still speculation really. No one has actually had a look at SF3D's or my board apart from the pictures posted here.

[Raja@ASUS]

Sascha sometimes likes speculation - I think we should call him 'Speco'...lol

[kiwi]

I don't think Intel can verify that particular cpu has been overclocked or not. Or they can?
So it still depends on user's ethics and honesty whether he decides to RMA it or not.

BTW, I remember I had Intel Bad Axe board when Conroe just came out and believe it or the board died in such a way that it saved all my overclocking settings without the ability to erase it (not even bios reset). So If I sent it to RMA they would certainly see that I run cpu out of spec That's interesting

so changing any clockspeed above what intel officially supports/specifies voids the warranty?

Core i7 1366 doc:


so wait, running an i7 above ddr3 1066 voids warranty?
intel certifies xmp profiles that go above 2000, but running above 1066 voids the warranty? i dont even know a SINGLE person who runs his i7 at only 1066... even pre-configured systems from dell and hp etc run their memory above 1066...

so intel basically made the official warranty so tight, about every single end user who ever uses an i7 voids its warranty and intel doesnt actually have to rma the cpu if they want to...

i think you guys are overdoing it a bit francois