Slashdot Mirror
Intel's Tick-Tock Cycle Skips a Beat
New submitter Ramze writes: Several outlets are reporting on Intel's confirmation that it will make three generations of 14nm processors, delaying the switch to 10nm. The planned 14nm Kaby Lake processor marks the first time Intel has skipped the "tick" of a die shrink on its regular "tick/tock" cycle. Production of Cannonlake processors on 10nm has been pushed back to the second half of 2017 — likely due to manufacturing difficulties. Intel reported earlier this year that it may have to switch away from silicon to exotic materials such as indium gallium arsenide to make the next shrink to 7nm.
Hi there, young entrepreneur! Welcome to the Indium Gallium Arsenide Valley! :)
I know - the transistor count should have enabled us to build neural nets to filter out inane AC comments by now.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Look at Intel's recent earning and revenues. Business is so bad it doesn't justify investing money in a new engineering shrink.
love is just extroverted narcissism
ACTUALLY, it is getting us into a fickle afterlife & fantasy world where the sum of your desires make the walls of your sepulchre.
MAYBE you are already in your afterlife, eternal and alone with your base ego.
All rites reversed 2010
Silicon-on-Silicon CMOS: non-toxic.
SOD-CMOS: non-toxic.
Indium-Gallium-Arsenide: toxic heavy metals combined with toxic metalloids. Holy fuck. If ROHS doesn't lift their ban on lead after this, they've got their heads up their asses. Mercury is a little worse.
Support my political activism on Patreon.
We did have a leap second this year, and next year is a leap year, so there's really nothing to see here.
This just in from Intel Markteting:
"Vee have vays of making you tock!"
Solving Unix problems since 1989...
Smaller transistors result in bloated and sloppy code. That's what's been happening for 20 years now. When I think about how much faster (at a hardware level) my current computer is relative to my old 486SX 33MHz, and how I still have to spend time waiting for the damn thing to catch up to what I'm trying to do, I just have to shake my head. At this point, I have no reason to not believe that this is the way it will be forever. All gains made in hardware speed will be wasted by software bloat.
wtf
Woah.. a computer technology article not submitted by Mojokid.. Shocking..
Back when I actually cared enough to be a signed in user I would have my prefs treat a Funny comment downward. Far. Funny comments are or were the scourge of /. - now I just don't care. /. is the scourge of /. Now I use alterslash as another layer of filtering and post as AC.
unfortunately this moron's comments were captured by alterslash...
I know - the transistor count should have enabled us to build neural nets to filter out inane AC comments by now.
Hardly. The brain has ~100 billion neurons and 100-500 trillion synapses, of which the latter is closest to a transistor. Leading CPU/GPUs have 5-9 billion transistors or less than 0.01% of that. Remember, we are approaching atom size but only in an extremely thin 2D slice. Current processors are about 100k*100k transistors big, if we could have the same density in three dimensions we'd have 100k^3 = 1000 trillion transistors in a 2.5 cm cube, comparable or even beyond the brain in density. I wouldn't try cooling it though as 100000k*100-250W power consumption means it'd consume >10 MW.
The brain only operates at ~100 Hz though, at least that's the rate synapses pass signals from neuron to neuron but it's not entirely clearly if that's equivalent to a CPU cycle or a network connection. Probably more like the latter as it seems each neuron has a form of local storage and programming of what to do. Like if you're looking at a picture and trying to determine if it's a cat people can reliably do that in half a second or 50 cycles which indicates quite a lot more processing per cycle and the neuron firing is more of a sub-result of a distributed process. So there's a lot more brain, though it runs much slower.
Live today, because you never know what tomorrow brings
I *knew* I shouldn't have partnered with Intel on my line of computer-chip lollipops!
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Well, thanks for making the world a better place by making your comments invisible (except when I'm moderating, anyway).
Hey, why so sensitive? IT WAS A JOKE!!
Oh, right. Sorry. My mistake. Humor ruins everything.
personally, i make a habit of blaming developers
By and large, language is a tool for concealing the truth. -- George Carlin
Why build neural networks when you can get the wild-type ones to do the same work in exchange for 'mod points'? Markedly more cost effective.
My very layman understanding of the brain is because of the way the synapses interact, they also act as part of the processing. You may think of the brain as an ASIC special purpose CPU, but which ASIC that is used is determined by the input. As input triggers a neuron to do something, that neuron may trigger several other neurons in different orders, timings, and strengths. Each of these many combinations causes a cascade of triggering other such neurons in the same way. After only "50 cycles", a cat is identified. Millions of neurons and billions of synapses may have been used during this time.
Using the same neurons in different timings and strengths results in different answers. It's like a hybrid of CPUs, memory, and routing. I was also under the impression that neurons use a hybrid of analog and digital processing, and even a possibility of quantum.
Because which input synapses and the signal strength of those synapses affect how the neuron sends data to the output synapses, you could say the neurons also have routing inside of themselves.
Pretend you have a neuron with 2500 input and 2500 output synapses. Lets assume each synapse has some 8 states it can be in. Not sure if it's even remotely close, but lets go with it. The neuron can have 8^2500 different inputs resulting in a probably constant-time O(1) result in output. That one neuron. If a single neutron can translate an 8^2500 number of inputs into a useful output of a partial amount of data processed, it doesn't take too many iterations to process data.
Synapse is NOT like a transistor, it is a schmitt-trigger circuit with hysteresis. Multiply your transistor count by order of magnitude or more per neuron
It certainly doesn't make the task compact or cheap(and appears to make it much harder to directly duplicate with either transistors or the inspired-by-but-not-models-of version of 'neurons' that 'neural network' usually implies in the context of computers); but the one nice thing about the very low 'clock rate' of the brain is that it suggests that however it does what it does; it can tolerate suffering fairly high latency per unit of distance between elements.
If you can tolerate latency, you at least have the option of buying more racks to compensate for what you can't achieve with available miniaturization and integration. If you can't, you run into relatively painful constraints on size. At the speed of light your hypothetical 3GHz processor is going to be waiting ~10clocks/meter for anything it needs from elsewhere in the system; and any practically available arrangement is going to be slower than that. If the brain were dependent on low latency, inability to replicate its density would(at best) mean being limited to replicating it slower than real time, possibly much slower. If it is relatively tolerant, replicating it at lower density might be horrifically expensive; but at least potentially doable with the ability to fabricate only rather small chunks of very high density.
" The planned 14nm Kaby Lake processor marks the first time Intel has skipped the "tick" of a die shrink on its regular "tick/tock" cycle."
They've not skipped anything - they're adding a second 14nm tock and delaying the Cannonlake 10nm "tick" by almost a year, in the same way they delayed the Broadwell 14nm "tick" by a year and filled in the gap with 22nm Haswell Refresh.
Given they've not called it Skylake Refresh, we can only assume Intel plan a more substantial change in architecture compared to the clock bumps and voltage regulation improvements we saw in Haswell Refresh.
You say ta-may-toh, I say ta-mah-toh (they say, "holy-fuck-this-die-shrink-is-a-bitch")
Is it just my observation, or are there way too many stupid people in the world?
I am not sure why there is anything more on wasted on desktop processors given the last 6-7 years of only ~10 percent gains? We are expecting almost zero improvement in desktop performance with Skylake over the 4790K processors, and barely a power reduction. Billions were spent to get us almost nothing tangible.
Laptop machines have come a long way, but the desktop is stuck at 4 cores and no hint at anything but maybe 10% performance gains per year for the foreseeable future.
We are instead getting integrated crappy GPU's in flagship processors that will mostly never get utilized, and that crappy GPU is half the die area. I'd rather have the same die size with 6-8 cores, or more L2 cache, or almost anything else that I might actually make use of. Sadly, intel reserves those kinds of features for their much more expensive Xeon or "Extreme" branded lines.
Please stop. The rate of synapses firing is not equivalent to a CPU frequency. I love when people outside their depth like to make comments....
The neuron does not send data. Signal strength is incorrect, you are looking at activity. Stop trying to compare the two. It really doesn't work.
Intel is stalling at 14nm. Everyone else stalled at 28nm. 28nm is still the cheapest node in per transistor terms. Since most chip makers are driven by cost rather than transistor performance, there have been few takers for 20nm and 14nm.
I know - the transistor count should have enabled us to build neural nets to filter out inane AC comments by now. Hardly. The brain has...
Your whole post is a non sequitur, because the one you're replying to said nothing of trying to duplicate a brain. Neural networks are already in use for various purposes, and in no way imply they have anywhere near the same complexity as the whole human brain. Sometimes, they can perform tasks otherwise assumed to be complicated, and end up right more often than wrong with far, far less complexity than a human brain, or even modern processor.
Even better the brain is 3d where chips are two 2d flat landers.
i thought once I was found, but it was only a dream.
When I think about how much faster (at a hardware level) my current computer is relative to my old 486SX 33MHz, and how I still have to spend time waiting for the damn thing to catch up to what I'm trying to do, I just have to shake my head.
Part of the blame goes to slow mechanical hard disks, though.
You are correct. The brain is like a hard drive and rate of firing of synapses is more analogous to rotation speed.
Of course it runs much slower. With the kind of density the brain has if you increased the clockspeed a lot you couldn't even cool down the brain properly.
This is why people need to buy more AMD. They're the only ones pushing Intel to do anything good for the mass market.
Actually in a GOP politician they function as a biode (e.g. length of wire)