I work on AM3 boards on a regular basis, and I have never, ever bent a pin. I use this little trick called patience, its awesome. Don't force the chip into the socket, when its positioned correctly it will fall in, when removing a heatsink gently twist the loosen, though I find that the ceramic based thermal compound largely remedy's the sticking you get from AMD's standard heatpad, or from polysynthetic silver.
I'm not saying the Intel design is not a little more user friendly, but when the manufacturing tolerances are off just a touch, you have what you have here.
Pins are old and busted. LGA is easily the most intelligent way to interface with a CPU. Sun and IBM have been using LGA for their sockets for almost a decade.
Oh, its not a personal attack, I'm just saying, I've had to work on a pin socket hundreds of times, I'm not kidding, I've never, ever bent one, not once. Am I careful, kinda slow? Sure. Is it the worlds perfect design? Maybe not, but It has some pros, like, everything making contact between the chip and socket.
Pins are old and busted, clearly a slot loaded CPU is the way to go. Behold, the past that is the future!
Oh, and Cliff, why do you even bother anymore? Even when you do make a point they just dismiss you as a fanboy anyway. Not that I'm saying you have a point here. All sarcasm aside, I don't really care how you mount a CPU as long as it works. I am in the camp of never having managed to bend a CPU pin though.
Pins are old and busted, clearly a slot loaded CPU is the way to go. Behold, the past that is the future!
Oh, and Cliff, why do you even bother anymore? Even when you do make a point they just dismiss you as a fanboy anyway. Not that I'm saying you have a point here. All sarcasm aside, I don't really care how you mount a CPU as long as it works. I am in the camp of never having managed to bend a CPU pin though.
I am unfortunately not in that camp. I managed to bend many pins on my northwood pentium 4. Through VERY careful realignment of the pins, the CPU still works, but it scared the shit out of me when I did bend them.
Have you ever dropped an LGA CPU on the socket sideways? I did! That smashed 3-4 springs in the socket. It took so much time and patience under a magnifier to bring them back to position. My overclock dropped by 200MHz after this event. After seeing this story, I wonder if those bent springs are still not making contact. But I have no intention to redo the operation again.
Could you imagine the width of a slot CPU if they mad them now, not to mention the cost. A shame it met its limitations in design, because it was as close to mistake proof as a design could be.
My old Slot A Athlon was the machine that made me a PC hardware enthusiast.
I think for LGA1156 is simply early after launch. It seems that no one remembers issues with LGA775... but there were all kinds of issues with all suppliers.. All the suppliers of sockets are going to have some growing pains with quality.. Anandtech is getting huge mileage out of one system that burned at 5.13GHz. This particular instance is totally blown out of proportion.. Intel spends 2 years doing extensive reliability testing with all tier 1 suppliers.
Every socket generation brings a significant reduction in pitch which reduces the current path cross section and subsequently reduces the current carrying capacity of the metal contacts in the socket. There are a of other technical inaccuracies in the articles (including Anandtech's) but I think that the real work needs to be done by Intel and Foxconn. Ultimately they are responsible to find the solution and determine if there is really a problem.. This is always part of the emotional issues of early adopters.
For those who understand thermodynamics... The CPU substrate is essentially an adiabatic boundary condition unless delta T (CPU/Socket) is huge. In other words, the cooler doesn't do a lot for the board or socket.
Ya I think too much is being made of these results seen on "xXxtreme overclocking and benchmarking" testbeds. With some of these non-standard pots being used, it might simply be a mounting pressure issue. It also sounds like part of the problem is the change in the LGA1156 latch mechanism. The LGA1366 required extreme pressure to close the latch, possibly ensuring all pins are making solid contact. LGA1156 doesn't function the same way and puts much less pressure on the CPU in the socket.
Anyways, I also prefer the pin/zif sockets of old, especially with the amount of pressure required with LGA1366. It was obviously in Intel's best interest to move to a LGA design, as that shifts much of the expense and potential for damage to the socket and board rather than the CPU. But I'd agree it seems much easier to damage those socket pins than the pins on a CPU, even though I haven't damaged either, I've seen far more users report socket damage on various support forums.
I don't think anyone is exaggerating the severity of this issue. One manufacturer has qc issues at unqualified clockspeeds. Ho noez! This is clearly circumstantial. That said, it is still news and it is still a potential problem that extreme OCers need to be aware of. Saying nothing about it because it has a narrow scope is patently absurd.
I don't think anyone is exaggerating the severity of this issue. One manufacturer has qc issues at unqualified clockspeeds. Ho noez! This is clearly circumstantial. That said, it is still news and it is still a potential problem that extreme OCers need to be aware of. Saying nothing about it because it has a narrow scope is patently absurd.
And therein lies the problem, if the product works at specified clockspeeds and voltages that fall within design tolerances there are no QC issues. You might as well publish a few warnings about going outside during a thunderstorm wrapped in tinfoil or swimming in the ocean covered in chum. Ho noez, you might get struck
by lightning or eaten by a sharkz!
Bummer. Both 1156 boards Icrontic has feature Foxconn sockets. Not that we're going to use an extreme phase change cooler or LN2, but nonetheless a bit disappointing. Overclocking will proceed with caution.
And therein lies the problem, if the product works at specified clockspeeds and voltages that fall within design tolerances there are no QC issues. You might as well publish a few warnings about going outside during a thunderstorm wrapped in tinfoil or swimming in the ocean covered in chum. Ho noez, you might get struck
by lightning or eaten by a sharkz!
It is a QC issue, chizow. According to Intel specifications, all pins need to be contacting the processor. Two socket manufacturers comply, one doesn't. You can't characterize it any other way.
It is a QC issue, chizow. According to Intel specifications, all pins need to be contacting the processor. Two socket manufacturers comply, one doesn't. You can't characterize it any other way.
Intel specifications may state all lands must be connected to their respective power planes, but that's clearly for redundancy purposes only as:
1) there are no reported problems with the Foxconn sockets running at spec or even with typical, non-extreme overclocking.
2) even the sockets from Lotes/Tyco don't meet spec as seen in the picture below, not all of their pads in the Vcc/Vss area are properly contacting the socket.
So what you have is two out of spec sockets, both of which handle normal operating frequencies and conditions without issue and passing any QA/QC requirements and another that handles extreme conditions and frequencies that 99.9999% of users will never subject their platforms to.
*Queue the Volkswagon future family to zap in and let us know if this'll be news in 10 years.*
That said, it is still news and it is still a potential problem that extreme OCers need to be aware of. Saying nothing about it because it has a narrow scope is patently absurd.
And of course...all of this is moot as the only way to replicate this problem is to run it so far out of spec that it clearly exceeds design thresholds. You claim there's a QC issue on one socket when there clearly isn't. I guess we need to burn up a couple Lotes/Tyco sockets at 8GHz and 2.55 Vcc instead of 6GHz at 2.1 Vcc to determine the same thing? That's nonsense.
In other news, if you drive 150mph and crash your car into a phone pole, you might die.
And I haven't disagreed about that point, my point was that the severity of the issue is clearly exaggerated, as it simply isn't exposed under normal operating conditions. The fact you keep insisting its a QC issue just illustrates how exaggerated the problem is, as once again, 99.9999% of users will never know there is an issue at all.
And I don't have a problem with the news bit as novelty or human interest piece, like watching a train wreck, but to characterize this as a QC issue is patently absurd.
It's causing spectacular failure in some extreme overclocking conditions because some of the vcc/gnd pads don't have connections. It could also be cause instability/crashes/no post / other problems in regular usage if some of the signal pads don't have good contact.
Additionally, intel specifies that those pads must all have contact, so they assume that a socket can deliver a certain amount of power divided over all the pads. Currently, only extremely overclocking a chip exceeds the amount of power that can be delivered by a reduced number of pads. If/when intel releases chips with higher power requirements (more cores, or just higher clock), they could also hit this limit, causing board/cpu/more damage.
EDIT: It may very well also be causing premature death due to increased heat.
Well guys, it really is an issue. in fact the way the foxconn retention bracket works is that it actually pushes the cpu half a mm down when you clamp it. My asus maximus iii formula just lost its processor...and i didnt overclock!!!! i am waiting for intel if they will replace it. ASUS already told me to f myself basically. There goes 300 on just a mobo not to mention my computer company ever using them again.
Thrax
🐌Austin, TX Icrontian
Comments
Many people have complained about bent pins; the heatsink gets glued to the CPU and it is SO DAMNED EASY to bend the pins.
I wish so badly that AMD would go to a pinless PCB like Intel did. I cannot stand the archaic and easily damaged pin system.
The pins are made of gold, which is why they are so easy to bend.
I'm not saying the Intel design is not a little more user friendly, but when the manufacturing tolerances are off just a touch, you have what you have here.
WOW CLIFF I'VE NEVER HEARD OF PATIENCE, I'LL HAVE TO GIVE IT A TRY, THANKS!
It's gonna be hard though, since I'm a totally impatient buffoon! Wish me luck!
Heaven forbid someone has a better design from an engineering and practicality standpoint than AMD
Oh, and Cliff, why do you even bother anymore? Even when you do make a point they just dismiss you as a fanboy anyway. Not that I'm saying you have a point here. All sarcasm aside, I don't really care how you mount a CPU as long as it works. I am in the camp of never having managed to bend a CPU pin though.
I am unfortunately not in that camp. I managed to bend many pins on my northwood pentium 4. Through VERY careful realignment of the pins, the CPU still works, but it scared the shit out of me when I did bend them.
Patience young grasshoppers........ patience.
My old Slot A Athlon was the machine that made me a PC hardware enthusiast.
Every socket generation brings a significant reduction in pitch which reduces the current path cross section and subsequently reduces the current carrying capacity of the metal contacts in the socket. There are a of other technical inaccuracies in the articles (including Anandtech's) but I think that the real work needs to be done by Intel and Foxconn. Ultimately they are responsible to find the solution and determine if there is really a problem.. This is always part of the emotional issues of early adopters.
For those who understand thermodynamics... The CPU substrate is essentially an adiabatic boundary condition unless delta T (CPU/Socket) is huge. In other words, the cooler doesn't do a lot for the board or socket.
Anyways, I also prefer the pin/zif sockets of old, especially with the amount of pressure required with LGA1366. It was obviously in Intel's best interest to move to a LGA design, as that shifts much of the expense and potential for damage to the socket and board rather than the CPU. But I'd agree it seems much easier to damage those socket pins than the pins on a CPU, even though I haven't damaged either, I've seen far more users report socket damage on various support forums.
by lightning or eaten by a sharkz!
It is a QC issue, chizow. According to Intel specifications, all pins need to be contacting the processor. Two socket manufacturers comply, one doesn't. You can't characterize it any other way.
Intel specifications may state all lands must be connected to their respective power planes, but that's clearly for redundancy purposes only as:
1) there are no reported problems with the Foxconn sockets running at spec or even with typical, non-extreme overclocking.
2) even the sockets from Lotes/Tyco don't meet spec as seen in the picture below, not all of their pads in the Vcc/Vss area are properly contacting the socket.
So what you have is two out of spec sockets, both of which handle normal operating frequencies and conditions without issue and passing any QA/QC requirements and another that handles extreme conditions and frequencies that 99.9999% of users will never subject their platforms to.
*Queue the Volkswagon future family to zap in and let us know if this'll be news in 10 years.*
In other news, if you drive 150mph and crash your car into a phone pole, you might die.
And I don't have a problem with the news bit as novelty or human interest piece, like watching a train wreck, but to characterize this as a QC issue is patently absurd.
Additionally, intel specifies that those pads must all have contact, so they assume that a socket can deliver a certain amount of power divided over all the pads. Currently, only extremely overclocking a chip exceeds the amount of power that can be delivered by a reduced number of pads. If/when intel releases chips with higher power requirements (more cores, or just higher clock), they could also hit this limit, causing board/cpu/more damage.
EDIT: It may very well also be causing premature death due to increased heat.