cpu sockets, necessary evil or rogue business tactics?

[CPUarchitect]

LGA1156 and LGA1155 have a completely different pinout:

With LGA1155 they moved the base clock generator from the motherboard onto the CPU die. It also adds support for DMI 2.0 with twice the link speed (affecting 6Gbps SATA and USB 3.0). It's also supposed to simplify routing on the motherboard, and improve signal integrity.

 

[gorydetails]

nnot important, let's see some bakwards compatibility instead.

 

[Magic Carpet]

Appreciate everybody's response. Thanks guys, quite an interesting read :thumbsup:

With LGA1155 they moved the base clock generator from the motherboard onto the CPU die. It also adds support for DMI 2.0 with twice the link speed (affecting 6Gbps SATA and USB 3.0). It's also supposed to simplify routing on the motherboard, and improve signal integrity.

From an engineer's point of view, at least... you reckon, the move was justified? Or is it just a "competing team" trying to establish their authority?

 

[imagoon]

I used to think about this but eventually I realized that by the time I wanted to upgrade my CPU, it was time to get some new feature on a system board anyway. PCI-e 3.0 / USB 3.0 being examples now.

 

[hans007]

its a necessary evil. you need to make a socket tailored to the new cpus tailored chipset .


the reason there are sockets is because many companies make boards for different uses or the board would just come with the CPU like an Xbox or something

it would only be a business tactic to force upgrades if people ever upgraded their CPU which 98% of people don't. enthusiast PC builders are not a factor probably ever

 

[dealcorn]

Now, seriously... what exactly was the need to go down the 1156>1155 route if not milking us, the end-customers?

Have you noticed that in every Formula One race, as it crosses the finish line, the rear left hand tire of the winning car is invariable rotating. It is true and verifiable, but if you bother to check, you have no clue what Formula One is about. You are lost in the trees when you need to hear the cadence of the forest.

The cadence comes from tick tock with its unyielding cycle. First you develop a better process technology so you can build something you know with unproven, better, cheaper transistors (tick). Then you build a new, better architecture that uses the now proven, better, cheaper transistors to deliver better performance (tock).

Intel could make it a tock design specification that the interface does not change but that constraint usually screws up performance. Right or wrong, Intel tries to deliver leading edge performance. Interface stability does make it easier for existing customers to pour additional funds into an obsolete platform as it falls further behind. As a business strategy it works fine but is not Intel's chosen route.

 

[BallaTheFeared]

I think he meant to say the P4 was the first CPU family that began using the LGA775.

My point was, sorry for not making it clear enough...

When Intel stuck to the same design and pumped out new sockets with refined chips AMD smoked them.

So instead, Intel has gone into a frenzy of new design + new socket. Which is fine by me, Pent 4 wasn't very good, and it spanning three sockets wasn't helping anyone here.

AMD, well now, AMD has been doing what Intel did from 2000 to 2006, refining the same design over and over, pumping out several sockets for the same "core" design...

Instead of Bulldozer being their "Conroe" however, it turned out to actually be their "Pent 4".

Back to the drawing board!

 

[jhu]

Don't think it really matters. Intel sells the cpus, not the chipsets.

Oh but they do make and sell chipsets.

 

[jhu]

AMD, well now, AMD has been doing what Intel did from 2000 to 2006, refining the same design over and over, pumping out several sockets for the same "core" design...

Instead of Bulldozer being their "Conroe" however, it turned out to actually be their "Pent 4".

Back to the drawing board!

Actually, Intel had been refining the same design from 1995 to 2011 until Sandy Bridge came out. Pentium Po -> Pentium II -> Pentium !!! -> Pentium M -> Core -> Core 2 -> Nehalem

 

[86waterpumper]

by the time I wanted to upgrade my CPU, it was time to get some new feature on a system board anyway.

Exactly there is always new tech on the boards that is nice to get even if it's not required. Think about even recently, you had usb 3.0, uefi, sata 6gb etc. There are probably amd boards out there without any of those features that would run a bulldozer cpu but who would even want that?
A system with a uefi bios feels alot different than a older system, and the system also boots up faster because of it. Most people that want cutting edge cpus are not going to be very happy with it stuck in a old system. Another thing to consider is that if you keep a system long enough it can be just as expensive to upgrade the cpu than a cpu and motherboard both. A good example of this is the prices some have gotten for the older socket 939 fx and dual core cpus.

 

[aigomorla]

I'm not quite sure as to why Intel is going from LGA1156 -> LGA1155 -> LGA1150 but it could be that there were unused pins in LGA1156 that was deemed unnecessary.

change in arch... this made the pin locations completely different from 1156 vs 1155.

This is why u guys got a new pin layout.
They couldnt hold onto the same layout and the same routing on Sandy-B.

Now if ur asking me why didnt they just keep 1366 i have no clue.... 2011 vs 1366 makes sense... but 1156->1155 and not 1366 im lost...

Intel makes chipsets... intel boards however i think Foxcon dominates the board maker business. Most company use foxcon, most cpu sockets are made by lotte.

 

[CPUarchitect]

From an engineer's point of view, at least... you reckon, the move was justified? Or is it just a "competing team" trying to establish their authority?

I certainly don't think Intel deliberately changes sockets to force people to buy more hardware. The laws of supply and demand make it clear you can't force people to spend money on something they don't want.

But the thing is, there will always be people complaining. It all depends on whether an individual appreciates the new features more than he dislikes the broken compatibility. Note also that even if Intel had delayed the new features, the next generation would still require a new socket. So to some the changes can't happen soon enough while to others it's always too early.

Personally I think requiring a new socket for every 'tock' and allowing an upgrade to the next 'tick' using the same socket is a very reasonable compromise. Obviously it means that if you start out with a 'tick' your next upgrade will require a new motherboard, but there's no pleasing everyone.

 

[Golgatha]

Don't think it really matters. Intel sells the cpus, not the chipsets.

No, they sell both Intel CPUs and chipsets. If a CPU socket gets changed, Intel makes money on chipset sales as well. Back when other companies could make chipsets for Intel motherboards, say the S775 days, you saw Intel trying to keep with the same socket (e.g. P4, Pentium D, Core 2). Now that Intel has all the patents and a monopoly on their chipsets, each new CPU takes a different socket. I'm to the point where I'm having a hard time keeping up with it all anymore and largely don't try too hard to keep up with it all anymore since the performance increases are marginal at best lately.

 

[Belegost]

So with bclk being moved on-die the older 1156 chipsets wouldn't have been able to support SB, so even if they had been able to maintain physical pin compatibility with 1156, if someone had stuck one in an old board, it would have failed, perhaps causing damage to the chip or motherboard.

So due to architecture changes a new motherboard was necessary, at which point why not make the socket different to prevent any accidents?

 

[Magic Carpet]

Thanks for your posts guys.

Personally I think requiring a new socket for every 'tock' and allowing an upgrade to the next 'tick' using the same socket is a very reasonable compromise. Obviously it means that if you start out with a 'tick' your next upgrade will require a new motherboard, but there's no pleasing everyone.

Personally, I don't have a problem paying X amount of money for new tech. But I do have a problem when the choice is limited. Often your next motherboard isn't like your previous... you get used to features, port location, slots, etc. Then, because of a cpu upgrade, you have to go thru this non-sense again. I'd rather build my own with the features I need. I want to be in control but I can't. Way too many compromises to make these days.

If it was a Lego-like builder, I wouldn't have a problem but it isn't.

 

[Ferzerp]

Locking in to a single socket means you're locking yourself in to that chipset feature set. I'd rather intel keep their agility than get stuck with rehash after rehash due to chipset limitations.

 

[386DX]

I don't care when they change socket because certain changes in CPU architect requires it. Besides changing motherboard also brings new up to date features.

What does bother me is why do they have to change the frigging mounting holes for the heat sink. you end up having to buy a new heat sink because finding an adapter is to hard and often costs more then buying a new heat sink.

 

[janas19]

No, they sell both Intel CPUs and chipsets. If a CPU socket gets changed, Intel makes money on chipset sales as well. Back when other companies could make chipsets for Intel motherboards, say the S775 days, you saw Intel trying to keep with the same socket (e.g. P4, Pentium D, Core 2). Now that Intel has all the patents and a monopoly on their chipsets, each new CPU takes a different socket. I'm to the point where I'm having a hard time keeping up with it all anymore and largely don't try too hard to keep up with it all anymore since the performance increases are marginal at best lately.

How do you explain Intel going from more than 14 consumer chipsets on LGA 775 to less than 8 on LGA 1155 then? If their strategy is to make money off changing chipsets, seems they would want to introduce more, not less.

 

[dma0991]

What does bother me is why do they have to change the frigging mounting holes for the heat sink. you end up having to buy a new heat sink because finding an adapter is to hard and often costs more then buying a new heat sink.

It's a conspiracy, heatsink manufacturers paid Intel to do that so that they would profit from users who is forced to change their heatsink every generation. :sneaky:

Unless of course you've bought from Noctua and you've claimed your free LGA2011 mounting bracket.

 

[Lonyo]

The OP is based on a false premise.
Changing sockets is something absolutely meaningless.

Intel could stick with the same socket for 10 years and it wouldn't change anything.
People have brought up the example of 775.
Try taking an original on-release 775 motherboard and running a last-gen 775 processor on it.
You can't.

Sockets are a byproduct of the other major elements of making a CPU work, the chipset and the CPU.
It doesnt matter what socket Intel are using, what matters is whether the chipset supports the CPU.
There's no point in making Socket 1156 IvyBridge CPUs, because the 1156 IvyBridge CPUs wouldn't work on the chipsets.

Sometimes a chipset will support different sockets, but usually the new socket comes with a new chipset which is required to support the new CPU.
If anything, changing sockets along with designs and chipsets keeps things cleaner for the end user, because you know what your motherboard will support. If you have a Socket 775 motherboard, you have to find out which Socket 775 CPUs will work on it.

If you have a Socket 1155 motherboard, you still have to do this, but to a lesser degree.

For example, IvyBridge will run on *most* socket 1155 motherboards, with a BIOS update, but SOME chipsets just cannot support it.
That's a new socket 1155 CPU on an existing socket, but the socket still doesn't support it because of the chipset.

No one should give the slightest crap about changing sockets, because they don't matter to the end user from a support point of view.

 

[Th3Loonatic]

The diagram on top. I'm pretty sure its not supposed to be released for public usage. If I'm not mistaken.

 

[scgt1]

Got a GIG board that I'm working on. The nipple on top of SA_DQS#[1] is missing. The data sheet states it's pin AP3 and DDR3 I/O anyone know what that may result in if the socket doesn't contact that pin properly with the nipple missing?

Would have been nice for a vss or vcc pin to be damaged.

Location on the diagram at the top of Page two is AP/3 far right second light grayish/blue row third in from the right.

 

[dave_the_nerd]

Moving from 775 to something else was necessary when they moved the memory controller on-die.

1366 supported triple-channel RAM, 1156 didn't.

No idea why 1156->1155 was necessary, or 1155->1150. But my guess would be it's a marketing reason as much as anything else. If you have a bunch of people sticking brand new SB i7s in their old 1156 motherboards with SATA2 and USB2, and maybe some slower interconnect tech or no PCI-E 2.0 support, or whatever, then the new CPUs don't have as much of a chance to shine.

 

[NTMBK]

Moving from 775 to something else was necessary when they moved the memory controller on-die.

1366 supported triple-channel RAM, 1156 didn't.

No idea why 1156->1155 was necessary, or 1155->1150. But my guess would be it's a marketing reason as much as anything else. If you have a bunch of people sticking brand new SB i7s in their old 1156 motherboards with SATA2 and USB2, and maybe some slower interconnect tech or no PCI-E 2.0 support, or whatever, then the new CPUs don't have as much of a chance to shine.

Haswell moved the VRMs onto the CPU package, making a new socket necessary.

 

[2is]

1.5 year old thread is old