Intel’s 8-core (Kaby/Coffee) Lake chip spotted

videocardz.com/newz/intels-8-core-kaby-coffee-lake-chip-spotted

wccftech.com/intel-8-core-16-thread-coffee-lake-8th-gen-cpu-spotted/

Attached: jim-keller-amd-intel-raja-koduri-ensemble.jpg (500x500, 43K)

Oy vey, time for another upgrade!

>new socket

Does it come with a soldered IHS?

>The chip was tested on the Kaby Lake client system test platform which confirms another critical detail about this processor which is that this chip will be compatible with current generation motherboards

cont.

>which is a lie

>2,6 GHz

Attached: th?id=OIP.CmFo7PVXl-C0GsfdpVWOTwHaHQ&pid=Api.jpg (474x464, 19K)

Coffeelake leaks had it running on Z270, those are internal validation boards with different firmware.

don't worry, now with Turbo Boost 4.0 (tm) it'll be able to boost a whole thread to 5.8GHz for 0.5 seconds!

Oh boy, another housefire from Intel!

IDGAF about 8c KL housefires, but OP pic is actually worth a thought regarding future products.
Which is stronger? Based Shitwrecker’s ability to clean house and right sinking ships, or “Designated” Raja’s ability to fuck up multiple architectures in a row?
The Unstoppable Force or the Unflushable Object?

raja was sabotaging RTG, this is not a metric for what he actually can do

seeing such low clocks i know its gonna be another hot chip.

what did you expect? same node, same architecture

their arch design does not scale well passed 6 cores, we've already seen that

This is gonna burn down some houses

I should expect nothing but horrible clock speeds from now on?

His best best days were pre-SIMT, and his pinnacle achievement was probably the fixed-func R300 ca. 2001.

Also,
> EVER trusting a POO to do anything but scheme and ladder-climb

if intel keeps doing this shit, yeah
there's no magic about these things

1 and 3 months in your 7700k and it's already Pentium tier cpu.
the greatest of all is that amd leads and intel follows.
their next move is multi-chip designs.

1 year and 3 months*

> 2.6GHz base

>their next move is multi-chip designs.
Like the 1950x

Attached: 000000000009.jpg (1280x878, 293K)

yeah, that's the projection for future xeons.
That's what IBM has been doing since the 90s, that's what AMD brought now to the consumers and apply it to the GPUs in the next releases and that's how Intel is pressured to follow the same path.
when your competitor gets a 64 thread chip with 1/5 of the cost and r&d of your 28core chip, then you know that it's a time for change.
That's why Kim Jeller joined Intel.

>amd leads and intel follows
well, yeah

In this case, intel would follow but they currently have a 10nm-sized dildo shoved directly up their asses.

Still 14nm I presume.

>0.5ns
Fixed.

It would be better than 10nm at this point.

AYYMD IS FINISHED & BANKRUPT

Mandatory 5 meters InRow cooling to avoid a housefire.

When amd releases their first 7nm cpu, intel will jump in and release a 10nm cpu no matter what's the cost.... But they will release something like 5 of them, just like they did with 5775

jim keller looks like tony hawk with a bit of linus torvalds

Attached: iCONICLA-TONYHAWK-FULL_1920x1080_330360.jpg (1920x1080, 96K)

>2.6GHz

Attached: IMG_1510.jpg (2048x1365, 461K)

Cost is only one dimension. Time is the other. There is a physical limit to how fast you can run a lot through a fab. It takes months.

DELID THIS
E
L
I
D

>2.6GHz
>probably gonna run at 120W+
>$500+
>still gonna be using cum in the IHS instead of solder
Some people are going to figuratively burn their money so they can burn down their houses.

This is what I was wondering, does bing bus even work for 8 cores or will they have to use mesh for this abomination?

Are you retarded or something? Mesh didn't exist before last year and Intel released their first desktop octa-core in 2014. You do the maths.

So all those 8 core plus xeon units before 2017 are using bingbus? jesus christ

Thanks faget can't unsee.

Yields are going to be crapptastic like 8-core Broadwell-E and Haswell-E chips.

They aren't going able to boost up to the turbo speed 8700K without high-end cooling. The dual-channel DDR4 is going to be a bottleneck for prosumer workloads.

This is just Intel trying to keep up appearances in the core-count race in the desktop market.

They should honestly just wait for Ice Lake to release a desktop-tier eight-core SKU.

Intel is going end-up cannibalizing its Skylake-X LCC SKUs for no good reason (better yields, superior platform for prosumer-stuff).

It is kinda funny that Intel is losing its shit over Ryzen and Ryzen+ when AMD has jack mindshare in the desktop/laptop arena.

Friendly reminder that Intel's previous two 8 core bingbus CPUs were $1000. Their yields are shit on higher core bingbus, AMD competition isn't the only reason Skylake-X is drastically cheaper than Broadwell-E. If Intel didn't switch to mesh then HEDT would have been a bloodbath last summer.

Yeah but they'll be able to send reviewers golden samples so they can OC the thing to 5ghz to run benchmarks on it. They just won't tell you. They spent $400 on a motherboard with beefy enough VRMs and $800 on a custom loop to cool the fucking thing.

>new chipset

Attached: All aboard the bingbus.png (5098x1500, 2.29M)

>unflushable object

lol

keller is on the soc team, not the core team

SHUT UP THAT LOOK GAY

>(Kebab/Coffin) Lake

Based Intel

AMD is finished

intel already has the designs, they are waiting on the process, I would be fucking shocked if they aren't at least saving the test chips

once amd goes ahead with its 7nm, intel will likely ditch the igpu or make it a modular part, something they may have already did, and either use amd, or use their own igpu, mcm it, and sell it regardless of loss.

they cat let amd have 7nm without something against it, because a 5ghz 8 core base clock cpu is going to fuck intel so goddamn hard even the fanboys won't be able to deny reality anymore if they don't put even a token effort into a 10nm part in the launch window for amd. they have a total of 1 maybe 2 quarters before they have to start scaling the process production up.

the overwhelming majority of cpus in the pro market are amd threadrippers, intels token efforts to compete are noted and found wanting. 8 core consumer isn't going to do shit to intel's hedt market for one reason, you don't but intel sum 16 cores

intel has had an amd product cycle to start soldering their cpus, but they wont, fucking why? thermal jiz was great money savings when they had no competition, fucking why?

because they still believe that they're 3 years ahead of AMD

soldered IHS on cooper plate ?
why do you want thermal paste on cooper IHS ?
help they want an aluminum IHS !
oy vey remember to delid and apply your own thermal paste !
>muh shekles 20 km.jpg

They already dumped the iGPU on the 10nm mobile CPU (which is complete shit BTW).

Do we know that? I mean they might just deactivated it, while its still part of the package.

>One cycle
Fixed

ringbus is fine on up to 8 cores.

For the higher core counts, they should be using 10nm. SHOULD be, but probably won't be.
They need the lower TDP. Their 14nm++++ process is just too hot.

but their 10nm process is absolute trash, won't get better than their 14nm stuff until the third gen

Their 10nm is lower absolute performance, but it's better perf/watt at lower clocks.

So what, you can hack older 1151 mobos to run Coffee Lake and yet there's no official support.

Is there any law I don't know about that forces intel to tell the truth about what Keller will be doing to the public? No? I thought so.

>they cat let amd have 7nm without something against it, because a 5ghz 8 core base clock cpu is going to fuck intel
There's no guarantee that the 7nm will actually improve the clock at all for AMD, it might very well be more efficiency focused.

I just bought a 8700k and they come out with this shit? the fuck? Will I be refunded the difference?

kek

>AMD Literally said 7nm would enable 5ghz processors

They didn't. GloFo did.
Zen 2 will have higher IPC and clock higher, but wether it actually reaches 5Ghz is pure guesswork and wishful thinking.

Why would they go and do something like that, and deprive everyone the thrill of delidding/voiding the warranty?

6 cores were hot. 8 cores are going to be nuclear

Which is far more then what what you would spend on a decent X299 + i7-7820X setup. It only would win at "gayming" shit while i7-7820X will pull ahead in prosumer shit (memory bandwidth does matter). Mesh-toplogy is superior for massively multi-threaded loads (far less resource contention and lower latency).

The only reason that Ring-bus topology is faster at gayming shit is because one or two CPU-cores has full access to that massive L3 cache block while mesh each core has it own dedicated L3 cache pool.

Our exclusive MultiCore Enhancement™ technology will help you overclock your CPU to 4.5 GHz for impressive 15 seconds at modest 300W TDP. Make sure to buy our brand new 1151v3 motherboard, goy.

They used multiple rings but those weren't rings of 4 nodes. A node may be a memory controller, pcie controller, core. They used rings of different size but with each node added the interconnection latencies increase, and adding more rings does too by introducing interring communication, much like in ryzen, but a bit faster.
Intel made rings as large as 16 nodes which is plenty for a 8-core CPU. 8-core 1151 cpu should have 10-node ring. It will indeed be slower than 6-node rings used in older 4-core cpus but nothing exceeding what Intel was doing before.

Attached: xeon_e5_v4_hcc_rings.jpg (1200x706, 183K)

>tfw Jow Forums cucks have been saying the same thing about coffee lake less that a year ago
>coffee lake has obliterated Ryzen
>intel sales soared, as well as their share price and market share %

the history is repeating itself

I can already see Zen 2 losing to Intel after a few months in 2019.
Intel's technological, intellectual and financial supremacy isn't something AMD can challenge.