http://www.overclock3d.net/articles/cpu_mainboard/intel_skylake-e_will_use_new_lga_3647_socket_and_use_6-channel_ddr4_memory/1
>LGA3647
>6 Channel Memory
>48 PCIE lanes (presumably for 16x + 16x +16x)
Looks like Broadwell-E will be the last generation of CPUs for LGA2011-3. What a crappy send off.
Also, it appears Intel is going somewhat retro and putting the CPU pins on the CPU itself, rather than in the socket. The socket only has holes in the pic.
LGA by definition means the socket holds the pins.
>>54834974
Also, here:
http://www.tweaktown.com/news/52336/intels-next-gen-skylake-processor-arrive-lga-3647-socket/index.html
>>54834991
Good point.
Wonder what we're looking at, then.
>>54835005
a prototype. The real mobo will probably hold the pins.
With how awful BL-E was. I'm not expecting much.
Please gives us a 128MB or 256MB L4 Cache, you cheap fucks.
>>54835091
You'd think with this much of a change (adding 1600+ pins) there would be a significant performance boost, but then again this IS Intel.
>>54834974
Holy fuck, look at all those pins!
>>54835135
What do you idiots even pretend you do that requires such huge increases in performance anyways?
>>54835170
I'm preparing for the day I can get a VR waifu, i'm assuming most AI processing will be done using GPGPU stuff as thats how most deep learning stuff is going right now, so more PCIe lanes for more GPGPU horsepower is gonna benefit me in the long run.
You can't kill my dreams
>>54835170
this.
I'm still using a C2D E8400 and it's fine. Obviously not for high-end gaming.
https://www.futureelectronics.com/en/Technologies/Product.aspx?ProductID=PE364727430021HFOXCONN5058446&IM=0
Hmm..
http://www.kitguru.net/components/cpu/anton-shilov/intels-28-core-xeon-skylake-cpus-to-support-6tb-of-dram-use-3467-pin-socket/
Wow, this is actually pretty big news, doesn't look like LGA 3647 will be for a consumer or even prosumer market, looks like it will be server and super computer markets only.
>>54835125
>Entire OS kernel and program binary in cache
>>54835295
Exactly, someone else fucking gets it. I was tempted to get the i7-5775C just because it has the 128MB L4 cache, but I do encoding so the 5820k made more sense.
If the 6800k had an L4 cache i;d have bought it in an instant, instead its just minor IPC improvement and I will need to wait for 2011-4 or this new LGA3647
>>54835274
Interesting. Sounds like this *isn't* Skylake-E, then.
>>54835170
>>54835324
Yeah rumors have already pointed at 2011v4 as a thing, so i wouldnt be surprised if LGA 3647 stays Xeon only.
>>54835135
Maybe they added more pins to future proof it so they don't have to introduce a new socket when they release processors that need more pins?
https://pbs.twimg.com/media/CUPUJo4VAAAtkiE.jpg
>>54835364
>5 DIMM slots
>>54835357
>intel
>future-proof socket
A socket chenge every year keeps goyim in fear!
Wait, what is 48 pcie lanes?
Up to 48 16x slots?
As in 24 way sli possible?
Oh dear god imagine the power draw.
>>54835418
Yeah but in the market segment they're targeting it may actually be beneficial for them to simply do chipset updates, it's not as if the average consumer will be buying one of these, it looks like even the low end chips will be $2k+.
The universities and governments and corporations who will be purchasing these are not looking for drop in upgradability anyway, when they upgrade their infrastructure they will be doing major upgrades anyway.
>>54835091
Broadwell in general was an awful abortion of an architecture, even before it was released to HEDT platform. Skylake has been very well received and has performed well in general. I actually expected Skylake-E to be pretty good, even with how bad the newest Broadwell-E chips are. Broadwell was always bad.
>>54835376
theres 6, he ripped one off
THANK YOU BASED INTEL
>>54835434
48 PCIe lanes as in x16 x16 x16 for triple SLI with full bandwidth or 6 way SLI with x8 bandwidth.
However it looks as if LGA 3647 will be for high end servers anyway so dual, quad and octo socket motherboards will be available, allowing for up to 384 PCIe lanes which would give you your 24-way x16 bandwidth SLI configuration.
Though obviously it would be useless for gaming as scaling would never work across that many GPUs and CPUs.
>>54835480
It would be useless for gaming.
But fucking cool to have anyway
>tfw 24 1080's
Hnng
>>54835766
Intel says they expect LGA3647 to allow for ExaFLOP performance by the end of this decade. The new UPI interconnects and the 100Gb/s omnipath fabric will allow for far greater scalability than ever before.
>>54834974
Maybe the extra lanes will be for xpoint support?
>>54834991
It is possible that when the locking mechanism is engaged the pins will rise up to interact with the CPU.
if that's the narrow version of the socket
how are oems going to fit the normal socket + 12 ram slots on to an atx motherboard?
>>54835911
there are 6 DIMMS in that picture,,,,
>>54835911
It's not meant for an ATX form factor so it doesn't particularly matter.
>>54835939
do you not see the other back of 6 dimms on the left?
>>54835964
bad eyes, thought that was a heatsink
Looks like the same socket as Xeon Phi (Knights Landing)
>>54836022
Now that's pretty interesting
>>54836022
What is the function of that tail? where do those connections lead?
>>54836078
Possibly a daughter board.
>>54836022
The frame around the socket even has the cutout for the extension. I think we are looking at the socket for the Xeon Phi.
>>54835170
When will your retarded ass realize 4K @ 120 requires serious CPU power to go along with GPU?
>>54834974
>Fumble CPU
>Socket irreparably ruined
>>54835219
>
I'm preparing for the day I can get a VR waifu, i'm assuming most AI processing will be done using GPGPU stuff as thats how most deep learning stuff is going right now, so more PCIe lanes for more GPGPU horsepower is gonna benefit me in the long run.
You better learn algorithms and some linear algebra, anon. Hardware is cheap and it can be bought when necessary.
Here is a library of useful books for you: libgen.io
>>54834974
A SOCKET A YEAR KEEPS THE GOYIM IN FEAR
>>54835005
Tiny pins.
a socket change a year keeps the goyim in fear
>>54836277
>>54836317
Loser
>>54836330
>>54836378
doesn't make you any less of a loser or idiot.
LGA 3647 isn't replacing LGA 2011 or LGA 1151.
>>54834974
that pic scares the shit out of me
i aint touching it with a 20 foot pole
>>54835418
I came to post this.
That socket looks huge, or my perspective is fucked,
>>54836378
So youre making the same joke someone made 4 minutes before that
Loser
>>54836586
I didn't even go to the bottom of the thread before I posted
>>54836078
that's where you put the anus
> you make bad joke
> no i didn
>https://libreboot.org/faq/#intelme
Enjoy paying to be fucked over again and again as a consumer.
>It specifies LGA - Land Grid Array, not PGA - Pin Grid Array.
>It's LGA, so either that socket is still not populated with pins (being a pre production sample, I assume) or the pins are so small that they don't look like other LGA pins under that light.
>>54836679
It has basically 150% of the pins of LGA2011, so what do you think the answer is?
>>54836687
>implying i asked a question
>>54836606
Sure buddy. And hitler is still alive.
>>54836739
You not believing me is your problem, I also don't really don't care desu
>>54836152
zen is fucking finished before it even came.
>>54836078
I'd assume that's just a test Connector. Won't make it to endproduction
fuck and i was just thinking of a sweet 6850K upgrade. wtf do I do now, wait?
>>54836782
>fucking finished before it even came.
>>54836804
Get a 6800k or 5820k, unless you NEED the extra PCIe lanes.
That or wait for Zen and hope it forces prices down a bit.
>>54836706
You gave two possible answers, but one was wrong.
>>54836831
WAIT WAIT WAIT AMD WAIT WAIT WAIT
wait for zen, wait for polaris, wait for vega, wait for bulldozer, wait for zambezi, wait for bobcat!
We got it right this time!!!!! WE PROMISSSS
Fucking AMDrones.
>>54836840
>both greentexts say LGA
how fucking retarded are you?
>>54836782
>>54836858
>Broadwell-E damage control squad
>>54836858
>Fucking AMDrones
You're talking to someone with a 5820k and soon to be GTX 1070.
I'm pretty damn far from an AMD drone, however if AMD can provide competition at the low end it will force Intel to lower prices somewhat, even at the high end. If AMD can release an 8 core CPU with ~10-15% lower IPC than intel, and they price it around $300-400. It will force the 6900k to ~$600 and the 6950x to ~$1000.
>>54836858
If you are waiting for PC hardware you are doing it wrong. PC is a dynamic platform, it changes with time and will forever. You can wait all you want, 1 year down the road there are upgrades available.
There is nothing wrong with the current chips/gpu's now, they will last you years. You cant wait past upgrading.
>>54836831
>6800k
fucking disgusting.
>>54836936
How is it any more disgusting than a 6850k? Ohh boy more PCIe lanes you'll never fucking use. And if you are NEEDING that many PCIe lanes you should probably going for a dual socket xeon or someshit anyway.
>>54836931
you can wait if it's 1-2 months, maybe 3. anything longer just means you will be perpetually waiting for new hardware.
>>54836899
Everyone is not expecting Zen to beat Broadwell-E, not by a long shot. The highest expectations we can have for it is that the 8-core chip will beat an i7-6850K and the 6-core will beat an i7-6700K. Zen can not compete core-to-core with Intel.
>>54837021
I literally just said that
>If AMD can release an 8 core CPU with ~10-15% lower IPC than intel
What do you think that line means if NOT exactly what you said?
>Zen can not compete core-to-core with Intel
I swear you faggots just like to argue over everything even when you're actually agreeing.
>>54837064
>I swear you faggots just like to argue over everything even when you're actually agreeing
yep. reading comprehension levels on this fucking site is abysmal.
>>54837064
What makes you think that AMD is going to sell Zen for less than their Intel counterpart.
Of course they're going to sell the same number of cores for the same price as their Intel counterparts.
Zen is a still-born AMD is pushing out in the hopes that people will overlook its many flaws, outdated performance, and backwards features (only two channels of DDR4? For an eight core CPU? REALLY?) because they've waited so long for it.
Zen is going to be a massive disappointment like everything AMD has made in the past 10 years and will continue the long AMD tradition of overhyping and underwhelming.
Cunt.
>>54835125
>128MB or 256MB L4 Cache
for what purpose ?
>>54837113
>What makes you think that AMD is going to sell Zen for less than their Intel counterpart.
the fact it wont be as fast.
>Of course they're going to sell the same number of cores for the same price as their Intel counterparts.
and no one will buy it if it's slower for the same price.
AMD is slower than Intel so they sell for a lower price. they have been doing this for the last 10 fucking years you imbecile.
>>54837113
>What makes you think that AMD is going to sell Zen for less than their Intel counterpart
We both know they can't release a $600+ SKU unless it had performance to match. I'm hesitant but with Jim Keller having done his magic on Zen, I am betting it wont be as big a flop as you seem to think it will. Intel has been focusing on efficiency and iGPU improvements for several generations, it's perfectly feasible AMD has made significant strides in IPC.
>>54837202
>they have been doing this for the last 10 fucking years you imbecile.
The FX-9590 started as a $350 SKU, you retard. It came out when Devil's Canyon launched.
>>54835125
Wouldn't there a delay/overhead issue ?
>>54836875
Apparently not retarded enough to communicate with someone as illiterate as you.
>>54837237
and then they lowered their prices according to performance, like they always do.
>>54837245
I mean sure compared to L1, L2, and L3. But a large L4 cache is still going to be much faster than RAM.
On die memory is always going to be fast as fuck because it's located on the die.
What's with all the Broadwell-E hate here? I didn't even know Intel had released it
>>54837219
>it's perfectly feasible AMD has made significant strides in IPC
>starting from shit IPC
Of course they're bound to make a giant leap in IPC, their Construction cores were stuck with early 2000s IPC. That's a slow as you can go.
But Zen's IPC leap is not enough to make it comparable with Haswell. AMD's underpaid engineers eeked out all of the performance gains they could from the lowest-hanging branches, of which there were many transitioning away from the failed Bulldozer architecture. It's that leap from Ivy Bridge performance to Kaby Lake that is going to be much more difficult, if not outright impossible, for AMD.
They will fail. They just don't have the money or time to beat Haswell, let alone Kaby Lake.
>>54837256
And they're still selling CPUs from 2013 and motherboards from 2011 with performance from 2007.
>>54837271
>What's with all the Broadwell-E hate here?
it's expensive. only reason
>>54837254
Try explaining what you're talking about then, it shouldnt be that hard since you're so smart...
>>54837282
>They just don't have the money or time
They just doubled their stock price and got a over $100M in cash from a chinese firm to license their x86 architecture. They are fine at the moment.
>>54834974
Sure glad I'm not getting a 6950X now
>>54837294
>financial equivalents of bandaids
>no long term fix on the horizon other than Zen
>not even AMD is certain if Zen will pay out for them in the long term
AMD is on their last legs. Rising stock prices are not always indications of financial recovery, nor are relatively minor revenue streams that have a significant downside in creating competition in the future. That was a patch to keep AMD's inflatable dinghy afloat for a few more months in the hope that a lifeboat made from Zen sales will arrive and bring them to dry land. That lifeboat could never arrive, or arrive too late.
>>54835888
Holy shit, that would be a godsend for many.
>>54837336
Yes licensing out their x86 architecture so they get royalty fees for every single CPU ever produced by the chinese firm, forever is a terrible business move.
Especially when you look at the fact high end Intel CPUs are banned from being exported to China, so the chinese government has been looking for a chinese fab to start making large x86 CPUs for Domestic chinese use. So one of the largest economies in the world is KNOWN to be on the look out for high end x86 architecture AND a chinese firm is offering to purchase your x86 license with the understanding you get $ for every CPU made.
Yeah terrible choice for AMD, they'll probably go bankrupt. It's not like china doesn't invest Billions per year into their defense research.
>>54837285
>>54837271
It can't overclock for shit. Temps go 100 C housefire over 4.4 GHz. You got to be mad to go for Broadwell-E instead of sticking to Haswell-E. If you want MOAR COARS just fucking go for Xeon
>>54835888
>>54837359
Seems like another thing that can break.
Honestly its not that hard to put a cpu in the socket without bending shit.
>>54837393
>they get royalty fees for every single CPU ever produced by the chinese firm, forever is a terrible business move
They just made a competitor in the near future who will, inevitably, break that contract and not pay AMD a single cent for future derivations based on their work. There is no guarantee that those firms will even hold up their end of the bargain in the short term, since the Chinese government has every ability to force AMD out of the contract or cause inconvenience to their facilities in China.
The Chinese are known to do this even to the biggest and largest Western firms. There was a car manufacturer (a joint GM venture) that was plagued with worker strikes, production delays, and "accidents" when they tried to take a Chinese car maker to court for blatant counterfeiting and copying their designs outright. Unfortunately for them, that car maker was owned by some very influential party members in Beijing. Not only did they lose in court, but they also were forced to stay silent on the "unfortunate events" that occurred during the lawsuit under the threat of government intervention.
AMD will suffer because of this, and those Chinese firms know that AMD can't do shit about it in the future, when the Chinese courts eventually dictate global copyright laws.
>>54837456
This China crooked as fuck. They wont need AMD more than to learn from.
>>54837393
do you have a sauce for high end intel cpus being banned in china? it sounds interesting
>>54837473
AMD are going to get their facilities in China forcefully taken over if they try to pursue any legal options in court against the people they licensed their design to. Most of those facilities are a core part of their silicon design and production validation strategy that China would love to see under 100% Chinese control.
AMD shot themselves in the foot big-time for some chump change.
>>54837488
>http://wccftech.com/us-government-bans-intel-nvidia-amd-chips-china/
Been happening for years
>new socket
HAHAHAHA
HAHAHAHAHA
HAHAHAHAHAHAHAHAHAHA
>>54837456
>>54837473
>>54837494
Wow, all this speculation because CHINA IS TEH EVIL
They have shit tons of money and AMD is a cheap bitch, $20 says the honor the deal and just pay them whatever small tithe they owe.
>>54837456
If they are not giving a fuck about that nothing should have stopped chinese to copy x86-64 until now, it's not exactly secret or anything. They have their own microprocessor architecture with MIPS ISA
>>54837507
>all this speculation
We've seen this happen time and time again since the early 2000s. China wants everything, and they'll take everything given the chance. It is their right to.
>>54837523
They did try to copy x86-64 designs, but not to any great success. They made usable copies for non-high-end compute applications, but they're in sore need of a competitive, powerful CPU for their military, scientific, and economic purposes. AMD has given them the keys to reach parity with modern x86-64 processors and leap ahead by several years.
>>54837560
Im not saying china couldn't fuck over AMD, but more than likely they want to make AMD emboldened to them, if AMD relies on china to survive, they're far more likely to either help with future designs, or be more inclined to do IP sharing and similar joint ventures.
>>54837560
>but not to any great success.
Or they realized the dumb fuckery called CISC doesn't worth the legal foreign sanctions.
> but they're in sore need of a competitive, powerful CPU for their military, scientific, and economic purpose
They can just apply to be a part of OpenPOWER if they need that.
>modern x86-64 processors and leap ahead by several years.
Nobody is leaping in x86-64 even Intel.
>>54837604
China wants competitive CPU manufacturers. AMD has offered themselves to them. China will want AMD under their control.
>>54837624
>They can just apply to be a part of OpenPOWER if they need that
Didn't Chinese research institutions already make overtures to join, only to get shut down or sidelined by their own government?
>>54837628
that means not fucking them over.
>>54837640
May be they are not in such a sore need
>>54835170
Emulation.
2500K OC'd has done pretty well by me for everything up to Wii but I will undoubtedly need more power for PS3 and 360 emulation and consoles beyond that.
>>54837681
Bullshit, they've been buying every high end chip they can easily get their hands on for a decade +. China.
The recently unveiled Tianhe-2 in China is the most powerful publicly announced super computer in the world.
>>54836554
MCM unit - multi chip module, two dies mated to a single socket board. The line through the contacts is a dead giveaway.
If this really will be for Xeon products we're going to either see similar core counts but larger space per core or a doubling-ish of total cores per chip.
>>54837740
They also not looking like they sidelined OpenPOWER.
http://openpowerfoundation.org/membership/current-members/#
At least 5 Chinese companies
http://openpowerfoundation.org/openpower-china-summit-2015/
summit in china
http://www.forbes.com/sites/patrickmoorhead/2015/09/06/ibms-openpower-a-lot-has-changed-in-two-years/#75da3f1312c0
>Many of these developments show IBM and OpenPOWER’s focus on China. China’s government-led operations in finance, military, research, healthcare and education have a government mandate to have all new systems on 100% controlled hardware, software and middleware. This is referred sometimes as “China 2016” in the industry and OpenPOWER wants the datacenter side of it to be based on their technology.
http://www.nextplatform.com/2016/02/17/china-lays-the-chip-foundation-for-its-next-platform/
AMD's involment look peanuts compared to others
Looks like shit. I'll be rocking my X99 system for a long time to come.
>>54839114
>Looks like shit
Literally how?
It's hand over fist better than X99 in every way, but it's also not meant as the successor to X99, it's in a higher tier entirely.
Broadwell-E confirmed for excellent overclocker. AMD FUD is shit...again.
Source: http://wccftech.com/intel-core-i76950x-overclocked-whopping-57ghz-details/
>>54836068
>>54836078
>>54836089
>>54836125
>>54836802
>yfw that's the connector for XPoint
>>54839345
>PCWorld has reported that during an overclocking event at the Computex 2016 trade show in Taipei on Tuesday, large tanks of liquid nitrogen stood on the ready to help overclockers keep the processors cool and to achieve their new speeds on the deca-core chip.
Excellent it is. You need liquid nitrogen to cool the housefire.
>>54840192
Damn, with clocks that high, an A10 might be able to beat a Hasmeme i5.
>>54839451
>x-point
This is vaporware.
>>54836858
polaris is already here you dunce
>>54837507
>I know nothing about China.
They take pride in ripping people off, they have no respect for honest businessmen, Scamming is their culture. What other country has problems with people selling fake eggs and refining cooking oil from sewers?
>>54842281
I took 4 years of chinese in university so I guess I'm retarded.
we'll see in a few years, but i'm betting china will keep the deal alive.
>>54842397
>I took 4 years of chinese in university
And I lived 15+ years near a Chinese community populated by the worst sort of mainlanders China has to offer. These people are worse than scum, I'd even go as far as to call them uber-Jews to Jew them all.
4 years of sheltered, carefully-selected and biased exposure versus over a decade and a half of living around actual Chinese people, working with Chinese-run businesses, and dealing with Chinese family issues that spill out from those businesses.
>>54843048
Still just speculation, there are example of China honoring deals as much as their are examples of them breaking deals. Only time will tell for sure.
Hell even if 5 years go by I bet you'll still be telling me eventually china will fuck them over so it's not really an argument i'm going to win with "you people".
>>54843117
And I bet you'll defend the Chinese when they suddenly decide that the license they signed with AMD was invalid because of bureaucratic reasons and no longer had to pay royalties because they changed two very minor components, which somehow constitutes into "an original and unique" design which invalidates any lawsuits AMD would try to pursue to recover the money they were owed.
Corruption is systemic in China. I wouldn't even be surprised if they tried exporting those licensed CPUs and sued AMD for "patent infringement" in that country to cement their position in the market. It's happened before.
>>54843048
Welcome to ultra-capitalism bitch
>>54843303
I've got two of those families deported in the last five years (or least made them go into hiding). Welcome to America, bitch.
>>54843303
thank mr capitalism
>>54836152
Looking at the key notches, it would go the other way around.
>>54843427
Obviously, its LGA, the picture I had available was upside, i can't exactly flip a 2D image around.
I'm going to be on the look out for an HP Z900 to use this socket for it's motherboard. Shit is going to be cash money.
>>54843927
>>54836129
when will your retarded ass realize 4K@120Hz is a meme
>>54835170
having a bigger dick than you