LGA 1156 Foxconn socket problems?

[Humminn55]

Just posted article on Anandtech about potential problems with Foxconn manufactured LGA 1156 sockets and OC'ing.......

http://www.anandtech.com/mb/showdoc.aspx?i=3661

[saaya]

its based on a thread on xs, heh, snail biting its tail

http://www.xtremesystems.org/forums/....php?p=4064451

EDIT: huh? no credit to xs or the people who brought it up? why?

[Humminn55]

its based on a thread on xs, heh, snail biting its tail

http://www.xtremesystems.org/forums/....php?p=4064451

EDIT: huh? no credit to xs or the people who brought it up? why?

First, I don't live on these forums and didn't see/didn't find the XS thread.

Second, AT's article did attribute XS's thread within the article...possibly read before criticizing, which may indeed be a new thing for you, I understand.

[jmke]

EDIT: huh? no credit to xs or the people who brought it up? why?

Maybe Raju... who posted in that thread is the author of the article; he experienced the burn with his setup; why would be have to name XS as source?

[Raja@ASUS]

Maybe Raju... who posted in that thread is the author of the article; he experienced the burn with his setup; why would be have to name XS as source?

Thanks. I was going to use pictures that the guys posted, but decided not to because I did not ask for prior permission. Other than that, respect is given to overclocking peers out on the forums. It's been a journey finding out the cause, and this at the end of it seems to be the common factor. I think has Sascha gotten a bit obsessive compulsive since leaving Foxconn with his posts..

[dctokyo]

Is Foxconn aware of this issue?

so it seems that LOTES/Tyco AMP sockets are safe for now?

[saaya]

First, I don't live on these forums and didn't see/didn't find the XS thread.

i didnt meant that as a poke towards you at all... i posted the link as a heads up...

Second, AT's article did attribute XS's thread within the article...possibly read before criticizing, which may indeed be a new thing for you, I understand.

where? there is no link to xs, no mention of xs, xtreme or systems or even forums at all. and no need to make snappy comments

Maybe Raju... who posted in that thread is the author of the article; he experienced the burn with his setup; why would be have to name XS as source?

he is, but i was under the impression that he only checked and found one burned pad on his cpu after massman and sf3d posted this thread. he then did some more testing and confirmed the socket starts to melt at high clocks and high vcore. even if he found this before and the thread didnt influence him at all im surprised he didnt at least mention xs or sf3d and massman...

i dont care... it just surprised me thats all...

[Raja@ASUS]

he is, but i was under the impression that he only checked and found one burned pad on his cpu after massman and sf3d posted this thread. he then did some more testing and confirmed the socket starts to melt at high clocks and high vcore. even if he found this before and the thread didnt influence him at all im surprised he didnt at least mention xs or sf3d and massman...

i dont care... it just surprised me thats all...

I found the burned pad 2-3 days before the thread went up, I spoke with two colleagues about it at the time and left scratching my head so did not say anything becasue the cause was unclear to me. They posted, I followed up with my pic that's all, as they seemed related. I just think you're always on the lookout for some way to diss AT (but like you maybe I'm assuming too right?).

People respond to you this way because it seems to happen quite often Sascha. You seem to hit the post button without thinking about what it is you're saying these days. I've got nothing against you personally (and I know you have not against me). I'd have given direct thanks if I'd not have had the issues. I've been working on a multiboard article and the issue cropped up with me twice during testing which is why I pursued it - I really had no choice.

[Foamy]

First, I don't live on these forums and didn't see/didn't find the XS thread.

Second, AT's article did attribute XS's thread within the article...possibly read before criticizing, which may indeed be a new thing for you, I understand.

[Chris Crocker modus] Leave saaya alone![mode off]

[Pt1t]

Raja,

I heard that Foxconn modified their LGA 1156 socket in the latest revision but i dont know what manufacturer allready use it !

[Raja@ASUS]

Raja,

I heard that Foxconn modified their LGA 1156 socket in the latest revision but i dont know what manufacturer allready use it !

Thx Thomas. As soon as I get word of that I'll update as required. I spoke with DFI direct this morning and as of yet they have not turned around with Foxconn, so let's see if we can get some concrete info.

[chuck4456]

FOXCONN CPU sockets have always had a degree of sloppy play anyway.

[Utnorris]

I doubt it was nothing more than a QC issue for the socket 1156 sockets made by Foxconn. A lot of manufactures use Foxconn for their 1366 sockets and we haven't seen the same issue. Every manufacture has had issues with QC at one time or another, including EVGA, but it will be interesting to see how the manufactures handle this since it only seems to happen under extreme cooling, which I doubt is covered under the normal manufacture warranty.

[STEvil]

I doubt it was nothing more than a QC issue for the socket 1156 sockets made by Foxconn. A lot of manufactures use Foxconn for their 1366 sockets and we haven't seen the same issue. Every manufacture has had issues with QC at one time or another, including EVGA, but it will be interesting to see how the manufactures handle this since it only seems to happen under extreme cooling, which I doubt is covered under the normal manufacture warranty.

You need to go read the thread. It does not only happen under extreme cooling and is not confined to foxconn. It IS confined to 1156 at the moment.

[saaya]

I found the burned pad 2-3 days before the thread went up, I spoke with two colleagues about it at the time and left scratching my head so did not say anything becasue the cause was unclear to me. They posted, I followed up with my pic that's all, as they seemed related.

ah alright...

I just think you're always on the lookout for some way to diss AT (but like you maybe I'm assuming too right?).

dissing and bashing and trashing is by my definition talking bad about somebody or something for no apparent reason. i hope thats not what you think im doing...

People respond to you this way because it seems to happen quite often Sascha. You seem to hit the post button without thinking about what it is you're saying these days.

lets say without thinking about certain people MIGHT understand it. you and some other people seem to be very sensitive to critique and feel offended easily. :P

I've got nothing against you personally (and I know you have not against me). I'd have given direct thanks if I'd not have had the issues. I've been working on a multiboard article and the issue cropped up with me twice during testing which is why I pursued it - I really had no choice.

cool... so it only happens on highend boards? like which boards? have you guys found any relation between LLC and this problem or the amount of phases or how the phases are arranged, ie duplex=bad?

or could this just be a side effect of the lower end boards simply not being able to produce that much current, hence you cant get the cpus clocked to the same speeds stable and cant load them enough to burn pins and pads?

Raja,

I heard that Foxconn modified their LGA 1156 socket in the latest revision but i dont know what manufacturer allready use it !

mhhh fyi, they said the same when the 1366 socket had problems and it took a while before things actually improved... remember boards not posting, only showing 2 of 3 installed dimms or being very unstable? a lot of that was caused by bent pins or pins making bad contact in the cpu socket...

I doubt it was nothing more than a QC issue for the socket 1156 sockets made by Foxconn. A lot of manufactures use Foxconn for their 1366 sockets and we haven't seen the same issue.

not the same, but similar... so i wasnt really surprised when i heard of this
like i said in the other thread, you cant even really blame them... they make a living off of making parts as cheap as possible while still maintaining the specs... and since they work fine at stock speeds, and even work fine at notably above stock specifications... you cant REALLY blame them... see it from their point of view, the socket works for what it was supposed to be used, pulling max 125W i guess, and in most scenarios 95W. asking them to make a more expensive, sturdier socket that makes them less money is kinda odd... yeah better is better, but 99% of the sockets will never get stressed that much, so... from their point of view this might sound like nonsense...

if they actually DO improve things then they most likely only do it cause they are under pressure from mainboard makers who would otherwise buy the sockets elsewhere, at least sockets for their highend boards...

[Raja@ASUS]

cool... so it only happens on highend boards? like which boards? have you guys found any relation between LLC and this problem or the amount of phases or how the phases are arranged, ie duplex=bad?

or could this just be a side effect of the lower end boards simply not being able to produce that much current, hence you cant get the cpus clocked to the same speeds stable and cant load them enough to burn pins and pads?

I think it's a simple matter of poor contact when inserted into some sockets rather than expensive boards vs cheap or LLC on or off. Of course, you'll have to draw highish levels of current in some instances, so it's just that boards that are OC'd hard have shown the problem to manifest. Too much current via too few pins is simply not that good.

What bothers me most is that I have 5 CPU's here. I checked the base of each of them with a 10xmag eye loupe. 4 of those CPU's have been in around 5 different boards (minimum) of which 4 would be Foxconn socket. Each CPU has almost the same pin indents missing. Further, I've even clamped those CPU's into some of the P55 boards with a stock Intel cooler to see if non standard cooling mechanisms skew contact. Unfortuantely, using the Intel cooler did not make anounce of difference to the non connecting/poor contact pads.

If however I take those same CPU's and insert them into the TYCO or LOTES sockets I get a contact mark on every pad and it's far better aligned than what I get from the Foxconn based boards I have here.

I should re-iterate as some here are still going by a comment I made a few weeks back; the ES 657 I have used a Foxconn socket and that was one of the boards I had a burned pin with. So, ONLY Foxconn socket based boards have been affected by non conctact issues and burnouts that I know of, not LOTES or TYCO AMP.

[saaya]

hmmm yes i read that part on anandtech, and its confusing isnt it? its the ground pins right?
why do the ground pins make worse contact than vcc pins? hows that even possible in a mesh like that?
for that to happen the ground pins would have to be made out of weaker metal or be physically diferent in shape or length than the vcc pins... but doing that would make it 10x more complex to produce the socket than having all pins the same, so im finding it hard to believe thats the case... yet, if thats not it, then why DO the ground pins make worse contact? its a mystery to me

[Raja@ASUS]

I know it does not make sense to me either which is why it took so long to believe this could be the problem. The indents tell the story I guess, and it's hard to refute..

On a sidenote, have you looked at Chew's picture from the other thread? Go and take a look, that CPU was on an air cooler; http://www.xtremesystems.org/forums/...&postcount=154

Compare those with the pads I marked in my pic in the article - anything familiar?

[saaya]

yeah same thing... and remember? i had a feeling theres something up with the vss pads before in that thread, if you look at the burn marks on some of the pics posted, it actually looks as if at least vcc AND vss pins got too hot, if not mostly vss pins actually! whats interesting is the ratio of 2 vcc to 1 vss pads which seems odd to me, but even moreso, the vss pads dont have any small dimples in them. do you know what those are actually for? i figured they might be to prevent the pins from sliding off the pads, instead the pin would slide into one of the dimples and be stuck there? but why do only vss pads have those and not vss?

so it seems that vcc and vss pads are clearly diferent from one another, and maybe pins are specced to be slightly different too?
maybe the intel spec asks for vcc pins to be stronger than vss pins, and while foxconn actually did implement different pin strengths, everybody else went the easy route and used the same pins for vcc and vss... which in this case might have saved them from the problems the foxconn sockets are seing?

meh, i dont know enough about this stuff to speculate, hahaha i better shut up

[BulldogPO]

BTW, is Foxconn only one who manufactures LGA1156 and LGA1366 sockets?

[Raja@ASUS]

yeah same thing... and remember? i had a feeling theres something up with the vss pads before in that thread, if you look at the burn marks on some of the pics posted, it actually looks as if at least vcc AND vss pins got too hot, if not mostly vss pins actually! whats interesting is the ratio of 2 vcc to 1 vss pads which seems odd to me, but even moreso, the vss pads dont have any small dimples in them. do you know what those are actually for? i figured they might be to prevent the pins from sliding off the pads, instead the pin would slide into one of the dimples and be stuck there? but why do only vss pads have those and not vss?

The 'dimples' are caused by the pins touching the pads under pressure. No dimples = no or little contact. Intel does not 'pre-dimple' the CPU's. The Foxconn pin tips are sharp and levae an indent in the pads (that's what the ones that connect well under pressure do). LOTES/Tyco seem to have a bigger surface contact point and leave a slight scuff mark on every pad.

[chew*]

I know it does not make sense to me either which is why it took so long to believe this could be the problem. The indents tell the story I guess, and it's hard to refute..

On a sidenote, have you looked at Chew's picture from the other thread? Go and take a look, that CPU was on an air cooler; http://www.xtremesystems.org/forums/...&postcount=154

Compare those with the pads I marked in my pic in the article - anything familiar?

Heh, when I took that pic I definitely noticed the pads that weren't indenting......But being an intel nub I wasn't sure if it was supposed to or not.

Better to look dumb than say something and remove all doubt I figured it was a clear enough pic and if it was a problem a few smart people would figure it out

[Raja@ASUS]

I should add when I say Intel don't pre-dimple CPU's I mean within the context of pin alignment. New CPU's out of the box may have some pin indents on pads becasue the CPU's are mounted in test jigs by Intel.

[STEvil]

BTW, is Foxconn only one who manufactures LGA1156 and LGA1366 sockets?

no. LOTES does as well.

[chew*]

That was that particular cpu's first time in a socket and I can confirm it had no indents prior. Heatsink mount that was used was the megahalems rev 2 for 1156 etc not makeshift or homemade macgyver routine.

I'm curious raja if diff heatsinks lets say a stock intel leave a different pattern versus say a megahalems......I'm going to go out on a limb here and say no as it appears to be a rather duplicateble pin pattern among the same socket.

Any way I can "fix" the troublesome pins?