Kaby Lake: Will it work in DDR4 Skylake motherboards?

[chucky2]

I'm sort of assuming that Kaby Lake will only support DDR4 (that Skylake is the memory bridge CPU between DDR3 and DDR4, and that Intel will drop DDR3L support with Kaby Lake).

So, if someone were to buy a DDR4 Skylake motherboard now, has Intel said that Kaby Lake will drop into that platform? Or will Kaby Lake need an entirely new motherboard?

I searched on this...just couldn't find anything on it.

 

[tential]

Honestly if Intel is going to keep switching the pins but not do a node shrink nor any significant change then that is screwed up at this point.

I couldn't care less in prior years. Now if we're on the same node for 3 years, I expect the cpu to be compatible and not need a new board. Sure new boards can be released with new features, but this is way too long on the same nodes now for Intel specifically to expect to launch new mobo/cpu combo that are incompatible with previous cpus

 

[crashtech]

The record for two CPU gens in one socket so far: 1156, no, 1155, yes, 1150, sort of. Three different answers for the last three sockets, there's no telling whether future CPUS will work with 1151.

 

[ShintaiDK]

Its unlikely that Kaby Lake drops DDR3L support. Its a Skylake CPU with an updated IGP. I cant imagine it wont work in current boards either.

 

[dark zero]

Kabylake could definately drop the DDR3 support due the iGPU update

 

[MrTeal]

The record for two CPU gens in one socket so far: 1156, no, 1155, yes, 1150, sort of. Three different answers for the last three sockets, there's no telling whether future CPUS will work with 1151.

Technically 1156 is more of a sort of than a no. There weren't any quad core 32nm parts on 1156, but there were 32nm Clarkdale dual cores available. No one would consider them an upgrade over Lynnfield though.

 

[chucky2]

Well at least I feel a little better that I wasn't missing any official announcement by Intel on Kaby Lake needing a new motherboard...every else is as in the dark as me...

 

[zir_blazer]

After that they did with Z97 and Broadwell, I wouldn't trust Intel in future proofing anything.
Kaby Lake should probabily work, but the next node shrink may not.

 

[jpiniero]

Cannonlake on the other hand is almost certainly on a new socket.

 

[chucky2]

It sure would be nice if Intel could give people 3 CPU releases on a socket like AMD is able to do. Heck, both my uncles have AM3 CPUs working on Gigabyte AM2 (not AM2+) boards, and with SSD, they are completely fine. Drop in something like a GTX 950 when they finally get cheaper, and I doubt for any kind of general use anyone could tell the difference between that rig and the Haswell Refresh i3 we built ones wife.

 

[ShintaiDK]

It sure would be nice if Intel could give people 3 CPU releases on a socket like AMD is able to do. Heck, both my uncles have AM3 CPUs working on Gigabyte AM2 (not AM2+) boards, and with SSD, they are completely fine. Drop in something like a GTX 950 when they finally get cheaper, and I doubt for any kind of general use anyone could tell the difference between that rig and the Haswell Refresh i3 we built ones wife.

Backwards compatibility should never be in the way of improvements or platform changes. Its more a problem for AMD than a benefit. Amount of users doing upgrade without replacing motherboard is very small.

 

[Insert_Nickname]

Backwards compatibility should never be in the way of improvements or platform changes.

There was a time when I'd have agreed with you. But with both memory controller, northbridge and GPU integrated, and since every Intel CPU since Nehalem has used DMI 2.0 for connecting to the PCH, would backwards really be -that- hard to achieve?

This only applies to socketed desktop chips. For mobile, it absolutely makes sense to not keep backwards compatibility. After all, who upgrades their laptops CPU?

 

[ShintaiDK]

There was a time when I'd have agreed with you. But with both memory controller, northbridge and GPU integrated, and since every Intel CPU since Nehalem has used DMI 2.0 for connecting to the PCH, would backwards really be -that- hard to achieve?

This only applies to socketed desktop chips. For mobile, it absolutely makes sense to not keep backwards compatibility. After all, who upgrades their laptops CPU?

The electrical properties of the socket, yes. Else we could argue that sockets should never change because it would be a DMI/A-link/HT/FSB/QPI etc. And we should essentially all be using LGA1156/FM1 for mainstream.

 

[Insert_Nickname]

The electrical properties of the socket, yes. Else we could argue that sockets should never change because it would be a DMI/A-link/HT/FSB/QPI etc. And we should essentially all be using LGA1156/FM1 for mainstream.

I look at it more like you could potentially drop-in an i5-750 in a Skylake board, and the other way round, dropping in a 6700K in a Nehalem board. It'd just need to work at the lowest common denominator.

So, f.x. if you're using a 750 in a new board, you'd only get the features 750 supports. Using newer CPUs in older boards, you'd only get the features the board supports.

Its a little abstract, but I hope you get the idea. Besides, socketed CPU used to work like that.

 

[mikk]

I'm sort of assuming that Kaby Lake will only support DDR4 (that Skylake is the memory bridge CPU between DDR3 and DDR4, and that Intel will drop DDR3L support with Kaby Lake).

So, if someone were to buy a DDR4 Skylake motherboard now, has Intel said that Kaby Lake will drop into that platform? Or will Kaby Lake need an entirely new motherboard?

I searched on this...just couldn't find anything on it.

There is no change, SKL PCH is supported according to the leak.

https://benchlife.info/intel-kaby-lake-scedule-show-10nm-cannonlake-will-push-back-to-2018-10162015/

 

[ShintaiDK]

I look at it more like you could potentially drop-in an i5-750 in a Skylake board, and the other way round, dropping in a 6700K in a Nehalem board. It'd just need to work at the lowest common denominator.

So, f.x. if you're using a 750 in a new board, you'd only get the features 750 supports. Using newer CPUs in older boards, you'd only get the features the board supports.

Its a little abstract, but I hope you get the idea. Besides, socketed CPU used to work like that.

In theory. But you would have to sacrifice quite a bit in trade offs. And forget Nehalem boards, that's not going to work in this case.

Lowest denominator also means power efficiency and performance loss.

AMD and Intel isn't changing sockets to annoy the tiny crowd that do CPU upgrades without new motherboards.

 

[Insert_Nickname]

In theory. But you would have to sacrifice quite a bit in trade offs. And forget Nehalem boards, that's not going to work in this case.

Lowest denominator also means power efficiency and performance loss.

Of course, that goes without saying.

But we're still talking desktop boards. A few W more to get an older CPU working with a new mainboard isn't as important as it is for mobile applications.

BTW, the Nehalem/Westmere boards I'm referring to are the LGA-1156 variety, not LGA-1366. Which from a high level look quite similar to LGA-1155/50/51.

I'm reminded of my old Super Socket 7 Epox MVP3-G2. That board could run everything from a Pentium 133 to a K6-2 550MHz, and even the K6-3. If you could get one. Or the 440BX, which ran (almost) everything Slot1 and Socket 370.

AMD and Intel isn't changing sockets to annoy the tiny crowd that do CPU upgrades without new motherboards.

More like Intel/AMD thinks when you buy a CPU, you should buy a chipset too. Its all business.

 

[MrTeal]

In theory. But you would have to sacrifice quite a bit in trade offs. And forget Nehalem boards, that's not going to work in this case.

Lowest denominator also means power efficiency and performance loss.

Of course, that goes without saying.

But we're still talking desktop boards. A few W more to get an older CPU working with a new mainboard isn't as important as it is for mobile applications.

BTW, the Nehalem/Westmere boards I'm referring to are the LGA-1156 variety, not LGA-1366. Which from a high level look quite similar to LGA-1155/50/51.

I'm reminded of my old Super Socket 7 Epox MVP3-G2. That board could run everything from a Pentium 133 to a K6-2 550MHz, and even the K6-3. If you could get one. Or the 440BX, which ran (almost) everything Slot1 and Socket 370.

Intel isn't going to accept performance loss in their new systems in order to support backwards compatibility unless there is a compelling reason to do so.

For example, in the case of LGA1151 you would need Skylake CPUs to handle the 1.5V memory signalling of old DDR3 boards. That doesn't come without tradeoffs, or SL would likely already support DDR3 instead of just DDR3L. Similarly, you would not be able to introduce new features like FIVR on the common socket.

 

[chucky2]

There is no change, SKL PCH is supported according to the leak.

https://benchlife.info/intel-kaby-lake-scedule-show-10nm-cannonlake-will-push-back-to-2018-10162015/

Thx for that, is that trustable do you think as far as accuracy?

 

[jpiniero]

This only applies to socketed desktop chips. For mobile, it absolutely makes sense to not keep backwards compatibility. After all, who upgrades their laptops CPU?

You can't anyway, at least on Skylake mobile since it's BGA only. It's only a matter of time before they move the desktop processors to BGA as well, but the timing really depends on how much OEMs care about the sockets.

 

[Arachnotronic]

Thx for that, is that trustable do you think as far as accuracy?

BenchLife is the best in the biz right now when it comes to leaking official Intel documents.