Lga 1151

So Intel are bringing a new CPU, the Skylake-S line, and along with it another socket refresh. Are they just out so screw customers and rake in cash for mobo makers?

One extra pin, a small bump in speed, and a heap of cash to get in on it on ly to have to do it again next refresh/generation. All this and exorbitant DDR4 prices.

Linkage: http://www.kitguru.net/components/cpu/matthew-wilson/intel-skylake-s-cpus-to-support-new-lga-1151-socket/



I think this appears to be more of an engineering thing. Since Broadwell got effed up pretty bad, they might not ever drop Broadwell LGA CPUs at all, which was basically a refresh of Haswell with a die shrink and better graphics and lower power usage. Hence, Skylake coming in on time. 

Since Skylake is a new architecture, the TOCK to Broadwell's Tick, there is a new socket. DDR4 is probably part of the reason to switch to a new platform, so is the rumored removal of FIVR. But the fact they are moving to a new platform for their next cycle is nothing new or surprising, and has nothing to do with "Screwing customers and raking in cash". There is indication that Skylake will offer a greater performance jump than we have seen for the past few cycles, and there is even more evidence which shows Skylake in low power scenarios will be an absolute beast. New beast, new platform


Perhaps the new socket is required for DDR4 to work properly?

I suppose probably not if it's just one extra pin.

It's ridiculous how frequently Intel is switching sockets, but, then again, there's hardly a reason to upgrade from even an i5 2500k (for gaming). I'm don't see myself upgrading from my 3570k for at least another 2 years.

Apparently DDR3L will work fine and the option is up to the Mobo makers to choose which they offer. 

Incremental changes, not sure if anyone upgrades every year. Heck if you have that type of money you make the jump to 2011 and the Intel-E line up and be done with the consumer grade kit. I always liked the AMD approach with some cross platform compatibility. As we all know by now AMD is playing a different game these days. The only thing I don't like about the Intel annual change is it become tricky to get the older stuff. I would love it if they continued to make the older CPU's for a bit longer so there was more of an overlap. Several years back you could build a system and then in a year or so buy a new CPU on the same platform. These days you are looking at a whole new rig, three years or five, wouldn't want to wait longer than five. 

Right but Haswell/Broadwell on Z87/Z97 ect were not designed for forward compatibility with DDR4. Just because the memory controller is in the CPU, doesn't mean retro-fitting it into an existing platform is an easy engineering fit.

But that is not the only change, as I stated before. Removing FIVR from the chip completely destroys any chance of backward compatibility.

This is just because they wanne abbandon DDR3, thats where this whole thing is about.

Z170 chipset DDR4 support, and allmost no ipc improvement over current haswell cpu´s..

Skylake will be there Q1 2015, same time wenn intel will bring out broadwell for Z97. Broadwell will be for the enthusiast overclockers, And skylake, will only get locked cpu´s for 2015. unlocked skylake cpu´s will be there mid 2016.

this is just a planned time period for intel to abbandon DDR3. Skylake will NOT support DDR3 anymore

Sockets change because of the actual architecture change. When you change your architecture usually your pinout changes with it. Intel has their Tick-Tock design philosophy which pretty much forces them to create a major architecture revision every other generation, that architecture revision leads to a new pinout on the chip mostly because when you add new silicon (depending on what it is it may require a new pin(s) on the die to interface with the chipset or other parts on the board) or change around the location of different parts of the die you can't make the same run to same pins.  They are not trying to screw over customers, pinout changes are unavoidable and honestly expecting every single generation to use the same pinout is kinda silly. At least they're still keeping the same 115x cooler mounting layout

I don't have a problem with working on new designs or changing things, even often. But at this stage there have will be 4 main lines all going at once.

On top of this with each socket they bring 3-8 revisions (Z87, Z77, Z97, H81 so on). They all have a place and a purpose, it can just be a little overwhelming and confusing.

And for all its new ness if the motherboard makers wanted to they could make a board that offers zero over what we have now. It would be stupid to do so but they could.

Ticks and tocks are nice but do they have to release ever revision, update, refresh and number bump. Can't they wait it out till they have something actually impressive? Something that looks like an upgrade.

I am not out to dump in them, just playing devils advocate.

Apparently is does still. Backwards compatibility with DDR3 will continue at the option of the motherboard manufacturer to ease the transition.

Maybe we'll get motherboards like this but with DDR3 and DDR4 slots http://www.gigabyte.com/products/product-page.aspx?pid=3505#ov (Doubt it though. More capacity vs backwards compatibility)

I think the main problem we are facing right now is that Intel was way behind schedule (roadmap was thrown way off kilter as well) with Broadwell due to all the problems and delays with the 14nm node (Broadwell is the problem here, Skylake is on schedule as early 2013 roadmaps had skylake pinned for Q2-Q3 2015, this is also why the refresh exists) I think what Intel is trying to recover their roadmap and because of that Intel is doing strange things like the supposed lack of unlocked Skylake processors. But if everything goes smoothly with the 10nm node, things will be back to normal when Cannonlake arrives in 2016

What documentation have you read that indicates no IPC increase? Everything I've read thus far, and admittedly it is still early, indicate that Skylake will be much more like Sandybridge and much less like Haswell in terms of the evolution and performance increase of the core.

Also, I'm not sure if you were saying this, but the chipset for modern Intel platforms has virtually nothing to do with how many instructions per clock the CPU can deliver. 

This is true.

This is entirely a possibility 

as far as i know intel, i dont think this will happen.

Only if they broaght out skylake on the same platform as broadwell then yes. like in the early days, the transistion from DDR2 to DDR3 on socket 775 motherboards.

Since skylake will be launched on a complete new platform, and Haswell-E also does not have DDR3 backwards compatibillity, i highly doubt that skylake will have DDR3 backwards compatibiliity. im allmost 100% sure it wont.

Whats would otherwise be the sense for intel to realse broadwell on Z97 the same time, with skylake on a new platform then?


It's basically already confirmed.

 As for Broadwell, the problem was that Broadwell was SUPPOSED to be released on Z97 this past summer. They couldn't do it. So their roadmap is all screwed up. That is why they will either push back Skylake-K or skip broadwell-k CPUs all together. It has nothing to do with DDR3 and DDR4



no DDR3 for skylake, desktop cpu´s, im allmost 100% sure about that. give me any reliable source that proofs DDR3 backwards compatibillity for skylake?

Is there DDR3 backwards compatibillity for haswell-E? same architecture then Haswell?..... Ehhhh no right?

i stand by my statement No DDR3 backwards compatibility for skylake, untill i proven wrong

edit: wccftech is not a reliable source!
i have readed articles about imc backwards compatibillity, for skylake, but  they are all rumours. maybe for the mobile market cpu´s, but not for desktops, i simply cannot believe that. it would also be a very strange move from intel in my opinnion. because then it means that broadwell launch will be kinda pointless.

But we will see what happens.

Haswell-E is not a valid comparison. You are talking about a high end platform that exists with CPU's that have anywhere between 6 and 18 cores; an 18-core 32-thread server CAN ACTUALLY MAKE USE of all that memory bandwidth. A quad-core hypther-threaded desktop or laptop CPU just plainly can't. There is virtually no performance increase. With the price of DDR4 it makes ECONOMIC sense for dual-compatible IMCs, much like AMD did with Phenom II and Intel did with the G41 chipset.

If LEAKED OFFICIAL INTEL SLIDES and historical parity are not a valid enough demonstration, then I am sorry, you don't understand the driving forces in the industry at all.


And the Broadwell launch IS kind of pointless. With Intel's processor cycle of 12 months, a chip 6 months late IS pointless. But they spent alot of time and money creating it, so they are, in fact, going to make it.

eh I'll stick with my 4670K and my Z87 Gryphon MOBO until AMD's Zen Architecture. Intel's Monkey Business with the Sockets and Features have been all over the place lately.