CPU'S Where do we go from here ?

Now that CPU Die sizes are reaching the physical limit , currently 7nm for AMD, Nvidia etc and Intel not able to produce sub 10nm until 2021. The question becomes where will we go once CPU'S, GPU'S, APU'S reach 5nm ? or is 7nm already the financial limit ? It currently costs about $16 Billion to build a forge to manufacture sub 10nm chips, a forge to produce 5nm chips will probably cost $20 Billion and will be the last of its kind due to quantum effects on the chips themselves. So Moore's law no longer applies, we will have reached the plateau for CMOS fabrication. We now have a dilemma, do we go for more cores and parallelism , which will require rewriting code if we opt for IPU's (Intelligent Processing Units) as these are specialised and not general purpose. With Apple moving to ARM based processors in the near future it looks like Apple have opted for specialised IPU's to run Apple Software. But where will general purpose computing go, or is the writing on the wall ? Do we go with Intel's stacked CPU's or AMD's modular approach which relies on memory speed or do we revisit the Cell Chip ? I have wondered for a while now why it was that the Cell Chip seemed to vanish from mainstream existence when it threatened the very existence of Intel and AMD, who knows maybe they made a deal with IBM to slow that horse , but the Cell had promise and was already faster than Intel or AMD. Personally I am hoping for the resurrection of the Cell Chip, but the options for Intel and AMD are dwindling, we may be on the cusp of a complete technology change, a move to specialised hardware and the end of the general purpose PC we all know and love.

a reply to: Cosmic4life

Need a paragraph or two mate.

That body of text is hard on the old skyballs.

AMD has 5nm Zen 4 processors in the pipeline, so there is that.

Apparently the Cell Chips low manufacturing yield, high cost, and lack of affordable hardware systems other than the PS3 and the result in failure to achieve significant performance differences compared to existing x86 based architecture lead to its discontinued use.

The future is quantum computers. They're in the lab now and once perfected, it will be comparable to going from the stone-age to the jet-age. That's when AI will take off and humans will no longer do the thinking.

But who knows, maybe it will lead us to utopia!

a reply to: LogicalGraphitti

A desktop variant of a quantum computer based on nuclear magnetic resonance (NMR) is already available to purchase from SpinQ Gemini.

It's only a 2-Qbit system all the same.

Not suited to handle large quantum applications, but could help pave the way where learning about quantum computing and/or how to program such a system.

As to utopia, you don't want that mate, as it would equate to humanity's stagnation on the vine.

You see the belief that humans are perfect or could be perfect, leads inevitably to mistakes, when a perfect society is designed for an imperfect species.

a reply to: Cosmic4life

5nm and under from AMD and 3D Stacking from Intel. ARM and RISC-V architecture will become more popular. I feel we are still a way out from quantum etc..

a reply to: Cosmic4life

Now that CPU Die sizes are reaching the physical limit

I heard that sentence in the 90´s also.

I do not believe the development of the CPU is reaching its limit..
It is still a new thing history wise, and it is only the beginning.

The concept of "Time-Crystals", is taking off as we speak.

Its rather complex, but here is the article:

Time crystals enter the real world of condensed matter

Just yesterday, the first observation of interacting time-crystals was made.

First ever observation of 'time crystals' interacting

And i'm afraid that it wont change anything related to AI, just the learning speed of mimicking the wanted behavior.

a reply to: kloejen

Oh yeah, I remember a thread on time crystals 5 years ago I think.
I'll try to find it.

a reply to: Cosmic4life

there will be no quantum computers in the household in the near future-if ever.

Moore's law has reached it's peak when it comes to computers, the bleeding of electrons due to the ceramics in the resistors, the smaller the resistors the greater chance of electrical meltdown. In my opinion Moore's law is in palliative care at this point in time.

However I could be wrong, maybe there is an alternative, after all idiots blame the pyramids on aliens rather than human ingenuity-we'll figure it out one way or the othe other.

Or we could back engineer tech from the craft rovered from roswell *insert x-files theme*

a reply to: kloejen

Time-Crystals are fascinating, take for instance, unlike clocks or any other known objects in our universe.

These crystals derive their movement, not from stored energy, but from a break in the symmetry of time itself, enabling a "special form" of perpetual motion without the violation of any known physical law.

a reply to: Thecakeisalie

Back in the 60s, we assumed computers would take up whole buildings and require massive amounts of power.

Look at us now?

Moore's law, whilst important will simply be superseded by another concept.

Take for instance the invention of carbon nanotube transistors, or even spintronics, which may very well transcend said law if taken to fruition.

As long as I can play solitaire I don't care what size the processors are.

Seriously, I have been in IT for 32 years and my enthusiasm for what's coming down the road is gone. Completely. I envy some of you that still have the drive and the curiosity but truth be told I've only done IT for the career, my passion is music and classic cars.

I remember the early years when I was building and selling PCs. That was fun at the time, looking for the best, fastest, most powerful components and putting them together with the store name on them. We built a great reputation and business to a point, then the owner couldn't handle the growth and some of the people issues and it turned south. Fast. So I got out and went to a school as the IT director and the career took off at that point.

I have said I don't think any new products will be created that we don't currently have. Some took that wrong thinking I meant no new products at all. But with technology continuing to downsize we still only have CPUs, storage, input devices and such. Nothing new. New ways of doing things, but nothing new. If the cost of making equipment for these devices is so expensive, where is the advantage? How can they profit? If you're talking millions I could see it over the life of the product but billions? That's insane. I'm too old to be playing these financial games I guess.

a reply to: LogicalGraphitti

How? I've never really gotten the big deal about quantum computers. So...you've got some kind of special bits that can hold simultaneous values until they're read from then they collapse to one or the other.

As far as i've read, it's expected that quantum computers will be able to perform some algorithms quite a bit faster than classical computers, cryptography and encryption is usually what I see referenced, but will still fall short of classic computers with other algorithms.

If anything, future computers are likely to have a quantum coprocessor alongside a traditional processor, much like how older 8 and 16 bit computers used to have adding or floating point coprocessors.

originally posted by: dug88
a reply to: LogicalGraphitti

How? I've never really gotten the big deal about quantum computers. So...you've got some kind of special bits that can hold simultaneous values until they're read from then they collapse to one or the other.

As far as i've read, it's expected that quantum computers will be able to perform some algorithms quite a bit faster than classical computers, cryptography and encryption is usually what I see referenced, but will still fall short of classic computers with other algorithms.

If anything, future computers are likely to have a quantum coprocessor alongside a traditional processor, much like how older 8 and 16 bit computers used to have adding or floating point coprocessors.

Good point, and I agree. The computer processing ecosystem has undergone fragmentation such that rather than focusing on and building out a single, core, incredibly high-clocked main processor, other specialized chips have been introduced to compliment the CPU.

- GPU for video
- ASIC cards for graphics and intense, domain-specific mathematical computations (e.g. hashing, RSA key manip)
- Network stack chips mounted directly on the NIC
- TPU for crypto

I could see introducing quanta chips to supersede ASCIS, processors capable of extremely high rate of instruction throughput, useful for applied problem areas, hashing, cracking keys, astronomical modeling, etc

originally posted by: andy06shake
a reply to: LogicalGraphitti

A desktop variant of a quantum computer based on nuclear magnetic resonance (NMR) is already available to purchase from SpinQ Gemini.

It's only a 2-Qbit system all the same.

Not suited to handle large quantum applications, but could help pave the way where learning about quantum computing and/or how to program such a system.

I'd love to have one to play with but it requires some ridiculous cooling. If I had one though, the first thing I would do is work on decrypting Blockchain keys.

As to utopia, you don't want that mate, as it would equate to humanity's stagnation on the vine.

You see the belief that humans are perfect or could be perfect, leads inevitably to mistakes, when a perfect society is designed for an imperfect species.

Lol, I need a quantum computer to work that out complex statement 🤣

I agree with most of you as per Quantum Computing .... decades away. TSMC are apparently heading towards 3nm in the next 5 years, they have 5nm which will probably be rolled out in the new ARM based Apple Macs late this year, early next. 3nm is definitely the physical limit unless they find a way to prevent bleed over. I think we are going to see specialised CPU's , Apple CPU's in Macbooks, Chrome CPU's in Chromebooks, Qualcomm Snapdragon's in Phones etc I expect Microsoft will keep Intel and AMD alive for now. Graphcore have developed an IPU for AI Machine learning, AI will disrupt the market a lot sooner than Quantum Computers.

a reply to: HalWesten
The thing that really grinds my gears is that every time new speedier CPU's or faster drives or Memory came along, Software got more and more bloated. I remember when software was small and super efficient, it just did what it said on the tin, even now I prefer to run older versions as the speed is incredible, run new software and it sucks up all the hardware to a point where seems like I'm still using a Pentium 2 233mhz ..... software developers need a huge boot up the rear end for being so damn lazy and producing rubbish that suck away all the gains made in hardware.

a reply to: LogicalGraphitti

It's not that complex mate.

Essentially, you cannot polish a turd.

Because it falls apart at the seams.

No matter how pretty the toilet or proficient the polish.

We aspire towards perfection mate, but we never attain it, and if we did we would not have it for long and simply make a mess on the carpet.

next gen. organic cpus !!a reply to: Cosmic4life

Graphene as a base material instead of silicon.

Would allow another clock speed race.

Quantum tunneling effects are actually considered for every new small transistor CPU as is.

I don't see a die shrink below 5nm having significant gains anymore.

So...

They gotta do something different.