Intel Coffee Lake CPUs Have Different Pin Configuration than Previous Generation

[Papahyooie]

Intel is set to release its newest generation of processors, known as code name Coffee Lake in just a few days. Recent controversy has focused around the inability of newer processors to be used with motherboards supporting the previous generation CPUs. Intel has released data sheets for Coffee Lake CPUs, and images of the socket's layout support Intel's statement that Coffee Lake will indeed need a new socket design. Specifically, there are more pins responsible for delivering the main power for the CPU cores, known as VCC pins, with Coffee Lake motherboards sporting 146 VCC pins illustrated in the first image below, compared to Kaby Lake and Skylake's 128 illustrated in the second image.

With these pins responsible for delivering power to the CPU cores, this may suggest that previous generation motherboards allowed for lower power operation than is possible with the higher core count of Coffee Lake processors. Whatever the reason, it definitely confirms that Coffee Lake CPUs are hardware-incompatible with previous generation motherboards, and not simply a software or BIOS-level lock.



View at TechPowerUp Main Site

 

[cadaveca]

Typo in third sentence. And interestingly enough, I had suggested a while ago that there would be more power pins, and here they are. I mean, adding two cores would obviously require more power pins, or perhaps clocking would be negatively affected by the existing pads sharing power to more areas of the chip...

 

[VSG]

Typo in third sentence. And interestingly enough, I had suggested a while ago that there would be more power pins, and here they are. I mean, adding two cores would obviously require more power pins, or perhaps clocking would be negatively affected by the existing pads sharing power to more areas of the chip...

Typo corrected, I'll take responsibility for this one

 

[Hood]

This also confirms that it won't work the other way, either. 6th and 7th gen CPUs can't work in Z370 boards, even though they use the same 1151 socket. The new boards would either fry a 6700/7700, or refuse to boot. Strangely enough, they all have identical alignment notches, possibly allowing inexperienced builders to make this mistake. I predict some confusion in a few years, when people building from old parts (dead PCs) cobble together parts that look exactly the same, but are incompatible.

 

[cadaveca]

This also confirms that it won't work the other way, either. 6th and 7th gen CPUs can't work in Z370 boards, even though they use the same 1151 socket. The new boards would either fry a 6700/7700, or refuse to boot. Strangely enough, they all have identical alignment notches, possibly allowing inexperienced builders to make this mistake. I predict some confusion in a few years, when people building from old parts (dead PCs) cobble together parts that look exactly the same, but are incompatible.

Actually, when situations like this happen, there are usually pretty clear indications on the packaging or even on the boards themselves to warn users of potential problems. Also, retailers should have staff educated enough to avoid this sort of thing if purchases are made in person, but online purchases could be problematic. This can easily addressed though, by a small bit of code that asks questions when people make such purchases...

 

[OSdevr]

Also, retailers should have staff educated enough to avoid this sort of thing if purchases are made in person

But seriously, this has always been dealt with by using a different socket. If it must have the same number of pins then it should have different notches and a new name (ala 2011-3). This may not be as bad as the days when the power and cassette drives had the same connector and mixing them up could fry your computer, but it is a similarly bad design.

 

[FR@NK]

At first glance, it still looks compatible as the added voltage pins are just reserved pins on the older socket.

 

[chaosmassive]

Okay okay.... Intel, we get it
its from technical standpoint that I can't put KL chip onto your new chipset 300 series
clearly not from business motive at all
Gee I thought Intel was greedy company, turn out it wasn't
Thanks Intel /s

 

[TheoneandonlyMrK]

Actually, when situations like this happen, there are usually pretty clear indications on the packaging or even on the boards themselves to warn users of potential problems. Also, retailers should have staff educated enough to avoid this sort of thing if purchases are made in person, but online purchases could be problematic. This can easily addressed though, by a small bit of code that asks questions when people make such purchases...

Your right , and were right but you aren't getting him , he means in a few years people will blow up second hand or old parts, probably true given the variety of intel chips and sockets around atm but not that important to oems or intel.
All in this is the first time intel has shown an imperical socket pin difference comparison like this , it's reasuring and reasonable , it's a shame imho they are not as clear consistently with specs and info but hopefully they might improve.
Oh and after a look at both pics just on a phone i think intels definitely just shown how easily it can puff out a bit of smoke and completely blind those that want to be, to just how devious they are..
Every chip has an interposer attached to its Beol this is a multi layer pcb chip mouting package that connects and defines all of the actual chips connections to the socket, some are virtually pass through , like memory channels others are not ,but this extremely important and unloved part of every chip is an easily changed part , the interposer costs significantly less to design and make then the chip.
It's a play for money in my book still ,but with at least some reasoning and fair play.

 

[eidairaman1]

We already know this, why is this news again?

 

[TheLostSwede]

We already know this, why is this news again?

Because now we got the official explanation from Intel? It wasn't clear as to why older CPUs wouldn't work in the new boards, although this seems like a half assed excused imho.

 

[Frick]

Actually, when situations like this happen, there are usually pretty clear indications on the packaging or even on the boards themselves to warn users of potential problems. Also, retailers should have staff educated enough to avoid this sort of thing if purchases are made in person, but online purchases could be problematic. This can easily addressed though, by a small bit of code that asks questions when people make such purchases...

Yeah, that'll happen. Small piece of code.

 

[silentbogo]

We already know this, why is this news again?

Because there was no 8th gen datasheet available at that time.

This also confirms that it won't work the other way, either. 6th and 7th gen CPUs can't work in Z370 boards, even though they use the same 1151 socket. The new boards would either fry a 6700/7700, or refuse to boot. Strangely enough, they all have identical alignment notches, possibly allowing inexperienced builders to make this mistake. I predict some confusion in a few years, when people building from old parts (dead PCs) cobble together parts that look exactly the same, but are incompatible.

In the old forum thread we've discussed this thing: this is not something new, and it did happen in the past. Just like the confusion with many iterations of LGA775, just like sAM2/sAM2+/sAM3 etc. etc. etc.
I believe that the leaked picture of the 8th gen CPU retail box design already has a label "Requires 300-series motherboard" or something like that. Same goes for motherboards: all of them have QVL and compatibility lists. So, if someone sticks a wrong CPU in a wrong board/socket - it's their fault. There is no excuse for being illiterate. Plus, there will always be dummies who will stick LGA1155 CPU in an LGA1150 socket, bend PGAs, put TIM on the bottom of the CPU, or clean the debris from LGA socket with their fat dirty clumsy fingers, buy a $300 tablet for their 4 y.o. kid to watch cartoons, perform bend tests on their own smartphones, throwing electronic things into liquid or spilling liquid onto electronic things. If those people did not exist, I'd be out of job right now.

 

[londiste]

How difficult would it be to have identically formatted images :/
I get it, these are copypasted out of Intel's datasheet and glued together but still.

 

[B-Real]

Because now we got the official explanation from Intel? It wasn't clear as to why older CPUs wouldn't work in the new boards, although this seems like a half assed excused imho.

This isn't an official explanation.

"With these pins responsible for delivering power to the CPU cores, this may suggest"

 

[GorbazTheDragon]

They are reserved pins, should have enabled them before

Zzzzzzzz

 

[john_]

And the 8 core model will be incompatible with 370 motherboards because Intel would use even more reserved pins with that model and when they where making the 6 core model they didn't knew they would use more pins with the 8 core model. So they regret they will have another incompatible platform again in less than a year. Also i5, i3 and Pentium processors, in the Coffee lake family, would also need more power than the top of the line Kaby Lake i7 processors, for stability and overclocking purposes, even the locked models, so it will be logical for those processors to also not work on 270 motherboards for example.

On the other hand they managed to fit an 18 core processor on a socket they where designing for 12 core processors, with no extra pins needed, having much higher power consumption than the 165W TDP limit of that socket. No problems there. But there they didn't wanted to compete with just a 12 core model against a 16 core model, at least from a marketing perspective neither had the time to create/market a new chipset/socket.

As I said in the past.
Does it has to do with Intel? Excuses, excuses, excuses.
Does it has to do with Nvidia? Excuses, excuses, excuses.
Does it has to do with AMD? The firing squad is ready.

 

[GorbazTheDragon]

That 3w TDP difference wew...

It's not like every overclocker has been running 200+ watts through this socket for ages...

 

[Parn]

And the 8 core model will be incompatible with 370 motherboards because Intel would use even more reserved pins with that model and when they where making the 6 core model they didn't knew they would use more pins with the 8 core model. So they regret they will have another incompatible platform again in less than a year. Also i5, i3 and Pentium processors, in the Coffee lake family, would also need more power than the top of the line Kaby Lake i7 processors, for stability and overclocking purposes, even the locked models, so it will be logical for those processors to also not work on 270 motherboards for example.

On the other hand they managed to fit an 18 core processor on a socket they where designing for 12 core processors, with no extra pins needed, having much higher power consumption than the 165W TDP limit of that socket. No problems there. But there they didn't wanted to compete with just a 12 core model against a 16 core model, at least from a marketing perspective neither had the time to create/market a new chipset/socket.

As I said in the past.
Does it has to do with Intel? Excuses, excuses, excuses.
Does it has to do with Nvidia? Excuses, excuses, excuses.
Does it has to do with AMD? The firing squad is ready.

Well said.

While it is true now that CFL pin-out is technically different from SKL/KBL, it still doesn't change the fact that Intel delibrately designed those sockets in such ways because of greed.

This also pretty much confirms the rumor that 8-core CFL and Z390 would not be compatible with 6-core/Z370.

 

[OSdevr]

In the old forum thread we've discussed this thing: this is not something new, and it did happen in the past. Just like the confusion with many iterations of LGA775, just like sAM2/sAM2+/sAM3 etc. etc. etc.

Intel's Slot 1 and AMD's Slot A also used the same connector but were very different electrically. I just thought we had moved beyond that stuff by now. LGA 2011 had three versions but all were keyed differently. You can't make something idiot proof but keying a socket differently is such an easy step.

 

[Durvelle27]

Because there was no 8th gen datasheet available at that time.


In the old forum thread we've discussed this thing: this is not something new, and it did happen in the past. Just like the confusion with many iterations of LGA775, just like sAM2/sAM2+/sAM3 etc. etc. etc.
I believe that the leaked picture of the 8th gen CPU retail box design already has a label "Requires 300-series motherboard" or something like that. Same goes for motherboards: all of them have QVL and compatibility lists. So, if someone sticks a wrong CPU in a wrong board/socket - it's their fault. There is no excuse for being illiterate. Plus, there will always be dummies who will stick LGA1155 CPU in an LGA1150 socket, bend PGAs, put TIM on the bottom of the CPU, or clean the debris from LGA socket with their fat dirty clumsy fingers, buy a $300 tablet for their 4 y.o. kid to watch cartoons, perform bend tests on their own smartphones, throwing electronic things into liquid or spilling liquid onto electronic things. If those people did not exist, I'd be out of job right now.

But AM2/AM2+ was cross compatibile just like AM3/AM3+

 

[GorbazTheDragon]

But AM2/AM2+ was cross compatibile just like AM3/AM3+

Backwards compatibility of CPUs on older boards is not very useful in many cases.

Say Z68 boards supported Haswell CPUs and you had a Z68+2500k which you swapped for a 4790k. You would still be left without USB3.0, only 2 SATA3 ports, PCIe2.0, a small DMI bus, no m.2, more restrictive iGPU display configs. In those cases it would only be sensible to rewire the socket because you are now having to add new interfaces without removing the older ones which may not even be used on the newer socket, may require additional I/O on the CPU chip, etc... And as a user you would upgrade the board as well more often than not...

That said, in this case it's pretty stupid unless I've overlooked something on the diagrams that is actually changed... They added 14% extra Vccin connections for a chip that seems to draw only marginally more power in the first place, on a socket that seems to have no trouble running way out of power spec, on pins that were unused...

 

[eidairaman1]

Backwards compatibility of CPUs on older boards is not very useful in many cases.

Say Z68 boards supported Haswell CPUs and you had a Z68+2500k which you swapped for a 4790k. You would still be left without USB3.0, only 2 SATA3 ports, PCIe2.0, a small DMI bus, no m.2, more restrictive iGPU display configs. In those cases it would only be sensible to rewire the socket because you are now having to add new interfaces without removing the older ones which may not even be used on the newer socket, may require additional I/O on the CPU chip, etc... And as a user you would upgrade the board as well more often than not...

That said, in this case it's pretty stupid unless I've overlooked something on the diagrams that is actually changed... They added 14% extra Vccin connections for a chip that seems to draw only marginally more power in the first place, on a socket that seems to have no trouble running way out of power spec, on pins that were unused...

Actually it is beneficial, 8100 vs 8300 series FX, 8300 fixed alot pertaining to single core performance, multicore was better too, plus didn't run nearly as hot, it was a drop in upgrade, not needing to change the mobo out helps save money.

 

[silentbogo]

Backwards compatibility of CPUs on older boards is not very useful in many cases.

I'm talking about stupid people trying to run Piledrivers and Bulldozers on 10y.o. AM2/AM2+ boards in total disregard of an absent DDR2 memory controller and some major architectural changes.
If it fits it should work though? Right? It worked for Phenoms and stuff...

P.S. I know, that you cannot do it backwards due to keying, but I've seen some people (clients) with bent CPU pins who tried to install their old CPU into a newer AM3/AM3+ board to save time and money.
P.P.S. People get upset when either Intel or AMD changes something, even though there was nothing engineeringly-challenging (gotta coin this term) to make it fully compatible with an old socket, but I still want to remind you all that it did happen in the past many times, and some of those even went unnoticed. At least today we have a "heads-up" message.

As I said in the past.
Does it has to do with Intel? Excuses, excuses, excuses.
Does it has to do with Nvidia? Excuses, excuses, excuses.
Does it has to do with AMD? The firing squad is ready.

Oh... just stop with this "beating on AMD underdog" bullshit. All of those companies get their fair share of criticism, both reasonable and unreasonable.

 

[john_]

You would still be left without USB3.0, only 2 SATA3 ports, PCIe2.0, a small DMI bus, no m.2, more restrictive iGPU display configs.

Maybe there are more examples to base your argument, but the fact is that full ATX boards have a number of PCI and PCIe slots for that reason. They are not there for decoration. You can add USB 3.0 in a PCie slot for less than $10, extra SATA ports for less than $10, M.2 adapters probably for less than $10. The iGPU limitations are irrelevant if you have a dedicated graphics card and the PCIe 2.0 bus is far from a limiting factor in most of those cases.

Oh... just stop with this "beating on AMD underdog" bullshit. All of those companies get their fair share of criticism, both reasonable and unreasonable.

Fair share? "Oh... just stop with this bullshit". In modern society talking about "fair shares" is synonym to naiveness.