Intel: Tick-Tock cycle dead, now instead "Process-Architecture-Optimization"

[DocSwag]

Source: http://www.anandtech.com/show/10183/intels-tick-tock-seemingly-dead-becomes-process-architecture-optimization

Quote

As reported at The Motley Fool, Intel’s latest 10-K / annual report filing would seem to suggest that the ‘Tick-Tock’ strategy of introducing a new lithographic process note in one product cycle (a ‘tick’) and then an upgraded microarchitecture the next product cycle (a ‘tock’) is going to fall by the wayside for the next two lithographic nodes at a minimum, to be replaced with a three element cycle known as ‘Process-Architecture-Optimization’.

Quote

From Intel's report: As part of our R&D efforts, we plan to introduce a new Intel Core microarchitecture for desktops, notebooks (including Ultrabook devices and 2 in 1 systems), and Intel Xeon processors on a regular cadence. We expect to lengthen the amount of time we will utilize our 14nm and our next generation 10nm process technologies, further optimizing our products and process technologies while meeting the yearly market cadence for product introductions.

Well, we have been hearing a lot about Moore's law being dead and stuff recently. I guess this is just another confirmation for that. Intel seems to be officially calling the tick-tock cycle dead as well, and are now moving onto a process architecture optimization cycle, or as I like to call it tick-tock-tack (Sorry if it sounds lame but it makes sense). 

We all know that instead of bringing Cannonlake this year Intel will be bringing Kaby Lake instead. It was pretty obvious Kaby Lake probably wouldn't be a brand new architecture. Now this is just more confirmation that Kaby Lake will only be some optimizations and not anything new. People have been suggesting a new graphics architecture, which sounds pretty likely to me, and perhaps a few small tweaks to the CPU itself. 

 

One thing I am wondering-other silicon chip manufacturers like TSMC seem to still be somewhat following Moore's law. While Intel has postponed 10nm to 2017 and 7nm to at least 2019 and most likely 2020, TSMC has instead said it will begin 10nm this year and have 7nm in 2018. Intel has always had the superior manufacture process at each node, but this probably won't hold true when comparing TSMC 7nm vs Intel 10nm. If Intel is moving to a 3 year cycle while companies like TSMC are not, could this mean Intel will start to fall behind when it comes to manufacturing process? Only time will tell, I guess.

[Fred Castellum]

Postponed?! Damit Intel, the speed of advancements should go up! Not slow down!

 

Dam market control probably gives them the leverage to not be as aggressive.

[Coaxialgamer]

Dude... One thing history has shown...

 

NEVER underestimate intel's engineers .

 

Companies went under because they did just that.

[pagani123]

3 minutes ago, Coaxialgamer said:

Dude... One thing history has shown...

 

NEVER underestimate intel's engineers .

 

Companies went under because they did just that.

thats a given rule like never try to invade Russia  

[_ASSASSIN_]

5 minutes ago, MyInnerFred said:

Postponed?! Damit Intel, the speed of advancements should go up! Not slow down!

 

Dam market control probably gives them the leverage to not be as aggressive.

or it could be that creating processes that are that damn small is difficult when you have to deal with electrical interference from circuits at that scale as well as heat dissipation.

But yes I'd imagine since they have such a grip on the x86 market, they don't need to worry about that market segment. Intel is focusing on mobile processing heavily at the moment and will continue to do so. 

[Fred Castellum]

Just now, _ASSASSIN_ said:

or it could be that creating processes that are that damn small is difficult when you have to deal with electrical interference from circuits at that scale as well as heat dissipation.

But yes I'd imagine since they have such a grip on the x86 market, they don't need to worry about that market segment. Intel is focusing on mobile processing heavily at the moment and will continue to do so. 

And yet their mobile processor track record has still been disappointing for investors. 

 

I understand the difficulty that comes with creating smaller and smaller chips but the way they worded their changes seems far more like a excuse to slow down on advancement because of a detour in technological capabilities while these lack of technological capabilities have been being discussed about for years now and should have various alternative options to approach or remedy.

[i_build_nanosuits]

seriously...an old dvd cut into halves for a graph...intel that's not serious right?

[Guest]

Well, it was more like

Tick *semi-tock* Tock *semi-tick* Tick + or - a few tick's and tocks

As we've seen with skylake, Kaby lake doesn't seem too promising 

Ex) Haswell, Broadwell, Skylake

Tick Semi Tock Tock

But yeah

Intel should focus on getting the nm's down and getting the GHz up towards 3.5+ on all CPU's, Power effiency should be kept to where it is now, and they should try to reduce it on i7E to the power efficiency of a regular i7, since not too many people buying i7's care about an extra $1 or 2 on their powerbill 

[Coaxialgamer]

1 minute ago, i_build_nanosuits said:

seriously...an old dvd cut into halves for a graph...intel that's not serious right?

well considering all the money they put into R&D , they probably couldn't afford more than a class of 11 year olds to make their graphs , would they

[Oshino Shinobu]

11 minutes ago, MyInnerFred said:

Postponed?! Damit Intel, the speed of advancements should go up! Not slow down!

 

Dam market control probably gives them the leverage to not be as aggressive.

It's a shame. Intel really doesn't have much competition on the high performance market, so they seem to just be minimizing their R&D costs from what I can tell. 

[Kherm]

4 minutes ago, i_build_nanosuits said:

seriously...an old dvd cut into halves for a graph...intel that's not serious right?

 I'm pretty sure they're supposed to be halves of silicon wafers.

[Guest]

14 minutes ago, i_build_nanosuits said:

seriously...an old dvd cut into halves for a graph...intel that's not serious right?

I hope you're kidding .-.

its a silicon wafer I really hope you're joking 

[Roawoao]

32 minutes ago, Coaxialgamer said:

Dude... One thing history has shown...

 

NEVER underestimate intel's engineers .

 

Companies went under because they did just that.

There is one other thing to not underestimate and that is called the laws of physics.

 

I highly doubt they will be able to keep up even a 3 year process node cadence given how they themselves predict dumping silicon for much more expensive high performance materials such as GaAs and other alternatives that have been in use for many years in applications where cost is no object and you physically need the material to even get it to work.

 

Given how I've noticed my 5 year old computers/laptops being perfectly usable Moore's law as effectively dead a long time ago for consumers especially since I remember way back decades ago where every few years was a literal giant leap (>2x) in computer performance while now it is at best 10-20% better. The only revolution that changed the way I use computers was SSDs in recent times.

 

What will likely happen is tons more fixed function hardware and the inclusion of FPGAs (Intel owns Altera now) to try and eek out differentiation every year.

[patrickjp93]

1 hour ago, MyInnerFred said:

Postponed?! Damit Intel, the speed of advancements should go up! Not slow down!

 

Dam market control probably gives them the leverage to not be as aggressive.

It slows down necessarily because of an exponential increase in difficulty. ASML is 2 generations late getting their EUV Lithography tech out the door (it was originally intended for use at the 14nm node, not the 7nm, and it's only viable down to ~3nm anyway). The entire industry is switching gears on new materials and revolutionary new technologies. Intel is going as fast as it can. That doesn't mean perceptually for us as consumers that they are moving quickly. Further, they want to become leaders in FPGAs. That requires a process node advantage. Intel is moving as fast as physics and its engineering prowess allows (and some internal politics, certainly).

 

1 hour ago, Maybach123 said:

thats a given rule like never try to invade Russia  

You can invade Russia, but don't try to fight in the Winter. You WILL lose.

 

58 minutes ago, Coaxialgamer said:

well considering all the money they put into R&D , they probably couldn't afford more than a class of 11 year olds to make their graphs , would they

Intel historically has very simple, logical presentation formats. Even their programming tutorials have the same theme throughout. Boring? Sure, but I care about the substance, not the frill.

 

59 minutes ago, Oshino Shinobu said:

It's a shame. Intel really doesn't have much competition on the high performance market, so they seem to just be minimizing their R&D costs from what I can tell. 

IBM Power 8 and Oracle Sparc 7 both say hello. Intel has a lot of competition still going strong. Simply follow the SAP business benchmarks. x86 gets creamed in large databases and mass transaction workloads. That's one reason the TSX instructions were created. And that's why Several of its partners are creating 32-socket x86 server boards where Intel only officially supports up to 16 (not having to move between boards/nodes reduces a lot of communication overhead). And that's why Purley expands the interconnect to 16-star-associative instead of 8.

 

30 minutes ago, Roawoao said:

There is one other thing to not underestimate and that is called the laws of physics.

 

I highly doubt they will be able to keep up even a 3 year process node cadence given how they themselves predict dumping silicon for much more expensive high performance materials such as GaAs and other alternatives that have been in use for many years in applications where cost is no object and you physically need the material to even get it to work.

 

Given how I've noticed my 5 year old computers/laptops being perfectly usable Moore's law as effectively dead a long time ago for consumers especially since I remember way back decades ago where every few years was a literal giant leap (>2x) in computer performance while now it is at best 10-20% better. The only revolution that changed the way I use computers was SSDs in recent times.

 

What will likely happen is tons more fixed function hardware and the inclusion of FPGAs (Intel owns Altera now) to try and eek out differentiation every year.

What do you think Intel's been researching the last 5 years while ASML has been floundering? Intel has several contenders for the next-gen material be it Silicon-Germanium as IBM demoed, one of the III-V family (Gallium-Arsenide being one iirc), or Carbon Nanotubes which it has been collaborating with MIT's research labs on. Transitioning nodes every 2-3 years will be possible, but the tuning time will be extended. AMD does 3 product refreshes per architecture, so is it really unfair to be hating on Intel for this (not you specifically, but in general)?

 

You are correct about fixed-function units. Intel is certainly learning from Apple on this one.

[themaniac]

1 hour ago, Maybach123 said:

thats a given rule like never try to invade Russia  

Japans done so before in fact they were the first eastern nation to successfully invade a western nation and guess what that western nation just happens to be Russia

[pagani123]

Just now, themaniac said:

Japans done so before in fact they were the first eastern nation to successfully invade a western nation and guess what that western nation just happens to be Russia

ok let me amend that never invade russia on the western front in the winter.

[Sakkura]

16 minutes ago, themaniac said:

Japans done so before in fact they were the first eastern nation to successfully invade a western nation and guess what that western nation just happens to be Russia

Japan never successfully invaded Russia. They won the Russo-Japanese war in naval battles and in land warfare in Manchuria and Korea. They were defeated by Soviet forces in the lead-up to WWII.

[azurezeed]

At least it's not AMD tick tock tack teck tuck

 

 

/jk

[Energycore]

This feels like a calm before the storm moment. Since we've pushed the limits of standard silicon transistors as far as they will go (dammit planck), we're about to see a revolution in computing that could both give us astronomically high performance increases, or not very much at all. The tension is high.

[KingUnicorn]

1 hour ago, patrickjp93 said:

or Carbon Nanotubes

carbon nanotube tranSISTORS! *pushes up glasses*

 

I saw the chance for a reference. I took the chance.

For this who don't know, it's a joke on net linked daily on NCIX.

[patrickjp93]

13 minutes ago, Potato_King said:

carbon nanotube tranSISTORS! *pushes up glasses*

 

I saw the chance for a reference. I took the chance.

For this who don't know, it's a joke on net linked daily on NCIX.

Well played sir!

[TrigrH]

Here is the real graphic:

[patrickjp93]

1 minute ago, TrigrH said:

Here is the real graphic:

I'm not a photoshopper or even a user of Paint, so here's my two cents.

Process (get the new node producing the tweaked old architecture well, add in new graphics, boost up core count for servers and possibly enthusiasts)

Architecture (introduce the new one with new demanded feature sets, even if there isn't time to tune up clock speeds or power, add new graphics if ready, possibly add cores)

Optimization (shave instruction latencies, introduce more instructions as demanded, tune up clock speed, add new graphics if ready)

 

And pray IBM doesn't get a massive breakthrough at the wrong time.

[Roawoao]

1 hour ago, Energycore said:

This feels like a calm before the storm moment. Since we've pushed the limits of standard silicon transistors as far as they will go (dammit planck), we're about to see a revolution in computing that could both give us astronomically high performance increases, or not very much at all. The tension is high.

I'd bet genetic engineering is going to be the next actual technological revolution vs some magical breakthrough that saves moore's law for another 2 years after being dead for 5 years or so. Instead of making faster computers why not longer living, smarter, healthier people. 

 

Not only that I'd almost certain AI is just going to take the cheap route and use a literal squishy brain on a chip using genetically engineered neurons that are adapted to live on an active electrode array.

 

There is no way chip scaling can go sub-atomic or even close to that and I highly doubt quantum computers will ever be practical in consumer applications let alone any given their highly statistic dependent operation (almost random) and almost all requiring cryogenic conditions to operate and a classical super computer clusters to process the results.

 

If you think about it biology has stuff down to 1nm and can both do self assembly, computing, energy, .... all using cheap/easy to make carbon based compounds. You want nanomachines well protiens in biology protiens literally are nanomachines everything from motors, electrical gates, ratchets, sensors, .. if you think about it gray goo already happened a long time ago given the planet is pretty much plastered with self replicating diverse lifeforms it is just more of a green goo situation. 

 

With the level of understanding in genetics just starting to take off now it is a pretty ripe area of development. Compared to Intel, life in general has been developing this stuff continuously for a few billion years and even if we just figure a tiny bit out it could be revolutionary (generally cure cancer kind of stuff). Of course there are also huge ethical, technical, safety pitfalls as well given any powerful technology.

[Big-G]

2 hours ago, themaniac said:

... that western nation just happens to be Russia

Russia = Western = Does not compute