Intel fabs on suicide watch

What's the average failure rate of a CPU die?
It's probably 97% at this point, everything is so fucking small now.

>>59707567
failure should be about the same % of surface area as before, meaning smaller the chips more yield.

WHY IS ITHIS HAPPENSING

>>59707266
10nm never

anything below 14nm is a meme on silicon

>>59707581
>failure should be about the same % of surface area as before
pretty sure that's not how it works. smaller structures, less room for error.

>>59707567
What? Smaller dies have less defects than larger ones.
Far smaller.

>>59707266
Welp that'll push Icelake to 2019 then, they were saying end of 2018 before.

>>59708022
You should consider the Intel roadmap dead, it's gonna be completely changed now.

1. Cannonlake for servers has been canceled.
2. Cannonlake for desktops is probably not gonna even happen, if it does it's gonna be 14nm, mobile only 10nm

>>59708048
Cannonlake is just die shrunk Skylake, so 14nm Cannonlake will just be Skylake.

Coffeelake

JUST
RELEASE
SKYLAKE
AGAIN

>>59707855
Because Moores law will be over soon and intel are cheap shits

You can tell they're desperate to get their desktop shit to 10nm so they can squeeze out a few extra MHz before hitting the TDP wall.

At least we will get a new socket :^)

>>59708143
|
|>
|x
|
|

Holy shit, at this rate AMD will release Pinnacle Ridge/zen2 before Intel.

Don't tell me even Coffelake is now Q4?

>>59708174
>ayymd moor law live

>>59707906
He's talking about the manufacturing process, genius, not the die size.

File: 1390758529558.jpg (82KB, 500x532px) Image search: [Google]

82KB, 500x532px

>>59707266
Ah shit, now we can't get yet ANOTHER new socket this year :(

>>59708174
Coffeelake is 250W TDP, they're having issues getting non-exploding motherboards ready

>>59707266
>Wait™
Justice being served to jewtel

>>59708219
The funny thing is it'll actually use that much power at around 4.6ghz

They did say 15% more performance though, which points to a 3.5ghz range hexcore. It'll basically be a 6800K for mainstream.

File: 1490399409123.png (973KB, 801x1500px) Image search: [Google]

973KB, 801x1500px

>>59708219
Mandatory

File: intel designed final solution.jpg (28KB, 550x550px) Image search: [Google]

28KB, 550x550px

>>59708219
>Coffeelake is 250W TDP
Hitler's getting hard in the afterlife right now.

>>59707266
Hahahaha Intel waitdrones on suicide watch.

>>59707895
That's not how it works, kid.

File: 1467129039190.jpg (165KB, 1024x656px) Image search: [Google]

165KB, 1024x656px

Intel is giving up. This is a bad sign actually. If Intel gives up, that means everyone else is going to give up.

File: 1489778876093.jpg (33KB, 640x621px) Image search: [Google]

33KB, 640x621px

So there's 10nm, 7nm then finally 5nm. Apparently 5nm and below is incredibly difficult due to quantum tunnelling causing electrons to fuck about.

We need to move away from silicon

>>59708478
Move to where?

>>59708527
Just use some Chinese transistor that you can buy from the store.
Let's bring back the room size computers.

Jewtel should sell HEDT CPUs for competitive price instead of mainstream 4 core babby CPUs.

File: Specs-9570.jpg (34KB, 424x295px) Image search: [Google]

34KB, 424x295px

>>59708354
wrong pic

>>59707266
>low power mobile Cannonlake launches in 2018
>Cannonlake server parts canceled
>Cannonlake desktop parts never
>a year later they'll say they have a new even better process called "10nm+"
>its the exact same thing but now yields are up to 10%
>good little goys will still pretend like intel has some foundry advantage because pointless back end area scaling metrics
>Krzanich will keep lying to investors, saying there is no problem
>next investor call says that the foundry business is only a quarter behind schedule
>meanwhile marketing has shifted roadmaps by a full year
>fanboys will still say nothing is wrong

>>59708629

>250W > 220W

Wow, who knew?

>>59708661
>250W
>no sources
nice try drone
meanwhile amd released this shit at $900 w/o wc and barely matched 2500k with constant fear of house fire

>>59707895
Wafers had small error, if one error are inside one Chip, need drop chip, if you had 50 big chips for wafer and had 20 errors you will need drop 20 chips, if you had 1000 small chips for wafer and had 20 error you will need drop 20 chips, this is why SoC for smartphone are great business and big GPU are expensive.

>>59708478
I'd just carve out the x86 instruction decoding into software and make the thing a proper RISC.

>>59707890
jesus christ

>>59708211
We're still getting this abomination on a new socket https://videocardz.com/67610/intel-core-i7-7740k-and-asrock-x299-professional-gaming-i7-spotted

File: 1488636448506.jpg (12KB, 199x216px) Image search: [Google]

12KB, 199x216px

>>59707266
> Intel has a goal of delivering a minimum of 15 percent improvement in chip performance on an annual basis.
Really activates those almonds.

>>59708763
thanks for sharing, pajeet

>>59708904
>15 percent improvement in chip performance on an annual basis.

Mobile only, in sysmark.

>>59708271
What's the point if it can't even compete with Zen+ 6 cores at the same TDP?

>>59708562
If you make a processor too big, desync becomes a problem. Light travels a foot in one nanosecond (one clock cycle at 1ghz). Electrical signals travel a bit slower than light, and IDK exactly how far the maximum distance these signals have to travel in one clock cycle is.

>>59708527
>Move to where?
More cores.

>>59708232
Intel made everyone wait for faster CPUs. Made everyone wait to be able to afford their 60% profit margin monopoly abusing prices. Made everyone wait for affordable 6+ cores. They deserve it.

>>59708629
>let's dredge up the past and compare it to the present!

Yes, Intel shills are already looking back to the good ol' days when AMD wasn't looking so hot and they could coast on by with 60% profit margins and pitiful upgrades forever.

>>59708982
right now the consumer is funding Intel's multi-billion dollar war on the laws of physics

eventually we'll have 1nm processors running at 500w tdp. 64 cores, 7ghz out of the box. and shit will still lag and be slow and you'll have to play on medium.

>>59709010

it's a noble war.

>>59709010
Let's lend them some unobtanium.

C, Go, Javascript

thats literally all you need

File: le_happy_merchantess.jpg (11KB, 213x250px) Image search: [Google]

11KB, 213x250px

>>59708982

Intel could just do a 6 core version of the 7700k and/or double the caches,

but...

>>59708412
They will remain the biggest players by sabotaging games together with nvidia and using that to sell their CPUs.

>>59708937
Sounds like bullshit. multi-ghz signals travel between cpu and gpu

>>59708300

That would be much funnier had AMD not followed Athlon64 with bulldozer.

>>59709108
Two devices communicating with each other =/= the same device getting out of sync with its own parts.

>>59708527

maybe Gallium Arsenide ?sp?

>>59709068
put some OC on it and you've got that 250w+ powerdraw

>>59709068
They can't do that for mainstream with muh iGPU.

>>59708821

The 90's called. Transmeta failed.

Heck even the DEC Alpha could not do translation fast enough to make a switch worthwhile, and that thing was a fucking beast.

>>59709108

In pcie the clock is encoded onto the data signals. Each lane is received independently. There is no real shared time base between the GPU and the CPU.

That rx/tx path is very different than what you have with synchronous logic where you share the same clock signal.

>>59709030
pretty much this, we need software optimization more than anything

>>59708584

If they were to do that, they would be way cheaper than server counterpart. In order to keep people paying $$$$ for Xeon cpu, they need to sell shitty consumer grade cpu.

>>59708635
>Krzanich
Krzanich is an accountant, he can't see the technological gap to close, only the numbers.

>>59707266
>Those 8th Generation Core laptops may be more attractive to customers. The first 10-nm Cannonlake chips will be slower than 14-nm 8th Generation Core processors. Intel acknowledged the speeds during the manufacturing event, with a chart showing 10-nm chips catching up with 14-nm chip performance in one to two years.

WHAT IN THE HAHAHAHAHAH WOW

This company has completely lost its marbles.

>>59708125
>be over soon

it is over for years now

>>59709514
will MUH GIGAHURTZ save intel?

>>59708763
>overlay errors don't exist
>correct overlay has nothing to do with structure size
ooooooooookay.....

How long before Intel is forced to sell off their fabs?

I don't want to overclcock my rig, I am seriously thinking of sending the cpu and motherboard back and buying an i7 7700k rig. I am not an engineer nor did I pay 1300 Canadian dollars to beta test a platform that AMD rushed out the door before it's ready!

>>59709694
you copy&pasted the wrong post in the wrong thread. No rupees for you today.