Why the pins on CPUs and motherboards?

[Sigfried]

(rferring specifically to cpu and their sockets)

 

Simply put: pins are long tiny pieces of metal = pins are are fragile.

 

Why don't/can't they develop a different more resistant contact mechanism?

[freeagent]

14 minutes ago, Sigfried said:

(rferring specifically to cpu and their sockets)

 

Simply put: pins are long tiny pieces of metal = pins are are fragile.

 

Why don't/can't they develop a different more resistant contact mechanism?

What did you have in mind?

[jones177]

They(PGA) are not fragile.

What will replace them on AMD CPUs is(LGA).

 

I have never had an issue with pins and I have been building PCs for about as long as there as been PCs.

 

I had an eyelash take out a motherboard and CPU on my Q9550 build. 

 

So for me it has been 3 LGA to 0 PGA loses. 

 

[Fasauceome]

4 minutes ago, jones177 said:

They(PGA) are not fragile.

To a point, anyway. I ordered a hard Ryzen 7 CPU and the shipper "creatively" used some cardboard instead of plastic clamshell to package the CPU. Had some negative results...

 

Either way, @Sigfried CPUs are small and complex. Small, complex things are fragile by nature

[HanZie82]

28 minutes ago, Sigfried said:

(rferring specifically to cpu and their sockets)

 

Simply put: pins are long tiny pieces of metal = pins are are fragile.

 

Why don't/can't they develop a different more resistant contact mechanism?

How else are you gonna convey the power, with little resistance and good contact. For many pads in close proximity?

The pins are just there so the socket has some gold plated copper to hold on to.

There are other styles. Like Intel's, where the pins are in the socket rather then on the CPU.

Same thing really just other way around.

And since flexible plastic doesn't conduct, metals it is.

The only other option is directly soldered, and we DO NOT want that for our main desktops.

[YoungBlade]

Modern CPUs are made up of many individual parts. These parts each need access to electrical and data lines from the motherboard. This means that you need a lot of connections.

 

The simple fact is that pins are the best way to achieve this in a compact area while allowing modularity. The only other way to do it is with a ball grid array that requires the CPU to be soldered to the motherboard, which does allow for many independent connection points, but makes modularity impractical.

 

If you have a better suggestion, I'm sure any of the major chip manufacturers or motherboard makers would love to hear about it, so they can patent it and have a competitive advantage by needing to process fewer RMAs.

[Jarsky]

First thing that comes to mind is it needs to be dense for modern CPU's. Looking at a modern CPU with 1300-1700 pins...you also need every single pin to make reliable contact. 

 

In a modern AMD socket using PGA, the pins slide into the socket which has metal contacts on each side of the pin. But its not until you lower the retention arm that the top of the socket moves across,clamping the pins into the metal contacts. 

 

In a modern Intel using BGA the "pins" in the socket are springed so they push up against the pads on the bottom of the CPU. Hence why the retaining bracket should be clamped closed to ensure each springed pin is pushing up against the pad. 

 

I personally have only ever damaged 1 CPU, and 0 Motherboards via socket damage in 30 years of PC building. As long as you take care handling the CPU, put cover on sockets when you have CPUs out for extended periods, and put PGA CPU's into clamshells when not in use. 

 

Be interesting to see what ideas you have though for a new design. 

[Guest]

The better option is BGA, but the draw back is non swap cpus. 

 

Perhaps instead of pins to pads, male female connectors. Which IMO wouldn't be any better or reliable.

[mariushm]

The most common processors types are PGA (pin grid array) and LGA (land grid array) 

PGA = there's pins soldered on the bottom of the CPU and you insert the CPU in a socket and you get a very good grip on every pin

LGA = there's tiny pads (squares, hexagons, rombs, whatever) or gold plated material on the bottom of the cpu and the socket has flexible contacts that press on the pads to make contact. When you place the CPU on the socket, the CPU's weight bends down the contacts a bit. 

 

There's pros and cons for each method. 

Pins are safer, there's less risk of damaging CPU compared to damaging motherboard socket. CPUs are shipped in plastic shells or boxes, so no risk of damaging pins during transport.  Only time you risk damaging pins is when you take it out and insert in socket - you may drop it and bend a pin, which can be straightened out. In very rare cases, you may break a pin and even then you can solder another one with the right equipment

 

Intel moved to LGA because they didn't want to deal with processors with bent or broken pins, they basically moved the losses from their end to the motherboard manufacturer's end. There's much higher risk to have a screwdriver or the corner of the CPU drop on the LGA socket and permanently damage one or more contacts in the socket, and your motherboard is gone... replacing a socket can cost up to $100 or more at authorized service places. 

 

On one hand there's the benefit of PGA that the pins are grabbed from two sides by the socket, so there's a lot of surface area between socket and pins, so you don't need a lot of pins to transfer power to the CPU... in contrast the pads often have very small tips, small area where it makes contact with the cpu, so you need more pins to carry the same amount of power. 

There's also higher risk of having some dirt or... anything, like for example a strand of hair, be between the LGA socket and the CPU and cause a contact to not touch a pad on the cpu. 

 

However there's some downsides. Pins are long and because of that, there's longer distances between the circuit board and the actual cpu, and the pins themselves can have extra inductance or act like tiny antennas picking up noise from the other pins. This can affect the CPU reaching higher frequencies, or it can be an issue where you have lots of pins close together.

With LGA, the spring like contacts are shorter, so  overall there's less length of wire between the motherboard and the cpu die.

 

LGA also allows you to pack more pins in an amount of space, that's why AMD went with LGA on Threadripper and Epyc, that and what I said in the paragraph above, shorter distance, better signals etc .. it's an issue when you have 4-8 channels and 128 pci-e lanes and lots of stuff you need pins for. 

 

Makes sense for AMD to move to LGA for better signals, when they're gonna do pci-e 5.0 (probably they're gonna assume socket will still be here on pci-e 6.0)  , ddr5, and other high bandwidth stuff. 

 

 

[jones177]

17 minutes ago, Fasauceome said:

To a point, anyway. I ordered a hard Ryzen 7 CPU and the shipper "creatively" used some cardboard instead of plastic clamshell to package the CPU. Had some negative results...

 

Either way, @Sigfried CPUs are small and complex. Small, complex things are fragile by nature

I look for the fraise "retail box" in the descriptions. 

They are also better for resale.

[Sigfried]

4 hours ago, HanZie82 said:

How else are you gonna convey the power, with little resistance and good contact. For many pads in close proximity?

The pins are just there so the socket has some gold plated copper to hold on to.

There are other styles. Like Intel's, where the pins are in the socket rather then on the CPU.

Same thing really just other way around.

And since flexible plastic doesn't conduct, metals it is.

The only other option is directly soldered, and we DO NOT want that for our main desktops.

Sorry, I f*cked up the phrasing, with "resistence" I actually meant to say "endurance".

[Sigfried]

3 hours ago, mariushm said:

The most common processors types...

Great answer, thanks for the explaination, that being said, isn't there a better solution? I'm not engeneer but how about pads against pads? or pads against short/roud/conic pins?

At least removing the lenght problem? Is it such a dumb Idea?

 

[Sigfried]

4 hours ago, Guest 5150 said:

The better option is BGA, but the draw back is non swap cpus. 

 

Perhaps instead of pins to pads, male female connectors. Which IMO wouldn't be any better or reliable.

oh nonono that is even worse, didn't even consider that one

[Mister Woof]

9 minutes ago, Sigfried said:

Great answer, thanks for the explaination, that being said, isn't there a better solution? I'm not engeneer but how about pads against pads? or pads against short/roud/conic pins?

At least removing the lenght problem? Is it such a dumb Idea?

 

Apple would say soldered, non user-serviceable CPUs are "better" options

[Sigfried]

4 hours ago, Jarsky said:

Be interesting to see what ideas you have though for a new design. 

I just looked up a BGA chip.

I know what the tiny metal drops are in this picture, definitely not the mechanism we want  BUT , is it that crazy to do pins in a similar shape? 

Maybe conic but not pointy? 

anything that would make the pin part of the equation shorter, so that is simply less fragile.

I have no idea how CPUs are built and I am learning now some rationale but... I don't know it seems crazy to me, even though I'm sure they have really good reasons that I am probably about to know as well.

[Kilrah]

The problem is you need something that is flexible and can absorb the tolerance between the 2 parts, if everything was rigid it'd be impossible to ensure all pins make contact.

On both LGA and PGA the flexible bit is in the socket, on LGA it's exposed which is what makes it fragile, on PGA it's protected but it means there need to be long enough pins to get into the protected socket and provide enough area for the contacts to engage with, and now these get fragile, although less...

[mariushm]

The BGA chips have balls of solder ... you put the chip over pads on the circuit board and when the solder is heated there's a chemical reaction between the solder and the pads and the metals bind together so you get a very reliable and strong connection...

With removable processors, you rely either on friction (pga sockets, friction between the metal bits surrounding the pin) or pressure (lga sockets, chip pressing down on the elastic contacts) .. ram slots also rely on friction.

 

You could make very tiny and thin pins, using other stronger materials .. and you could maybe have holes in the cpu  bottom where pins would go in, and have more or less 1-2 mm of pins going inside the cpu base. Then, have the cpu tightened from the back of the motherboard using 4 thin screws.. you'd get very low height... but again, you risk the pins on the motherboard getting damaged, bent etc and those pins would be hard to solder onto the motherboard. 

[Mister Woof]

Generally speaking if it isn't broken, don't fix it. 

 

Both Intel and AMD know a lot more about this than we do, given the enormous amount of RMA data they have, and it's more profitable to keep doing whatever they're doing because the failures aren't common enough to be a problem.