Ars: Moores Law dead at 51

[Arachnotronic]

Because Intel slashed their research budget because no one is buying high-performance chips anymore.

Could you provide proof that they slashed their research budget, or are you just making things up?

 

[frozentundra123456]

What they slashed was the money going to mobile, was it not?

 

[DrMrLordX]

wat

That's beyond wat. That's more like Angkor Wat. Wat doesn't get much more wat than that.

 

[DigDog]

it was badly worded from the beginning; the number of transistors does not equate computing power, and aside that, there never was any ground to stipulate the law in the first place. it ought to be called "Moore's Observation" rather than law.

 

[zephyrprime]

The end of the line for Moore's law was really the P4. That's when we couldn't get any more clock speed which was a disaster for the industry. I know moore's law talks about transistor density and not clock speed but without more clock there was only a little that could be done to get more speed.

 

[ed21x]

Lets be honest, without Apple, TSMC and Samsung wouldn't be able to afford anything. Its Apples willingness to pay "any price" for 100s of million chips that have gotten the foundries so far. Apple in trouble means TSMC and Samsung foundries are in trouble.

Actually, isn't Samsung its own largest customer, and Huawei the largest TSMC customer? I believe both Asus and HTC are buying from TSMC as well in the smart phone arena. As much media presence as Apple has, it's not exactly the largest by volume in phone or tablet, and they don't acquire components are a premium either, otherwise Apple margins wouldn't be so legendary.

Um, no. Intel isn't on 10nm right now because they just barely got 14nm figured out.

both cherrytrail and skylake have been out in quantity for almost half a year. the shortage of skylake was only for the 6700k due to demand by enthusiasts for the top of the line chip, while the non-K chips were always in ready supply, so the issue wasn't really being able to produce at that node, but intel's yield for chips that can meet 4 ghz with headroom to overclock.

 

[ShintaiDK]

Actually, isn't Samsung its own largest customer, and Huawei the largest TSMC customer? I believe both Asus and HTC are buying from TSMC as well in the smart phone arena. As much media presence as Apple has, it's not exactly the largest by volume in phone or tablet, and they don't acquire components are a premium either, otherwise Apple margins wouldn't be so legendary.

For SoC revenue and the segment to push new nodes? Absolutely not.

Remember Apple is willing to pay pretty much "any price" for leading edge wafers. Unlike the others, Apple only sell high cost products. You can somewhat compare them to Intel that sits on ~60% of all MPU revenue. The revenue that matters in terms of who can afford what.

The next after Apple would be Qualcomm due to SoCs and modems.

 

[crashtech]

Yeah, I am a bit jealous of Apple's latest flagship phone, doesn't look like any of the Androids will be eclipsing its performance anytime soon.

/OT

 

[StinkyPinky]

Yeah, I am a bit jealous of Apple's latest flagship phone, doesn't look like any of the Androids will be eclipsing its performance anytime soon.

/OT

Performance isn't everything though. In a phone I look for camera, battery, display over raw performance. But no denying the A9 is a beast.

 

[tolis626]

Performance isn't everything though. In a phone I look for camera, battery, display over raw performance. But no denying the A9 is a beast.

More than anything, the A9 accepts the state of mobile applications and that's why it excels in my opinion. Is it an excellent chip? Yes, yes it is. But in that utopia where all applications were properly threaded and optimized for all the quad- and higher core count CPUs available, the A9 would probably be getting its ass handed to it. But alas, things are as they are and apps are lightly threaded and more cores are usually a waste in mobile. For all its excellent attributes, I think its greatest strength is that it's not forward looking. And financially speaking, that's not a bad thing. It works now, and it works exceptionally well. Bulldozer was forward looking, and look how that turned out.

Just something I wanted to throw out there. Sorry for the off topic. Also, give my condolences to mr. Moore about his law. It will be sorely missed.

 

[videogames101]

More than anything, the A9 accepts the state of mobile applications and that's why it excels in my opinion. Is it an excellent chip? Yes, yes it is. But in that utopia where all applications were properly threaded and optimized for all the quad- and higher core count CPUs available, the A9 would probably be getting its ass handed to it. But alas, things are as they are and apps are lightly threaded and more cores are usually a waste in mobile. For all its excellent attributes, I think its greatest strength is that it's not forward looking. And financially speaking, that's not a bad thing. It works now, and it works exceptionally well. Bulldozer was forward looking, and look how that turned out.

Just something I wanted to throw out there. Sorry for the off topic. Also, give my condolences to mr. Moore about his law. It will be sorely missed.

10/7nm will be here before you know it.

 

[ehume]

10/7nm will be here before you know it.

Baked in. After that?

 

[Rakehellion]

Also, give my condolences to mr. Moore about his law. It will be sorely missed.

Every time I see this thread I think it says Gordon Moore is dead.

 

[Rakehellion]

More than anything, the A9 accepts the state of mobile applications and that's why it excels in my opinion. Is it an excellent chip? Yes, yes it is. But in that utopia where all applications were properly threaded and optimized for all the quad- and higher core count CPUs available, the A9 would probably be getting its ass handed to it.

So it's a chip designed to run apps that exist and not apps that don't exist? Imagine that.

 

[positivedoppler]

Newton's Laws are laws because they were mathematically derived and its result drove the advancement of mankind. Moore's Law is not a law, it's a human prediction or anticipation based on an educated guess. The semi conductor industry would be in the exact same place it is now regardless of rather or not Moore made a prediction about transistor doubling. Moore's law died at the gate because it was never a law to begin with. It was a human hunch.

 

[tolis626]

Every time I see this thread I think it says Gordon Moore is dead.

Well, that makes two of us.

So it's a chip designed to run apps that exist and not apps that don't exist? Imagine that.

That's exactly my point. It's not like AMD's typical "It will be great in the future when apps need this and that". Most mobile CPUs don't seem to be designed with here and now in mind.

 

[videogames101]

Baked in. After that?

5nm will probably bring another device change, cost will surely be an issue.

 

[Fjodor2001]

More than anything, the A9 accepts the state of mobile applications and that's why it excels in my opinion. Is it an excellent chip? Yes, yes it is. But in that utopia where all applications were properly threaded and optimized for all the quad- and higher core count CPUs available, the A9 would probably be getting its ass handed to it. But alas, things are as they are and apps are lightly threaded and more cores are usually a waste in mobile.

At least on Android, multi-core use is much more common than what you might think:

http://www.anandtech.com/show/9518/the-mobile-cpu-corecount-debate/18

Chrome and to a lesser extent Samsung's stock browser were able to consistently load up to 6-8 concurrent processes while loading a page
[...]
What we see in the use-case analysis is that the amount of use-cases where an application is visibly limited due to single-threaded performance seems be very limited. In fact, a large amount of the analyzed scenarios our test-device with Cortex A57 cores would rarely need to ramp up to their full frequency beyond short bursts
[...]
On the other hand, scenarios were we'd find 3-4 high load threads seem not to be that particularly hard to find, and actually appear to be an a pretty common occurence.

Also, I agree Apple A9 is a nice chip, but primarily in ST performance. In MT performance it gets is ass kicked by Samsung/Qualcomm/Mediatek.

 

[StrangerGuy]

Meh, its already long irrelevant at least for CPUs. There is only so much you can throw transistors at when you are well within the deep diminishing returns territory and software scales much, much worse.

 

[nenforcer]

Um, no. Intel isn't on 10nm right now because they just barely got 14nm figured out.

This is perfectly fine - because the rest of industry still hasn't gotten .14nm products on the market yet in mass quantities while Intel unleashed there full .14nm product stack late last Summer.

Intel first .14nm products were the Core M series released in late 2014 - with Samsung Exynos release in early 2015 for the Galaxy 6 series.

Intel has just reiterated that it's .10nm products are on track for 2017 - whereas the rest of the industry (AMD CPU, NVidia and AMD GPU, all other mobile providers) will just be "figuring out" .14nm Samsung and .16nm TSMC.

Intel's what used to be a 2 year gap is now down to 1 year or less - but they are still leading the industry with their own foundries.