Rant in here if you hate the socket game with Intel

[Ctek]

rant

I sure hate the socket/platform game with Intel...

I gotta give it to AMD, they have been running the AM2/AM2+/AM3 intercompatible socket for a long time while Intel has been changing sockets on us at least 3 times without any interoperability. Plus thay have segmented the mainstream and high-end with different sockets.

Is it a money grab? I belive that early reports indicate that Ivy Bridge could be compatible with earlier chipsets (6 series, even 5) but it is simply not on the same socket. Makes me think back at the Pentium D/Core era where you had several LGA775 chipsets that had some interroperability.

The platform where I think got kicked in the face the most by those socket changes is the 1366. It's actually starting to be considered a dead platform, despite the premium you had to pay for it.

I for myself think that the 1156/1155/1366 and soon to be 2011 sockets are a big marketing scheme to grab the most money possible, forcing a new platform as soon as possible.

We all know Intel whoops AMD in most desktop segments in terms of raw performance (not always price/performance though) so if you are enough of a power user, you will buy Intel most of the time. I just built a new Sandy bridge rig for work and I am satisfied with the performance I get but I get a mixed feeling in my mouth for the next year or so if there will be any worthwhile upgrades for this platform or it will end up like 1156/1366, dead.

Anyways it's a rant thread about me not 100% agreeing with Intel's strategy with the current Platform/socket chess game.

/rant

 

[novadaemon]

Gigabyte has already released Ivy bridge compatible bioses for their Z68 products. IB engineering samples are also running on z68 boards so that's just FUD you are spreading.

Last gen we has 1366 and 1156
this gen we have 2011 and 1155.

With AMD their sockets have only been compatible between one product cycle and the next.

AM2+ works in AM2+ and some AM2.
AM3 works in some AM2+ and AM3, but not the other way around.
AM3+ will work in some AM3 boards, but no AM2+ boards that I know of.

You only ever got 2 product cycles out of a single board if you were lucky.

The only fault I find on Intel's end is having 3 different chipsets on the launch of one socket. just a single Z68 would have been good enough for everyone, then the MB manufacturers could pick and choose features from there.

This is not a chess game as you suggest.
1366 was triple channel and it brought added cost with it.
2011 is quad channel and that also brings extra cost.

You can still be a computer enthusiast on 1155. Most users don't need that kind of memory bandwidth anyway. The entry level Quad 2011 CPU is rumored to have a locked multi. If that's the case then the 2600k will still be the superior quad.

 

[gwarren007]

I belive that most of what Intel is moving to benefits us and that its not a money grab. One of the socket changeovers is because of the memory controller being put on the cpu die, and not on the mobo. I am glad they did that so you wouldn't put in the wrong cpu without setting the required jumper (s) and fring the cpu and mobo. I see the next infusion from intel requiring all (2011) mobo makers have the capacity to support 8 core and 16 core cpus in the VRM's. That isn't so required now.

My .02

 

[dnottis]

What socket game? I love buying a new motherboard every generation... it keeps the dust bunnies at bay

I wouldn't call it a cash grab as they do make improvements add features with each new board... what I really hate is that the change the mounting holes around the CPU so I have to keep changing my block / block hardware for my water cooling. If Intel and AMD would create a common standard for block and cooler mounting I wouldn't have to play the games with my cooling setup every upgrade. I managed to keep my socket 775 block setup onto the 1156 platform with the EVGA board that had 775 mounting holes. Even managed to carry the 775 block setup over to the 1155 platform as Asrock again gave us that option.

I finally upgraded to a block with native 1155/1156/775. etc support... but I'm sure the next generation will negate that. Now that really tweaks me.

 

[Ctek]

Gigabyte has already released Ivy bridge compatible bioses for their Z68 products. IB engineering samples are also running on z68 boards so that's just FUD you are spreading.

Last gen we has 1366 and 1156
this gen we have 2011 and 1155.

With AMD their sockets have only been compatible between one product cycle and the next.

AM2+ works in AM2+ and some AM2.
AM3 works in some AM2+ and AM3, but not the other way around.
AM3+ will work in some AM3 boards, but no AM2+ boards that I know of.

You only ever got 2 product cycles out of a single board if you were lucky.

The only fault I find on Intel's end is having 3 different chipsets on the launch of one socket. just a single Z68 would have been good enough for everyone, then the MB manufacturers could pick and choose features from there.

This is not a chess game as you suggest.
1366 was triple channel and it brought added cost with it.
2011 is quad channel and that also brings extra cost.

You can still be a computer enthusiast on 1155. Most users don't need that kind of memory bandwidth anyway. The entry level Quad 2011 CPU is rumored to have a locked multi. If that's the case then the 2600k will still be the superior quad.

While I agree with some of your statements, I just don't agree with Intel platform disparity and short socket life. LGA775 had a good 4 years of active lifespan and yet, 1156/1366 are barely cutting 2 years that another platform shows it's nose.

Talking about AMD's interroperability, just a reminder that SOME motherboard meant the majority of motherboards, with bios updates, were interroperable with the neighboor generation.

 

[Ctek]

What socket game? I love buying a new motherboard every generation... it keeps the dust bunnies at bay

I wouldn't call it a cash grab as they do make improvements add features with each new board... what I really hate is that the change the mounting holes around the CPU so I have to keep changing my block / block hardware for my water cooling. If Intel and AMD would create a common standard for block and cooler mounting I wouldn't have to play the games with my cooling setup every upgrade. I managed to keep my socket 775 block setup onto the 1156 platform with the EVGA board that had 775 mounting holes. Even managed to carry the 775 block setup over to the 1155 platform as Asrock again gave us that option.

I finally upgraded to a block with native 1155/1156/775. etc support... but I'm sure the next generation will negate that. Now that really tweaks me.

What I meant by a cash grab is that they could use the same platform to bring new evolution and simply push BIOS updates. Even if let's say a motherboards IOH/chipset could not fully tap the features and raw power of the new gen CPU, at least it could work, like the AM2+ CPU into an AM2 motherboard (crippled HT). It's a far less costly solution as far as the consumer is concerned.

A lifespan comparable to what LGA775 had was acceptable in my book. But that's only IMHO, everybody has their own.

 

[ND40oz]

The only fault I find on Intel's end is having 3 different chipsets on the launch of one socket. just a single Z68 would have been good enough for everyone, then the MB manufacturers could pick and choose features from there.

3 chipsets? There's 7 Cougar Point chipsets for 1155: B65, H61, H65, P67, Q65, Q67 and Z68.

Unless you mean 3 more then 1156 had, since there were 4 Ibex Peak chipsets: P55, H55, H57, Q57.

 

[Nathan_P]

My AM2+ board will take ANY AMD cpu from the AM2/2+/3 product range and i'm pretty sure that my busted AM2+ board will as well, yes they are both different chipsets. Makes it a lot easier to upgrade if you only have to do a BIOS update.

I cannot be bothered with the whole change everything routine every time i need some extra cpu power. My current board has had 3 different CPU's in it and will serve for a time yet

 

[iconiK]

1156/1366 are barely cutting 2 years that another platform shows it's nose.

1366 will be 3 years old when 2011 hits.
1156 is one year old, yes, that sucks.
1155 will be available for at least 2 years, maybe more if Haswell uses the same socket.
2011 is likely to last 4-5 years, as it unifies the dual and quad socket markets, which brings over the loooong turnover of quad socket.

 

[Dan_D]

How short people's memories are. AMD has segmented their market with different sockets. Not all that long ago they had Socket 754 which was pretty much a dead end. Socket 940 and 939 concurrently. Then they moved to socket G34, AM2, AM2+, AM3, and AM3+ looks like it will go no where. There market is still segmented with AM3 / AM3+ and G34. Intel has two sockets but once LGA2011 hits and they discontinue LGA1366 they'll still have two. LGA1155 for the mainstream and performance segment and LGA2011 for the high performance and server / workstation markets.

Sometimes companies have to change sockets. It's not that big a deal. Motherboards advance along with CPUs. Intel kept socket LGA775 for a long time but we still needed new motherboards for the most advanced CPUs as FSB's and memory clocks increased. The only good thing about keeping LGA775 around was being able to put older CPUs you might have carried over onto newer motherboards. With AMD you get to put newer CPUs on older motherboards but you either miss out on features compromise overclockability, etc. You also limit the processors performance in some instances. Remember AM2? When AM3 came out you could still use your AM3 processors on AM2+ motherboards but at a cost. Using DDR2 led to less performance than DDR3 based systems using the same processor because the newer processor was optimized for DDR3 memory but still supported DDR2 as a legacy solution. Some people are cheap and I get that but people who whine about sockets have a short memory or are turning a blind eye to the games AMD has played and is playing going forward. Oh, let's not forget AMD has a 95w and a 130w spec for CPUs. VRM design on many AMD boards was so haphazard that some boards were only designed to that 95w spec and as a result they were incompatible with 130watt CPUs. Yeah, that's nice isn't it?

You want a new CPU? Chances are you will want a new motherboard too. If you don't replace the motherboard you may be stuck to older memory and storage technologies or worse yet, reduced performance with your shiny new CPU.

Also, that socket compatibility AMD is famous for has left them with a screwed up thermal mounting solution and due to the proximity of the memory slots to the CPU, certain thermal solutions simply don't work on the AMD side or they create issues with regard to memory height even leading to the loss of usability of half your DIMM slots. With some solutions like the ever popular Thermalright Ultra 120 Extreme, you are forced into mounting the heat sink in a way which takes up as much DIMM real estate as possible and you are left with a less than ideal fan orientation. When AM2 came out the mounting holes weren't a big deal because coolers were smaller back then. With 6 core processors the heat sinks have had to get bigger and bigger. Now the mounting holes are really a problem. Either companies have to design separate AM3 compatible coolers or they have to modify (and sometimes do a poor job of it) their existing Intel compatible coolers to work on the AMD's as well. Yeah, AMD's approach to sockets sure is better. Each methodology has it's strengths and weaknesses but AMD's reluctance to move on to another socket (and thermal solution mounts) to make cheap bastards happy is hurting the platform at this point.

 

[jeremyshaw]

While I agree with some of your statements, I just don't agree with Intel platform disparity and short socket life. LGA775 had a good 4 years of active lifespan and yet, 1156/1366 are barely cutting 2 years that another platform shows it's nose.
[...]

[...]

A lifespan comparable to what LGA775 had was acceptable in my book. But that's only IMHO, everybody has their own.

LGA775 has a common socket for many CPU, but no forwards or backwards compabiility inbetween chipset generations and their related CPUs. So on LGA775, the chipset controlled what CPU you could use, and there was no "compatibile with everything," chipset. Either you had P4, 45nm C2D, or 65nm C2D support, or if you were lucky, some partial combination of the two.

 

[drescherjm]

I am still very mad at AMD for AM2. I mean there was a promise of quad cores however Phenom I was a disaster and then motherboard makers did not update their BIOSs to officially support Phenom II quad cores. I have a dozen or so such boards at work that I wish I could have upgraded.

 

[Dan_D]

LGA775 has a common socket for many CPU, but no forwards or backwards compabiility inbetween chipset generations and their related CPUs. So on LGA775, the chipset controlled what CPU you could use, and there was no "compatibile with everything," chipset. Either you had P4, 45nm C2D, or 65nm C2D, or if you were lucky, some partial combination of the two.

Until the 975 chipset compatibility was a mess. 975 could support Pentium D, Pentium D EE, Core 2 Duo, and Core 2 Quads I believe. Though it couldn't do 45nm duals or quads I don't believe. P35, X38 on up broke compatibility with older CPUs too. You couldn't use early Pentium 4 LGA775 CPUs with those P35 and newer boards.

 

[Forceman]

I guess I just don't see the issue. Historically, when you moved to a new processor generation, you moved to a new motherboard/socket with it, which is all Intel is doing (when you went from Pentium 2 to Pentium 3 - new socket, Pentium 3 to Pentium 4 - new socket, etc). I think the anomaly is that AMD has recently been trying out the "one socket to rule them all" approach, not that Intel keeps releasing new sockets for new generation chips. No one said you had to move from 775 to 1366 to 1156 to 1155 - you would move from 775 to one of the new generation sockets, and then keep that until the next generation hits. I guess you can complain about Intel segmenting the market by having two different sockets per generation (1366 and 1156, 2011 and 1156) but AMD tried that before also (939 and 754).

 

[MrXROK]

3 different sockets with miniscule difference. Wait another socket coming soon. :/

Intel is the suck.

 

[drescherjm]

3 different sockets with miniscule difference.

How many different sockets does AMD have now? Also do you realize that AM3+ is supposed to be a dead end.

 

[Mr Spocko]

Wow Dan D really hammering AMD here!

It's like this I went from Socket A..passed on 754 went to 939 and now AM3+. Yeah you're right AM3+ and FM1 are a dead end they should have just gone to FM2 directly. But hey that's life.
Cannot agree about holding things back if BD runs as well as some say on AM3+ then that's proof enough ot can cut it performance wise.

AMD have been slow to adopt 32nm too slow to move to that which has a lot more to do with thermals and power than a board ever has. Rant wise a bit OTT for my own tastes because I think the OP is correct is suggesting Intel have changed sockets too much..even if you want to point out some AMD things.

 

[soulsavior]

It's nice when a socket/chipset is supported for at least one CPU line after the original. Going out of their way to maintain compatibility beyond that is shooting themselves in the foot really.

 

[MrXROK]

How many different sockets does AMD have now? Also do you realize that AM3+ is supposed to be a dead end.

Why even bring up AMD?

Intel is suck when trying to milk or confusing the consumer.

I liked Intel a lot more when C2D days when they stuck out with 1 socket.

 

[Nettwerk]

You hate computers

 

[jeremyshaw]

Why even bring up AMD?

Intel is suck when trying to milk or confusing the consumer.

I liked Intel a lot more when C2D days when they stuck out with 1 socket.

yeah 1 socket that didn't work with 20-80% of all other chips (depending on the chipset) that could be physically inserted into the socket

AMD has 2 sockets AM3+ FM1 (upcomming FM2)
Intel has 2 sockets LGA1155 LGA1366 (upcomming LGA2011)


not counting BGA for Zacate/Atom, since it's not easily user replaceable
both not counting mobile sockets.

Even so, FM1 will only see one gen of APU, then will die in a year or so to make room for the FM2 socket. At least LGA1156 got to see two different generations of CPU, and even added in a video output without changing the socket (or was always designed to be there).

 

[teletran8]

1155 is the biggest chipset money grab of all time.

The whole idea of locking OC'ing on almost all the Intel 1155 CPU's is a major jackass greedy move IMHO (there's no 2100k for a reason to get your money cuz if u like 2 OC well you have to pay 219.99-314.99+99.99 for the cheapest lamest Intel chipset with a CPU that can get you a realistic OC. Because a 100mhz OC in 2011 is not an OC IMO sorree !)

LOL it's 2011 and Intel says your not gonna OC your chip unless you pay the premium for the amazing (bs'ing imo) engineering Intel gods and their almighty 1155 platform.

Yet on AM3 or AM3+ ALL chips can OC and there's no price premiums to be had either. No boards that INTENTIONALLY restrict OCing! Hmmm...

LOL! H61/H67 <-- worthless NO OCing chipsets, but they start @ 49.99 hmmm? Good deal + Good engineering right?

P67/Z68 <--- The P67's and Z68's start @ 99.99 on Newegg, hmmm kinda high if you ask me for a 16x/4x PCI-X board layout that doesn't support CFX.

AM3's start @ 44.99 , then the 16x/4x starts @ 64.99, and 8x/8x CFX like my mobo starts @ 89.99 and does 4 GHZ with ease. Hmm that's nice.

People who purchase intel are not into thinking, I will laugh when the only chips you can OC on Intel is the Extreme Edition chips. But ohhhh the performance they will all tell us AMD users and we will be like but I have a grand and you don't suckers.

 

[drescherjm]

AM3's start @ 44.99 , then the 16x/4x starts @ 64.99, and 8x/8x CFX

Which is the $5 less than lga1155 motherboards.

and does 4 GHZ with ease. Hmm that's nice.

At 4Ghz an AMD AM3 quad core chip becomes equivalent to a SB at stock.

 

[Andy735]

I don't think its a rip off because I tend to buy a new motherboard with each new CPU. Most recently I had a motherboard that while it could have supported the Phenom II quads would have been limited by FSB. It would be interesting to see how much the lower FSB, older pci express standard and reliance on DDR2 would have affected performance.

 

[BBA]

If a new socket is needed to gain features, such as for example more memory channels, then I see the socket changes as a good thing.

 

[Crosshairs]

While I agree with some of your statements, I just don't agree with Intel platform disparity and short socket life. LGA775 had a good 4 years of active lifespan and yet, 1156/1366 are barely cutting 2 years that another platform shows it's nose.

Does your old stuff suddenly stop working when new stuff is released?

I ran my 775 rig until 6 months ago....Im running an 1155 now and probably will be for 3 or 4 more years..........new cars come out every year..do you buy one every year?

1155 is the biggest chipset money grab of all time.

The whole idea of locking OC'ing on almost all the Intel 1155 CPU's is a major jackass greedy move IMHO (there's no 2100k for a reason to get your money cuz if u like 2 OC well you have to pay 219.99-314.99+99.99 for the cheapest lamest Intel chipset with a CPU that can get you a realistic OC. Because a 100mhz OC in 2011 is not an OC IMO sorree !)

LOL it's 2011 and Intel says your not gonna OC your chip unless you pay the premium for the amazing (bs'ing imo) engineering Intel gods and their almighty 1155 platform.

Yet on AM3 or AM3+ ALL chips can OC and there's no price premiums to be had either. No boards that INTENTIONALLY restrict OCing! Hmmm...

LOL! H61/H67 <-- worthless NO OCing chipsets, but they start @ 49.99 hmmm? Good deal + Good engineering right?

P67/Z68 <--- The P67's and Z68's start @ 99.99 on Newegg, hmmm kinda high if you ask me for a 16x/4x PCI-X board layout that doesn't support CFX.

AM3's start @ 44.99 , then the 16x/4x starts @ 64.99, and 8x/8x CFX like my mobo starts @ 89.99 and does 4 GHZ with ease. Hmm that's nice.

People who purchase intel are not into thinking, I will laugh when the only chips you can OC on Intel is the Extreme Edition chips. But ohhhh the performance they will all tell us AMD users and we will be like but I have a grand and you don't suckers.

Thanks for the laugh.....

 

[Poisoner]

So much fail in this thread. 1155 was needed because the clock generator and and a few other things were placed on the die itself.

Do any of you understand anything other than a double click?

 

[RickyJ]

It's a good thing they change physical sockets. Saves people mashing the wrong peg into the wrong hole, and saves the forums from endless "will X processor work with Y mobo" threads. I'd rather have a specialized socket and fully optimized pin configuration.

 

[evanisthecoastie]

I was unaware a board that can do crossfire, cant do crossfire, thanks teltran.

 

[BinarySynapse]

So much fail in this thread. 1155 was needed because the clock generator and and a few other things were placed on the die itself.

Do any of you understand anything other than a double click?

Sandy Bridge still uses an off chip clock generator. 1156 had two clock source, one for BCLK, and the other for PCIe, But with Sandy bridge, Intel changed the processor so that a single clock source ran ALL the integrated devices, Since PCIe is particularly sensitive to overclocking, you can't overclock the baseclock much anymore.

 

[MrGuvernment]

Wow Dan D really hammering AMD here!

It's like this I went from Socket A..passed on 754 went to 939 and now AM3+. Yeah you're ........................
I think the OP is correct is suggesting Intel have changed sockets too much..even if you want to point out some AMD things.

it isn't hammering, it is pointing out the truth and that Intel isn't so evil and that AMD has done the exact same thing.. both companies are guilty of numerous socket changes so to still claim Intel made to many changes is just wanting to turn a blind eye to poor little amd..

And in the end, no one is forcing you to buy a new CPU or motherboard, this sounds like the same people who cry that Vista wouldn't run on their ancient single core processors with 1G of ram....

 

[iconiK]

LOL! H61/H67 <-- worthless NO OCing chipsets, but they start @ 49.99 hmmm? Good deal + Good engineering right?

P67/Z68 <--- The P67's and Z68's start @ 99.99 on Newegg, hmmm kinda high if you ask me for a 16x/4x PCI-X board layout that doesn't support CFX.

AM3's start @ 44.99 , then the 16x/4x starts @ 64.99, and 8x/8x CFX like my mobo starts @ 89.99 and does 4 GHZ with ease. Hmm that's nice.

You do realize that the cheap motherboards are designed to handle 95W tops, and any overclocking is very likely to send the power delivery system way out of spec?

Also, you just told yourself that the good motherboards cost the same in either platform. x8/x8 CrossFire on AMD is $90, on Intel it's $100 for Z68.

People who purchase intel are not into thinking, I will laugh when the only chips you can OC on Intel is the Extreme Edition chips. But ohhhh the performance they will all tell us AMD users and we will be like but I have a grand and you don't suckers.

You don't really NEED to overclock an Intel chip to beat an AMD CPU. Take the i5 2500 for example, it mops the floor with Phenom II. Oh, you want to beat an overclocked Phenom? Xeon E3-1235. There, that will destroy OC'd Thuban.

You speak as if overclocking was a thing everyone does and provides massive benefits... eh, not really. Right now Intel is far better value than AMD, face it.

So much fail in this thread. 1155 was needed because the clock generator and and a few other things were placed on the die itself.

The clock generator is not on-die.

 

[Ctek]

So much fail in this thread. 1155 was needed because the clock generator and and a few other things were placed on the die itself.

Do any of you understand anything other than a double click?

And you don't find this a huge flaw in their planned lineup to kill the 1156 only 1.x year then release this revised model?

Some people here seems to think that I'm thinking (we, including people who agreees with the thread) that we should stick to one socket for the longest possible time. It's not true at all. I just find it irritating that socket generations gets almost EOL after around 2 years or less. 1156 and now 1366 are pretty much in this bandwagon.

 

[drescherjm]

1156 and now 1366 are pretty much in this bandwagon.
Reply With Quote

But so is AM3+ and FM1. Also remember that DDR4 most likely will require replacing the socket for AMD and Intel. To me I would prefer the motherboard to last more than 1 generation but it's not that big of a deal. I mean you buy a $200 motherboard and 2.5 years later you sell it for $125 to $150. Remember that replaced sockets tend to keep their value a lot better than sockets with a longer lifetime.

 

[iconiK]

1156 and now 1366 are pretty much in this bandwagon.

Again, 1156 was killed alive, we all agree (or should).

But 1366 will be 3 years old when 2011 hits, that's average socket lifetime.

 

[Dan_D]

Wow Dan D really hammering AMD here!

It's like this I went from Socket A..passed on 754 went to 939 and now AM3+. Yeah you're right AM3+ and FM1 are a dead end they should have just gone to FM2 directly. But hey that's life.
Cannot agree about holding things back if BD runs as well as some say on AM3+ then that's proof enough ot can cut it performance wise.

AMD have been slow to adopt 32nm too slow to move to that which has a lot more to do with thermals and power than a board ever has. Rant wise a bit OTT for my own tastes because I think the OP is correct is suggesting Intel have changed sockets too much..even if you want to point out some AMD things.

No. I'm just putting things in perspective for the fan boys demonizing Intel about their sockets. AMD has had plenty of changes, market segmentation and now dead end sockets with a very short life span. When I say AM3+ holds AMD back I'm speaking with regard to certain layout problems. All AM2 on up boards have a shitty layout. The memory sockets are too close to the CPU. Actually 939 suffered from this too but back then cooling solutions were smaller so it didn't matter.

 

[Kenjiwing]

All I have to say about AMD and sockets is fucking Socket 940.

I was dumb enough to by a FX-51 with socket 940.. months later socket 939 was announced.

Fuck you amd.

 

[drescherjm]

I really liked socket 940 but I had dual socket 940 and Opterons purchased at ebay prices (well at least at home). We still have 3 to 5 dual Opteron 940 systems running at work as file servers although I want to replace them because their pci-x slots are limiting my hba options and performance.

 

[thesecond]

as you move more controllers onto the die - memory, uncore, etc etc, you need more pins to power and control those components.

Sure 775 was all cute and sweet, because it was just cores. 1366 rolled around with a triple channel memory controller and needed the extra 591 pins for that.

now 2011 has moved a whole pile of the northbridge function onto the die, and needs more pins to reflect that.

You cant add more functions without more power and more lanes for i/o

 

[Dan_D]

Again, 1156 was killed alive, we all agree (or should).

But 1366 will be 3 years old when 2011 hits, that's average socket lifetime.

All I have to say about AMD and sockets is fucking Socket 940.

I was dumb enough to by a FX-51 with socket 940.. months later socket 939 was announced.

Fuck you amd.

Well both companies are guilty of the same shit, just not at the same time. AMD and Intel have both been guilty of short lived sockets and unnecessary sockets for the purpose of market segmentation. The list of "offensive" sockets is compiled based on their short lived nature primarily with some sockets lasting only a year or two or being specific to a very narrow scope of CPUs.

Intel's Offenses
Socket 1 (i486)
Socket 2 (i486)
Socket 3 (i486)
Socket 4 (Pentium 60 / 66 only)
Socket 5 (Replaced by socket 7, though socket 7 was at least backwards compatible.)
Socket 6 (i486 - Updated socket 3.)
Socket 370 (Not FCPGA type for Celerons only)
Socket 423 (Early Netburst Processors)
Socket 603 (Early Netburst Xeons)
Socket LGA1156 (Basically lasted about a year)

AMD's Offenses
Super 7 (Very short lived)
Slot A (Very short lived)
Socket 754
Socket 940 (Opteron, Athlon FX)
Socket 939
Socket AM2 (replaced by AM2+)
Socket AM3+ (Short life projected.)
Socket F1 (Short life projected.)

AMD used Intel's sockets for a long time otherwise they'd probably have a greater list of offenses. They simply used whatever Intel had designed at the time. From the onset of Super-7 onward they are about the same as Intel. It's just that their main stream socket seems to hang around awhile, but specification updates convolute CPU compatibility. Motherboard power specifications are also erratic leading to more CPU compatibility issues. AMD's socket A and Intel's LGA775 are probably the best examples of long lived sockets, but again chipset changes and FSB improvements lead to many CPUs needing a new chipset to get their necessary FSB support for full speed operation. Intel's socket LGA775 did have good backwards compatibility but the design always moved forward. AM3 has had good backwards compatibility with AM2+, just like AM3+ does with AM2, but they are still specification updates which cause some headaches.

No one comes out of the "socket game" smelling like roses. This has been my point from the very beginning. I'm not bagging on AMD for anything other than not leveraging the development cycle of Bulldozer to update their socket design and thermal solution to alleviate short comings of the current AM3 / AM3+ design.