Will Sklylake-E fit into upcoming skylake motherboards?

[wunkley]

Doing a new gaming build in the next few months. Trying to decide bewteen x99/ 5820k and the new skylake platform/ 6700k. I know lots of people are in a similar boat......

I'm looking for the best longish term investment with upgradability. Basically - if I get a 6700k 4 core 6700k but want to upgrade to a skylake E, 6 or 8 core (is that correct?) in the next year or so - will the E chips fit in the new 'normal' release skylake motherboards that launch soon? I'd like some sort of future proofing. Sounds like x99 might not have any future chip upgrade path?

 

[Random Guy Esq.]

No. Skylake-E will be as different to Skylake as Haswell-E is to Haswell i.e. a completely different socket, I'm afraid.

 

[wunkley]

Really? Not that I'm doubting you but do you have any links to solid info, I couldn't find much.

If that is true then I think I'm Deffo going x99 then but will wait for skylake release and reviews anyway.

 

[eddiew]

AFAIK the entire E-series has always been a different socket, generally coming with a bunch more DIMM slots, which necessitates extra pins on the CPU.

It would be very out of character for Skylake-E to be on the regular 'desktop' platform. Sadly.

 

[Freddie1980]

I don’t know what determines how many pins a CPU socket requires but the top E has always had a much bigger socket then mainstream desktops. If I had to guess it would due to the extra PCI Lanes, power, memory configuration/bandwidth/channels and all the other bits and bobs you get with this range.

Skylake-E won’t be out for some time, Haswell-E is fairly recent and Broadwell-E is next so we probably won’t see the light of day for at least a year if not longer. If you need prove just look at the all the previous E CPU’s since Sandy Bridge against the mainstream parts, I don’t see why Intel will break the pattern.

 

[VincentHanna]

Sandy Bridge/Ivy Bridge = LGA 1155
Sandy Bride E/Ivy Bridge E = LGA 2011 (X79)

Haswell/broadwell = LGA 1150
Haswell E = LGA 2011-3 (X99)

Skylake = 1151
Skylake E = ?

But it will be a different socket.

 

[drunkenmaster]

I don’t know what determines how many pins a CPU socket requires but the top E has always had a much bigger socket then mainstream desktops. If I had to guess it would due to the extra PCI Lanes, power, memory configuration/bandwidth/channels and all the other bits and bobs you get with this range.

Skylake-E won’t be out for some time, Haswell-E is fairly recent and Broadwell-E is next so we probably won’t see the light of day for at least a year if not longer. If you need prove just look at the all the previous E CPU’s since Sandy Bridge against the mainstream parts, I don’t see why Intel will break the pattern.

Broadwell isn't doing well with dual core, quad core both in mobile and desktop got delayed over a year and is still a problem, chances of them making 6-8 core versions properly is extremely low. It's likely cancelled in the same way Broadwell non e has been cancelled. As in launched but with no volume, with lots of talk about Skylake, done in every way possible to kill demand and not actually have to supply it. The lowest key launch Intel has ever done. It's not a 'real' product, Intel have put no weight behind it, they very clearly don't intend it to sell in volume.

Skylake is effectively on schedule, with Broadwell a year behind, they may launch the chips as a token gesture a month or two before the Skylake successors but they aren't serious launches and aren't serious chips. Skylake-E should still be out early next year.


In terms of what to get personally I decided against buying into another DDR3 based platform at this stage and am waiting for Skylake, possibly but not likely the E version.

Depends what kind of gaming you do, if more than 2 cards you have the money and the potential need for a 6-8 core Intel platform. If single card then frankly a x99 or skylake-e (x1199, Broadwell-e x1099?) is overkill now and will get more so with DX12 for a while.


6-8 core *-E platforms are great for really good multithreaded CPU performance, and in benchmarks they will be great but in real games today the extra cores do not a lot. With DX12, for at least a year or two we'll need less CPU power for the same games. In the future with better/more consistent CPU power available devs will start to use it more and CPU usage should increase. But short term 6-8 cores are likely to lose what advantage they have over 4 cores.

I'd say Skylake looks a good bet for cheapness, not wasting money on DDR3 and being great for DX12 for a year or two, then you have your memory, saved money and when games start to scale to use the available CPU power buy something more heavy hitting in a couple of years.

 

[Dave2150]

Broadwell isn't doing well with dual core, quad core both in mobile and desktop got delayed over a year and is still a problem, chances of them making 6-8 core versions properly is extremely low. It's likely cancelled in the same way Broadwell non e has been cancelled. As in launched but with no volume, with lots of talk about Skylake, done in every way possible to kill demand and not actually have to supply it. The lowest key launch Intel has ever done. It's not a 'real' product, Intel have put no weight behind it, they very clearly don't intend it to sell in volume.

Skylake is effectively on schedule, with Broadwell a year behind, they may launch the chips as a token gesture a month or two before the Skylake successors but they aren't serious launches and aren't serious chips. Skylake-E should still be out early next year.

Err? What nonsense is this?

Broadwell has had 8 core CPU's out for months now - it's called Xeon-D (8 core SOC). It was launched in March. Intel can make a 8-core Broadwell no problem.

Broadwell quad cores for desktop has been available for weeks, it's even in stock on OCUK right now:

https://www.overclockers.co.uk/showproduct.php?prodid=CP-577-IN&groupid=701&catid=6&subcat=567

 

[drunkenmaster]

Err? What nonsense is this?

Broadwell has had 8 core CPU's out for months now - it's called Xeon-D (8 core SOC). It was launched in March. Intel can make a 8-core Broadwell no problem.

Broadwell quad cores for desktop has been available for weeks, it's even in stock on OCUK right now:

https://www.overclockers.co.uk/showproduct.php?prodid=CP-577-IN&groupid=701&catid=6&subcat=567

First, it hasn't been available for weeks, it launched what early june and stock was supposed to be available within a week, everyone posted pre-order dates, they were all missed world wide.

As for the 8 core soc. I'm under the impression they did a LOT of things to get that working. For one it's a SOC so it's more than one chip on the package. I'm not entirely sure as I've seen people say it is two dies when referring to the chip as well as the SOC. SO I'm not sure if it's two quad cores stuck together or not. But they also appear to have stripped out a lot of the IO and put that together as a south bridge made separately and on package. IE, it's made quite differently to a normal desktop chip.

If you strip out bits that are perhaps causing trouble and make various bits to different chip designs then combine them on package, well, that is where the industry is going to improve yields just via interposers rather than on package(lower power and higher bandwidth/lower latency connections between the chips when done via interposer).

It's hardly a standard chip and carries server pricing. What you can make and sell at £250 a chip and £500 a chip are very very very different things. Crap yields could mean you lose money selling at £250, but make a £200 profit selling at £500.

It's very clear to anyone who is unbiased that Broadwell desktop is a small scale release. NEver before has Intel launched something, while talking about the next product due what 2 months later, while promising dates and missing them by a month... after a year delay. Then pricing those chips at way more than the previous chips making absolutely no one want them.

Let me guess, the year delay, the 14nm delays, the higher voltage, the lower clock speeds, the low clock speeds on that 8 core xeon, the lack of higher clock speed desktop Broadwell parts for a year, the lack of higher clockspeed Xeons. The entire world knows Intel is having major 14nm problems.

 

[Dave2150]

First, it hasn't been available for weeks, it launched what early june and stock was supposed to be available within a week, everyone posted pre-order dates, they were all missed world wide.

As for the 8 core soc. I'm under the impression they did a LOT of things to get that working. For one it's a SOC so it's more than one chip on the package. I'm not entirely sure as I've seen people say it is two dies when referring to the chip as well as the SOC. SO I'm not sure if it's two quad cores stuck together or not. But they also appear to have stripped out a lot of the IO and put that together as a south bridge made separately and on package. IE, it's made quite differently to a normal desktop chip.

If you strip out bits that are perhaps causing trouble and make various bits to different chip designs then combine them on package, well, that is where the industry is going to improve yields just via interposers rather than on package(lower power and higher bandwidth/lower latency connections between the chips when done via interposer).

It's hardly a standard chip and carries server pricing. What you can make and sell at £250 a chip and £500 a chip are very very very different things. Crap yields could mean you lose money selling at £250, but make a £200 profit selling at £500.

It's very clear to anyone who is unbiased that Broadwell desktop is a small scale release. NEver before has Intel launched something, while talking about the next product due what 2 months later, while promising dates and missing them by a month... after a year delay. Then pricing those chips at way more than the previous chips making absolutely no one want them.

Let me guess, the year delay, the 14nm delays, the higher voltage, the lower clock speeds, the low clock speeds on that 8 core xeon, the lack of higher clock speed desktop Broadwell parts for a year, the lack of higher clockspeed Xeons. The entire world knows Intel is having major 14nm problems.

I'm pretty sure that come August 5th we'll see bucketloads of 6600K and 6700K (all 14nm Skylake desktop CPU's) available.

 

[Silent_Scone]

He's an irrationally deluded man.

 

[drunkenmaster]

I'm pretty sure that come August 5th we'll see bucketloads of 6600K and 6700K (all 14nm Skylake desktop CPU's) available.

and where exactly did I say we wouldn't see lots of Skylake CPUs available... no where?

When you decided to post about a chip you really didn't understand thinking it was a 'normal' higher clocked, higher power single chip Xeon but wasn't you did the usual thing, claim something completely besides the point while quoting me in an attempt to smear what I was saying with whatever garbage you made up.

Silent Scone fell for it, quelle surprise. But I didn't say anything about delays to Skylake, the 66/6700k's are SKYLAKE, not Broadwell. Broadwell got a token desktop launch with failed delivery dates and distributors lacking product precisely because they dumped it and moved to Skylake.

Here's a hint, when 14nm volume production is promised for 2013 and desktop, quad core, high powered Broadwell is promised for mid 2014..... then it finally gets launched 1-2 months before the next parts are due, with no volume, in mid 2015.... yeah, the process has problems. Find someone serious in the industry who doesn't think Intel didn't have a monumental problem with 14nm... or 10nm, because Intel has admitted to 10nm issues and after a year of pretending the delay on 14nm was business as usual they admitted to problems with that as well. 14nm is WAY behind schedule, it's STILL behind schedule. Yields on chips today won't be anywhere near where they'd be if they managed to ship Broadwell as planned a full year ago, the process would be far more mature. They usually have 6-12months with only 5-15% of production on the newest node but over the 12-24 month period they normally ramp their production towards 50-60% all on the new node, none of which has happened yet with 14nm.

 

[Dave2150]

and where exactly did I say we wouldn't see lots of Skylake CPUs available... no where?

When you decided to post about a chip you really didn't understand thinking it was a 'normal' higher clocked, higher power single chip Xeon but wasn't you did the usual thing, claim something completely besides the point while quoting me in an attempt to smear what I was saying with whatever garbage you made up.

Silent Scone fell for it, quelle surprise. But I didn't say anything about delays to Skylake, the 66/6700k's are SKYLAKE, not Broadwell. Broadwell got a token desktop launch with failed delivery dates and distributors lacking product precisely because they dumped it and moved to Skylake.

Here's a hint, when 14nm volume production is promised for 2013 and desktop, quad core, high powered Broadwell is promised for mid 2014..... then it finally gets launched 1-2 months before the next parts are due, with no volume, in mid 2015.... yeah, the process has problems. Find someone serious in the industry who doesn't think Intel didn't have a monumental problem with 14nm... or 10nm, because Intel has admitted to 10nm issues and after a year of pretending the delay on 14nm was business as usual they admitted to problems with that as well. 14nm is WAY behind schedule, it's STILL behind schedule. Yields on chips today won't be anywhere near where they'd be if they managed to ship Broadwell as planned a full year ago, the process would be far more mature. They usually have 6-12months with only 5-15% of production on the newest node but over the 12-24 month period they normally ramp their production towards 50-60% all on the new node, none of which has happened yet with 14nm.

I specifically referred to Xeon-D as a SoC. Here, I'll quote my own post to demonstrate:

it's called Xeon-D (8 core SoC).

SoC = System-on-a-chip

So, no, I didn't incorrectly refer to Xeon-D as a "normal' higher clocked, higher power single chip Xeon" - I correctly referred to it as a SoC.

I'm not entirely sure as I've seen people say it is two dies when referring to the chip as well as the SOC. SO I'm not sure if it's two quad cores stuck together or not.

I'll answer that for you. It's not "two quad cores stuck togetther" - it's simply the 8 Core Broadwell on one die, along with the PCH on another die. They both reside inside the SoC, covered by the IHS.

To quote Anandtech's description of it:

In a nutshell, the Xeon D-1540 is two silicon dies in one highly integrated package. Eight 14 nm Broadwell cores, a shared L3-cache, a dual 10 gigabit MAC, a PCIe 3.0 root with 24 lanes find a home in the integrated SoC whereas in the same package we find four USB 3.0, four USB 2.0, six SATA3 controllers and a PCIe 2.0 root integrated in a PCH chip.

Since it would seem that you are the one who lacks an understanding of this SoC, I felt it appropriate to refer you to the excellent Anandtech review of Xeon-D, it contains a vast amount of information for you to further educate yourself about this SoC:

http://www.anandtech.com/show/9185/intel-xeon-d-review-performance-per-watt-server-soc-champion/2

Hopefully after you have taken the time to fully digest both my post and the Anandtech review, you'll come to the realization that Xeon-D is in fact an 8 core SoC, not "two quad-cores stuck together"....

 

[Troezar]

I thought it was common knowledge regarding the almost paper launch? It's all about the investors, Intel had to make an effort to at least launch a tiny volume otherwise they would be seen to be struggling to stay on course. Now Skylake had them back on target they can forget about the previous issues and sell that fire all they're worth. Sounds like good business practice to me.

Skylake didn't sound as good as I was hoping though. I'll be sticking with my 2500k until the next round.