Moore's law & the Raspberry Pi

[Woll]

I've been wondering about Moore's law, that the speed of computers are said to double approximately every 18 months.

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

[expandables]

I've been wondering about Moore's law, that the speed of computers are said to double approximately every 18 months.

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

I think moore was talking about x86/x64 archiecture computers.

[rpdom]

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

I think moore was talking about x86/x64 archiecture computers.
[/quote]
Did the x86/x64 archiecture (sic) exist in 1965? or even 1975 when he revised the "law"

[expandables]

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

I think moore was talking about x86/x64 archiecture computers.

Did the x86/x64 archiecture (sic) exist in 1965? or even 1975 when he revised the "law"
[/quote]
Moore was looking into the future and it was known at that time that he had the DeLorean.

[fruitoftheloom]

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

I think moore was talking about x86/x64 archiecture computers.

Did the x86/x64 archiecture (sic) exist in 1965? or even 1975 when he revised the "law"

Moore was looking into the future and it was known at that time that he had the DeLorean.

Well the grand-daddy of the x86 the 4004 was released 1971

[rpdom]

Will this apply to new generations of Raspberry Pi's or is this law now obsolete?

I think moore was talking about x86/x64 archiecture computers.

Did the x86/x64 archiecture (sic) exist in 1965? or even 1975 when he revised the "law"

Moore was looking into the future and it was known at that time that he had the DeLorean.[/quote]
A TARDIS is the one true time machine

[Woll]

What about the graphene Revolution that has often been talked about. When will graphene be getting into everyday electronics?

[gsh]

Moores law stopped around 2005.

Since then you've not actually seen faster processors (Moore's law specifically discussed clock frequency). Just more and more of them...

Welcome to the future of computing...

[W. H. Heydt]

Actually...Moore was specifically talking about the number of discrete circuit elements integrated on a single chip. He wasn't addressing clock speed at all. Even with that correction, gsh is correct that "Moore's Law" (which isn't what Moore ever called it) has broken down relatively recently as it has become harder and harder to shrink circuit elements as the technology approaches materials and lithography limits. The ultimate limit (so far as anyone knows) is when you get down to single atoms being important, and at the cutting edge of the nodes now being looked at, we're getting rather close.

Now as to how that applies to the Pi...it's doesn't really. The current Pis are using a 40nm node. Intel is getting ready to try to make chips using a 10nm node. It's going to be a Whole Bunch of generations of Pis before this is an issue.

[gsh]

Actually no you are wrong, the cost to create a set of masks is now around 10x at 10nm what it costs at 40nm but the extra transistors you get per mm2 doesn't make up for it. So it's not efficient unless you're earning massive markups (i.e. Intel) or you need the high speed (Intel) and don't care about the power consumption (Intel)

Currently the sweet spot is 28nm so an improvement is available. But you still need to dig up $5m to create a new chip

Gordon

[Woll]

I find it interesting as I'm a bus driver. But as soon as there is driverless technology then bus drivers will be obsolete. This would cut companies costs as there are no wage bills or driver hours regulations and is in the companies interest. But no very good for the drivers.

It's interesting how technology will shape the future. Graphene is supposed to be set to revolutionize technology of the future, but they were saying it would be available by now a few years ago.

[Woll]

GSH, isn't graphene resistance free because it's only an atom thick?

[Woll]

W. H. Heydt, if technology is reaching its design limits, is it restricted in what it can do? Much the same as you can only do so much with steam power?

[W. H. Heydt]

Actually no you are wrong, the cost to create a set of masks is now around 10x at 10nm what it costs at 40nm but the extra transistors you get per mm2 doesn't make up for it. So it's not efficient unless you're earning massive markups (i.e. Intel) or you need the high speed (Intel) and don't care about the power consumption (Intel)

Currently the sweet spot is 28nm so an improvement is available. But you still need to dig up $5m to create a new chip

Gordon

I agree. Right now, going to 28nm makes a lot of sense. This is a lot of performance to be wrung out of 28nm before going smaller That's why I said that there are going to be a lot of generations of chips in future (20 years? 30 years? I probably won't live long enough to see it.) of Pis before the problems Intel is facing become an issue for the Pi.

And, by the way, Intel does care about power dissipation. Getting rid of heat in their CPUs is a major issue and they're trying to drive TDP down while driving perfomance up. And that's totally separate from their desire to grab off as much as they can of the phone/tablet/phablet market, for which they need power requirements close to, or better than, ARM designs. Hasn't worked too well for them so far.

[jamesh]

W. H. Heydt, if technology is reaching its design limits, is it restricted in what it can do? Much the same as you can only do so much with steam power?

I suspect there is still life in steam power yet. We just don't spend any money developing it for everyday use, and after all, steam turbines produce all our electricity.

[TideMan]

I suspect there is still life in steam power yet. We just don't spend any money developing it for everyday use, and after all, steam turbines produce all our electricity.

Meaning you have no hydro in your country?
I didn't know that.

[W. H. Heydt]

W. H. Haydt, if technology is reaching its design limits, is it restricted in what it can do? Much the same as you can only do so much with steam power?

It's not a case of "technology reaching it's limits". It's a matter of this particular technology reaching a limit of how small circuit elements can be (and at what level does it cease to be economically viable to go smaller). We've already seen (as Gordon noted) a "wall" on clock speed. That was actually driven by the thermal dissipation issue.

At one point a number of years ago some honcho from Intel got up and showed a chart indicating that if clock speeds continued to rise as they had been (and heat generation with it), within a few years the surface of the CPU chips was going to be hotter than the surface of the Sun. Now, quite obviously, you can't do that. The solution (for Intel, at least) was to shop increasing the clock speed and try something else...enter the multicore CPU chip. If you can't ramp up the speed of the CPU, go parallel and improve throughput that way. Note that recent boards are pretty consistently showing up with actual clock speeds around 1.2GHz, and with heat sinks installed, even at that speed. This suggests that SBC SoCs are hitting pragmatic thermal limits.

At this point, some of the server CPU chips are touting over 20 cores per physical chip. This works well when the system load can be partitioned across many processing elements. Doesn't work well at all when your code is, basically, a large, single thread. So at least part of the solution is to develop programming languages and tools to make multi-threaded applications easier to build, plus the training to understand how to design code that can take advantage of having multiple, concurrent processes going on.

Much of this is why I say that holding the Pi at 4 cores is probably enough. As Gordon said about 28nm node...it's a "sweet spot". This is in terms of SoC complexity and efficiency as well as enough cores that people can learn/study/analyze multi-processing without the knock-on effects of adding a lot of other costs that would make to Pi much too expensive for it's role. (See my remarks--and those of at least one other poster--about the 512MB PINE64 board.) Right now, the 1GB of RAM on the Pi2B is adequate, given that it has 32-bit cores.

When--eventually--some version of the Pi move to 64-bit cores, then the memory will probably have to be increased to *at*least* 2GB and quite possibly more. If one assumes that in the middle future (say 5 to 8 years out), 4K video goes mainstream and the ability to handle it at reasonable frame rates--say 60fps--and decode in real time, whatever GPU is in use is going to eat up a lot more memory, and the combination of 64-bit cores (even keeping to 4 cores) and 4K video would suggest 4GB RAM--or more--would be in order. Clearly, prices for the various components need to come down quite a ways before all of that can be sold for $35.

[Woll]

Chips will go the same way as big CRT televisions. A slight change in the way things are done will do to chips what LED TVs did to televisions.

Jamesh, I was amazed to see on a recent TV documentary about Sellafield that steam turbines were used in nuclear power plants.

[W. H. Heydt]

W. H. Heydt, if technology is reaching its design limits, is it restricted in what it can do? Much the same as you can only do so much with steam power?

I suspect there is still life in steam power yet. We just don't spend any money developing it for everyday use, and after all, steam turbines produce all our electricity.

No gas turbine "peaker" plants?

I suppose one should note how far the principles of the Turbinia have fallen. The RN still can't catch her.

[W. H. Heydt]

Chips will go the same way as big CRT televisions. A slight change in the way things are done will do to chips what LED TVs did to televisions.

Jamesh, I was amazed to see on a recent TV documentary about Sellafield that steam turbines were used in nuclear power plants.

What? Did you think the reactor provided electricity directly? The reactor is just heat source...a replacement for coal or oil. (It's also one of the few energy sources we've got that isn't--even indirectly--derived from sunlight, and the others aren't being used very much so far. I will leave it as an exercise for the reader to determine how coal and oil are derived.)

[Woll]

W. H. Heydt, thanks for your reply. It's interesting to see your answers because I'm not in this field of work. I wondered why the Raspberry Pi never had a heat sink.

It's interesting to see what Graphene will do to chips.

And no, I didn't think they relied on steam in nuclear power plants. The program was very interesting.

[timrowledge]

The reactor is just heat source...a replacement for coal or oil. (It's also one of the few energy sources we've got that isn't--even indirectly--derived from sunlight,

I think that's actually not entirely true, since IIRC all those nicely radioactive elements were created in supernovae and so doin a sense stem from starlight.

[jamesh]

The reactor is just heat source...a replacement for coal or oil. (It's also one of the few energy sources we've got that isn't--even indirectly--derived from sunlight,

I think that's actually not entirely true, since IIRC all those nicely radioactive elements were created in supernovae and so doin a sense stem from starlight.

But the starlight came from fusion....everything came from the big bang really.


And I should add that I completely forgot about Hydro, wind turbines, solar, and certain gas power stations when I said all our electricity came via stream turbines. Apologies.

[W. H. Heydt]

The reactor is just heat source...a replacement for coal or oil. (It's also one of the few energy sources we've got that isn't--even indirectly--derived from sunlight,

I think that's actually not entirely true, since IIRC all those nicely radioactive elements were created in supernovae and so doin a sense stem from starlight.

But the starlight came from fusion....everything came from the big bang really.


And I should add that I completely forgot about Hydro, wind turbines, solar, and certain gas power stations when I said all our electricity came via stream turbines. Apologies.

Isn't there a project in Scotland to install a tidal power system? Or is that somewhere else and I have a memory leak?

[W. H. Heydt]

W. H. Heydt, thanks for your reply. It's interesting to see your answers because I'm not in this field of work. I wondered why the Raspberry Pi never had a heat sink.

The clock speed and power dissipation of the SoC the Pis use is low enough that simple convective airflow will keep it with its thermal limits.

And no, I didn't think they relied on steam in nuclear power plants. The program was very interesting.

That's interesting. Did you have particular idea how it was done or was it a case of "nuclear reactor' == something == "electric power out" without thinking about what the "something" is?

Please note...this is NOT a slur on your knowledge or intelligence. There are tons of fields about which I know, essentially, nothing. My only saving grace is being an avid reader (though not nearly as much now as I used to be), so I've been exposed to a fairly wide range of technology and (mostly) physical sciences.