P55-UD6 socket burn

[Chr0n1c]

Do all eVGA boards have the LOTES socket or just the classifieds?

This thread made me nervous as hell lol, was testing stability at 4Ghz 1.3vcore/1.25VTT on my i7 860 and FTW in LinX.... for 2 hours. =/

Everything runs ok and I didn't smell anything but this makes me want to go and break the rig down and check it.

[A2C]

Well boys, thank you for lot interesting informations
i hope other brands will do something what can stop this socket burning challenge

[Hondacity]

Do all eVGA boards have the LOTES socket or just the classifieds?

This thread made me nervous as hell lol, was testing stability at 4Ghz 1.3vcore/1.25VTT on my i7 860 and FTW in LinX.... for 2 hours. =/

Everything runs ok and I didn't smell anything but this makes me want to go and break the rig down and check it.

1.3v isn't alot

[Raja@ASUS]

It's the current you need to be conerned about, most CPU's won't pull more than 160w until you have them to 4.3~4.4GHz and then run LinX. VID could be anything from low to high, what matters is the current.

[Chr0n1c]

Ok, I misinterped that vcore would impact the current, well I haven't gone that high on the chip for stability testing with linx. Tried 4.2Ghz but it failed ~2 minutes in. Only have done 2hour tests @ 4GHz.


Soooo it appears I'm A-okay! I'm done using LinX though! lol!

Thanks guys.

[Raja@ASUS]

primary factor is current and frequency, don't be too concerned about the VID. The leakage of processors varies, but by and large to be stable at X frequncy they all draw similar levels of current (at least with ambient temps). Of course, VID becomes an issue if you start pumping in a lot more than the chip actually needs.

[Dahmer]

I think you're right about the amount of VCC pins/pads raju.
As far as I remember from my electromechanics studies a few years ago the circumference of a conductor is based on the current flowing through it. Higher current needs higher circumference and in this case higher current -> more pins/pads

Thinking about that the burning pins/pads make me believe those pins had slightly (or much) higher resistance than the other pins probably because of bad quality/impurities on or in the pins.
They were running a bit out of spec but I still don't like seeing this happen, especially not on boards that have a sign saying "better overclocking"

I'm hoping this is a one in a million issue since I'm thinking about buying the exact same motherboard real soon =o

[lesstutrey]

ya know, this picture reminds me of something i saw, where a guy was trying to fake an RMA, claiming his socket melted, but it was obviously just damaged by something hitting it, this is actually what a melted socket looks like. ahh.. that's some crap.

[saaya]

hmmmm i just looked it up in the intel doc again and... the burned pins seem to be vss, not vcc? do you have more pics of it raja?
actually makes sense cause there is only 1 vss pin for every 2 vcc pins?

[Raja@ASUS]

hmmmm i just looked it up in the intel doc again and... the burned pins seem to be vss, not vcc? do you have more pics of it raja?
actually makes sense cause there is only 1 vss pin for every 2 vcc pins?

Don't read too much into that. The bottom of the processor shows the burn marks on the VCC pads not the VSS. The VSS pins got hot and fell away (from the neighboring heat) while the VCC pins left indents in the VCC pads (with coloration).

[kiwi]

I doubt any i5 can run linx @ 4.4 on air Maybe in winter with low ambients only but temperature would be a major issue

So raju, you want us all burn our cpus and mobos?

[Raja@ASUS]

So raju, you want us all burn our cpus and mobos?

On the contrary, if I wanted you to burn them, do you think I would have posted all of this?

I think the real answer to that question is in the sig of this post;

http://www.xtremesystems.org/forums/...&postcount=118

[Eeky NoX]

Wish it don't disturb you ? In that case I'm sorry and I could delete it.

I've seen so much people only talking about burning instead of try to feel Spi essence
Long stress times for a stabilitaty that don't mind or they don't know how it had been acheieved...8h...12h...24h...I've never understood really.

That's why I love what you said lol and what you're doing here is very respectfull mate

peace

[railmeat]

Burnt out when benching with high vcore ?

[TinTin]

I saw a similar case for MSI GD80 MB, similar location, air-cooling!

[Raja@ASUS]

Seems to be primarily related to an issue at the CPU end.

[saaya]

the gd80 also uses a foxconn socket...

raja and hipro, do you think a redesigned socket with better shaped pins and maybe thicker pins and contact heads and higher pressure onto the pads will make a big diference?
how many extra watts can a lotes or other more robust socket deliver without running into these problems?
so far it hasnt happened on lotes sockets it seems...

thx for all the input and testing raja!

[Raja@ASUS]

It may help a little (contact point density and width), fundamental issue though is the lack of VCC pads - 33% fewer VCC pins than 1366 with almost the same current draw. The LOTES make it slightly less likely, but don't prevent it from happening as shown in the fist pic I posted all the way back at the start of the thread.


I should add that I received pic of a another CPU/socket (FOxconn) that burned out this morning at around 7-10 amps of draw at 5Ghz running a single thread of 32M. So I really don't know what is safe at this point nor do I believe this is soley a socket related issue, could be soemthing fundamentally weak within the CPU die itself, no idea. Put that together with the reported 4GHz OC from a member above and something deeper may be amiss with some CPU's.

[kiwi]

eek, that sounds serious...

[STEvil]

It may help a little (contact point density and width), fundamental issue though is the lack of VCC pads - 33% fewer VCC pins than 1366 with almost the same current draw. The LOTES make it slightly less likely, but don't prevent it from happening as shown in the fist pic I posted all the way back at the start of the thread.


I should add that I received pic of a another CPU/socket (FOxconn) that burned out this morning at around 7-10 amps of draw at 5Ghz running a single thread of 32M. So I really don't know what is safe at this point nor do I believe this is soley a socket related issue, could be soemthing fundamentally weak within the CPU die itself, no idea. Put that together with the reported 4GHz OC from a member above and something deeper may be amiss with some CPU's.

CPU die cant be the problem otherwise the burn would occur in the die itself.

What might be a problem is if each core has its own vcc/vss pads which would mean single core ops are as dangerous or more-so than multicore ops.

[Raja@ASUS]

CPU die cant be the problem otherwise the burn would occur in the die itself.

What might be a problem is if each core has its own vcc/vss pads which would mean single core ops are as dangerous or more-so than multicore ops.

Not if there's a power distribution issue. Doubt Intel would give each core it's very own pads, the whole grid should be in parallel (easy to check for continuity btw). The fact it's always in the same place of the VCC pads points at something deeper.

[cirthix]

I doubt that intel would have done anything different from a dedicated vcc plane, but it is curious that the mark appears at the same spot on different boards...

[saaya]

does anybody have a dead i7? would be interesting to sand the pcb down layer by layer and see how the pads are meshed together...
that might reveal something to explain why it always happens at the same spots...

or, like i said previously, if you look at the socket and the lever and the way the socket moves and closes, it seems the pressure isnt distributed evenly among all pins... the top of the socket seems to get more pressure than the rest, and vcc is only on the left side...

[Raja@ASUS]

or, like i said previously, if you look at the socket and the lever and the way the socket moves and closes, it seems the pressure isnt distributed evenly among all pins... the top of the socket seems to get more pressure than the rest, and vcc is only on the left side...

If that's the case and only the cornermost VCC pads were making contact, it would require an extreme amount of lift at the bottom right corner of the CPU/socket. Given the amount of signal pins in the lower corner, I don't think that it is possible the CPU would even post if the socket were that borked.

[G.Foyle]

The socket has a "seating plane" which enforces fixed distance between pins and package, you can't really push the CPU too hard (i.e. overtightening pot screws). How can the pressure against the pins be too low or too high?