Intel Says Chips To Become Slower But More Energy Efficient

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
This comes straight from the company’s Executive VP, who said that “the best pure technology improvements we can make will bring improvements in power consumption but will reduce speed

Intel has said that new technologies in chip manufacturing will favour better energy consumption over faster execution times – effectively calling an end to ‘Moore’s Law’, which successfully predicted the doubling of density in integrated circuits, and therefore speed, every two years.
 
Considering how few things tax most chips to their full extent these days, I'm fine with this.
 
Yeah I remember the time you could take all the CPU power you could get your hands on.

Now I just look at taskmanager and sigh watching software hammering just two cores of the 8 available or sipping 20% of each one.

I think Handbrake is the only software (other than a couple of benchmarks) I have that runs all cores at 100%.
 
The speed of chips is not always directly tied to density directly. Many of the doubling's have been due to the integration of separate chips and functions which although benefiting the overall computing speed don't always translate to a MegaHz increase. That said, with the current consumer and enterprise consumption of products, a more efficient chip would benefit more buyers than a faster chip would. I suspect, however, that Intel would still continue to produce some limited volume high performance chips for workstations and enthusiasts.
 
When the competition is so weak you can start to downgrade your own products you know the market is a mess.

Nevertheless, software is really stagnant, it really is time for miniaturisation.

It would be a nifty thing to keep your phone in your pocket, sit at your desk, and use it as a wireless desktop. I suspect that's probably fairly distant as of today.
 
What's that stupid noise? Not everyone's computer usage stops at facebook and games.

I could have 10x times computation power tomorrow, I'd still say it's not enough.
Smartphones and tablets are not substitutes for workstations. What dream world is he in? And this guy calls the shots? OMG.

I hate when someone applies their own needs universally to everyone. "I don't need it therefore nobody needs it" The ego, is strong with those.

Yes, there absolutely should be smaller and more power efficient chips, for the areas that need those. But that doesn't mean we should completely abandon the pursuit of more computation capability at the same time.
 
Hmm.. Moore's Law doesn't really say anything about speed. It really just pertains to transistor density.
 
Yes, there absolutely should be smaller and more power efficient chips, for the areas that need those. But that doesn't mean we should completely abandon the pursuit of more computation capability at the same time.

And you'd be foolish to think they had. They work on these processors for years before release. He's likely referring to the next 5-6 years. Within a decade I'm sure there will be a "leap" in speeds.

That is, if we don't kill ourselves off first.
 
From the article, it appears that what was said was that as we transit to semi-quantum-based production, priority will be given to efficiency, and as a result, we may see the performance starting off slightly lower than where we left off during the early years of semi-quantum.

He never said Intel will intentionally slow the chip down. The talk is about technology of the future, not current technology. It may be the case that as we transit to something new, we will have to give something up.

Take SSD for example. At the moment, we are prioritizing speed over cost per capacity with SSD. when we purchase SSD, we are sacrificing capacity (when compared to mechanical HDD). Someday SSD may be a better storage in every way over mechanical HDD, but it cannot start off that way. Likewise, such may be the case for future ways of manufacturing semiconductor chips.
 
As previously mentioned, as long as we have crappy software that doesn't take advantage of all our cores, then we need faster/more efficient chips. Period. Power savings be damned.

Nothing irks me more than having something peg one logical processor while I have the other 7 IDLING.
 
And you'd be foolish to think they had. They work on these processors for years before release. He's likely referring to the next 5-6 years. Within a decade I'm sure there will be a "leap" in speeds.

That is, if we don't kill ourselves off first.

He's telling me, that they're going to make slower more efficient chips.

Then he's telling me that we don't need more powerful cpus anyway because of tablets and smartphones.

OFC most of what he's saying is not fact, but marketing bullshit. It sounds better to say what he says instead of coming out with the cold hard facts.

The truth is probably something along the lines of: We focused our R&D on a new technology that is unable to yield powerful enough cpus yet, but we're going to sell them anyway and build our marketing around efficiency instead of power.
 
What's that stupid noise? Not everyone's computer usage stops at facebook and games.

I could have 10x times computation power tomorrow, I'd still say it's not enough.
Smartphones and tablets are not substitutes for workstations. What dream world is he in? And this guy calls the shots? OMG.

I hate when someone applies their own needs universally to everyone. "I don't need it therefore nobody needs it" The ego, is strong with those.

Yes, there absolutely should be smaller and more power efficient chips, for the areas that need those. But that doesn't mean we should completely abandon the pursuit of more computation capability at the same time.

Nothing in his statement indicated that Intel would abandon anything. Intel is segmented for a reason, so that they can satisfy almost all market needs and not leave a beachhead for a competitor to carve out a foothold. Intel will continue to support the enthusiast/performance market (even thought it is smaller than the regular market) with older technologies until the new technologies are ready to take over that function.

Also as Hornet mentioned, technologies align to the market demands. Most of the CPU market is currently focused on mobility (probably 80% or more) so Intel is going to do their best to satisfy that market's needs (which revolve more around efficiency than speed right now). That in no way indicates that Intel will ignore the small but lucrative performance market where people are not so cost and efficiency oriented. The SSD market is the opposite right now, most of us don't care about cost yet, we want performance (and are willing to pay the premium for that).
 
My computer puts out more heat than a space heater and is louder than a vacuum, I think that about sums up how much I give a shit about "efficiency".
 
I guess Intel is not researching gallium or germanium as a silicon replacement. Those techs can/ will enable 5+ghz chips
 
I guess Intel is not researching gallium or germanium as a silicon replacement. Those techs can/ will enable 5+ghz chips

Pentium 4 was supposed to do 10ghz+, remember?

Gallium is so last decade. The next thing is carbon nanotubes
 
Pentium 4 was supposed to do 10ghz+, remember?

Gallium is so last decade. The next thing is carbon nanotubes


All research says the new silicon replacements can and will be transitioned to. Intel makes it sound like they're just giving up.
 
All research says the new silicon replacements can and will be transitioned to. Intel makes it sound like they're just giving up.

If you read the full article you don't see anything about them giving up. I think the operative quote is:

Holt said Intel doesn’t yet know which new chip technology it will adopt, even though it will have to come into service in four or five years. He did point to two possible candidates:

There is no indication that Intel (or the rest of the semiconductor industry) has committed themselves to any alternative technology at this time
 
If you read the full article you don't see anything about them giving up. I think the operative quote is:



There is no indication that Intel (or the rest of the semiconductor industry) has committed themselves to any alternative technology at this time


I suppose "Sounding" of the articles I've read, vs the "reality" of what intel will do is the issue then.

Instead of focusing on how we'll not get the raw performance. Intel should have said they'll be using new tech to reach more ghz.
 
No need to upgrade for at least another 10 years then, right? Suites me just fine. No new cpu, no new video card, no new Win10. save save save = win win win
 
No need to upgrade for at least another 10 years then, right? Suites me just fine. No new cpu, no new video card, no new Win10. save save save = win win win

You can bet they'll make you upgrade by, for example adding a new instruction set.
 
So, either I pick up a 5930k asap or I buy a new Zen processor when they come out. Interesting decision since I will not buy a processor that is slower than what came out last year just because they say so. I guess this is now tehe oftforums because people now have no issue with less performance than before.

Maybe video cards will be next because "Power Efficiency". :rolleyes:
 
I suppose "Sounding" of the articles I've read, vs the "reality" of what intel will do is the issue then.

Instead of focusing on how we'll not get the raw performance. Intel should have said they'll be using new tech to reach more ghz.

Again, you assume that is what the mass market wants. Intel and AMD both went multiple core since that was what the market wanted (not faster single or dual cores). One of the trade-offs of all those cores is speed (it is hard to dissipate heat from all those cores at higher speeds). Intel went for power efficiency (along with the GPU manufacturers for mobile) because that is what the laptop market wanted. Intel, AMD, and others aren't stupid. If there was really a market for 10 GHz laptops and servers then they would be building those.

I suspect that Intel will always be building performance chips to satisfy that market. They will not want to give it up to a competitor and those of us that want performance don't expect that to come at bargain basement prices like the regular consumer products. I don't think any of us expect m.2 SSDs or the newer PCIe SSDs to be cheap anytime soon but we will continue to buy them because performance trumps price for us. As long as we remain loyal to that then the companies will remain loyal to us too.
 
So, either I pick up a 5930k asap or I buy a new Zen processor when they come out. Interesting decision since I will not buy a processor that is slower than what came out last year just because they say so. I guess this is now tehe oftforums because people now have no issue with less performance than before.

Maybe video cards will be next because "Power Efficiency". :rolleyes:


We usually do okay as long as we don't have unrealistic expectations of Intel and AMD. Expecting high performance chips from them is reasonable. Expecting "cheap" high performance chips from them is not reasonable since that is not what the majority of consumers want. As long as we don't cheap out there will always be performance products available for the [H] consumer. As nice as it would be to have Titan performance for $250, that is not reasonable and most of us can shell out the $1000 for a Titan to leverage maximum video performance, if we want or need it.
 
Again, you assume that is what the mass market wants. Intel and AMD both went multiple core since that was what the market wanted (not faster single or dual cores). One of the trade-offs of all those cores is speed (it is hard to dissipate heat from all those cores at higher speeds). Intel went for power efficiency (along with the GPU manufacturers for mobile) because that is what the laptop market wanted. Intel, AMD, and others aren't stupid. If there was really a market for 10 GHz laptops and servers then they would be building those.

I suspect that Intel will always be building performance chips to satisfy that market. They will not want to give it up to a competitor and those of us that want performance don't expect that to come at bargain basement prices like the regular consumer products. I don't think any of us expect m.2 SSDs or the newer PCIe SSDs to be cheap anytime soon but we will continue to buy them because performance trumps price for us. As long as we remain loyal to that then the companies will remain loyal to us too.

There is a market for super fast single core speeds in the server market, not to mention gaming desktops. However obviously that's a niche market.
http://www.techpowerup.com/219275/i...chip-based-on-the-broadwell-architecture.html
 
There is a market for super fast single core speeds in the server market, not to mention gaming desktops. However obviously that's a niche market.
http://www.techpowerup.com/219275/i...chip-based-on-the-broadwell-architecture.html

And I would expect Intel to continue to try and service that market. We just shouldn't expect that kind of performance at the same price as those products designed to sell to 100 million consumers or enterprise customers. But as long as we are willing to pay to play there will always be products for the [H] consumer too
 
I was waiting for this thread. :p The original source article on MIT Technology Review has context of what Holt said last week at ISSCC. Unfortunately it's a mess, since the author doesn't seem to understand that Moore's Law has to do with doubling of transistors, not clock speed and mixes up the entire context of the talk. Worse, is this was probably the best treatment before the tech fail-o-sphere got to it. https://www.technologyreview.com/s/...-to-sacrifice-speed-gains-for-energy-savings/

Holt was talking about things like quantum tunneling transistors and spintronics-based replacements for silicon transistors, which are nowhere near ready for production. In that context:
Holt said:
...But the two new technologies can’t work on data as fast as silicon transistors. “The best pure technology improvements we can make will bring improvements in power consumption but will reduce speed,” said Holt.

...

“Particularly as we look at the Internet of things, the focus will move from speed improvements to dramatic reductions in power,” Holt said.

What I got out of it and the ISSCC program description* is that Intel is still working on high speed transistors and there's no reason to suspect Intel is planning on regressing clock speeds. However, there is much Intel can do in the future for other applications to lower power consumption.

* said:
Moore’s Law has served as the guiding principle for the semiconductor industry for 50 years.
But now there are growing concerns and doubts over the vitality of Moore’s Law going
forward, given the scaling challenges we face. This talk will directly address those concerns
and explore future opportunities for the industry. We will present the scaling benefits for
power, performance, and cost using specific product and design examples based on
state-of-the-art 14nm CMOS technology, for applications ranging from high-performance
computing down to ultra-low-power mobile applications. In additional to the scaling path of
CMOS technology beyond 14nm, this talk will also discuss some leading technology options
on the horizon beyond CMOS and their potential design benefits in advancing Moore’s Law
well into the future. Novel 3D heterogeneous integration schemes and new memory
technologies will be discussed for their potential in optimizing the memory hierarchy and
addressing bandwidth challenges in processor performance and power.
 
So, either I pick up a 5930k asap or I buy a new Zen processor when they come out. Interesting decision since I will not buy a processor that is slower than what came out last year just because they say so. I guess this is now tehe oftforums because people now have no issue with less performance than before.

Maybe video cards will be next because "Power Efficiency". :rolleyes:


I don't think you've read the article.

Holt has stated not just that Moore’s Law is coming to an end in practical terms, in that chip speeds can be expected to stall, but is actually likely to roll back in terms of performance, at least in the early years of semi-quantum-based chip production, with power consumption taking priority over what has been the fundamental impetus behind the development of computers in the last fifty years.

He's talking about completely new technology, where he believe it may not start off where the old one left in terms of raw performance, because power efficiency will take precedence. He never said they will be intentionally made slower, or that they will no longer improve on performance.

You can laugh all you want, but the reality is that power efficiency is an important factor. Most of the power consume are dissipated in terms of heat, and you cannot focus on performance alone while allowing power consumption to run lose. It's a design approach that simply isn't going to work. It's not about electrical bill, it's about whether you chip will work at all.

If you want to improve the performance of a new design, you will need to do so without increasing it's power consumption. In other words, you are improving it's power efficiency every time you make advancement. It is all about power efficiency.
 
It has probably already been mentioned, but Moore's law doesn't necessarily mean increasing speed. Simply the density on the chip. Lower power, but slower, chip design and Moore's law are not mutually exclusive.


They need to start making processor cores the size of transistors

Given a transistor is a component of a CPU, somehow, I don't see it happening. :p
 
For those of you guys with extra processing power to spare, why not engage in the HardOCP folding team or one of the many other DC projects that forum members are engaged in?

Apparantly, AMD never got the memo detailing Moore's law although I have been happy with my Turion 64 X2 2.5ghz ever since it first came out.
 
I can see the appeal for laptops, but for desktops... for the most part I could give two craps. 240mm water AIO coolers are cheaper and more reliable now than ever. I want a 24 core super beast that I need to hook a 240V outlet for in my room, and vent into the attic.
 
For those of you guys with extra processing power to spare, why not engage in the HardOCP folding team or one of the many other DC projects that forum members are engaged in?


When electricity becomes free or I move back in with my parents...maybe.
 
I thought transitioning to new materials was supposed to allow faster speeds though? You can still take advantage of a faster processor when it comes to alpha games that use a single core and workstation applications.
 
I thought transitioning to new materials was supposed to allow faster speeds though?
It does. The actual quote from Holt is that the rate of clock speed increases may slow in the future, but the most benefits from the two named materials (both quantum tunneling and spintronics-based gates are nowhere near ready for production) would be in low power properties running at lower clock rates. As an example, Holt uses IoT as an application which could benefit from that.

Since that design goal is not suitable for high performance MPUs, it's highly unlikely to be the future direction for Intel's PC/server chips. Given prior focus, it seems that Intel will be embedding "novel" gate characteristics using III-V materials in upcoming nodes for the foreseeable future.
 
You can bet they'll make you upgrade by, for example adding a new instruction set.

If I don't buy any new software or games I will never need a new instruction set. Already have everything I need now so they can all take a hike.
 
Back
Top