Scientists build first light based hardware that competes with silicon

[Space Lynx]

https://www.techspot.com/news/79737...st-light-based-hardware-competes-silicon.html


"The technology is closer than you think to being released"


this is interesting...

 

[SomeOne99h]

Waaaait! So RPG colours that was used for the look only, will become the hardware itself XD ... Such unexpected turn of events!

 

[FordGT90Concept]

It is possible to do math operations using wavelengths of light. The limitations are in interpreting the waves back into electrical impulses. I'm not holding my breath for this but the prospects are interesting.

I think this is better news for cheap networking equipment and interconnects than data processing. Like...maybe fiber optic-in-home networks can now become a reality.

 

[John Naylor]

Is it solar (light) powered ?

 

[moproblems99]

fiber optic-in-home networks

Not seeing a benefit for this. Is this a just because type of thing? Latency is already practically non-existent in a LAN (sub 1 mil). Gigabit bandwidth is already sufficient for most people but I suppose if most ISP ever get off their ass and make higher speed access more affordable then Gb will not be sufficient in many cases.

 

[FordGT90Concept]

The reason why bandwidths greater than 1 Gbps is rare is because equipment faster than that is so expensive to produce. Bandwidths greater than 40 Gbps rely on parallelism. These new transistors can cheaply and efficiently gate light entering/leaving fiber lines.

 

[moproblems99]

The reason why bandwidths greater than 1 Gbps is rare is because equipment faster than that is so expensive to produce. Bandwidths greater than 40 Gbps rely on parallelism. These new transistors can cheaply and efficiently gate light entering/leaving fiber lines.

40 Core CPUs are 'rare' too but it isn't because they are expensive. That's part of it but there is that whole need thing too. This is especially true consumers but is also true for most businesses. Not many do things that require more than a Gb.

 

[Athlonite]

So now I can look forward to RGB CPU & Ram and DNA storage whats next the Unicorn fart bus that runs at the speed of a rainbow

 

[FordGT90Concept]

40 Core CPUs are 'rare' too but it isn't because they are expensive. That's part of it but there is that whole need thing too. This is especially true consumers but is also true for most businesses. Not many do things that require more than a Gb.

If you're doing a file transfer between two modern systems over a gigabit network, you won't see performance better than 113 MB/s because the network is saturated. Your average computer has a hard drive or SSD capable of double that these days. Home wired network speeds haven't improved in 15 years.

 

[NdMk2o1o]

40 Core CPUs are 'rare' too but it isn't because they are expensive. That's part of it but there is that whole need thing too. This is especially true consumers but is also true for most businesses. Not many do things that require more than a Gb.

This is true, also Bill Gates said that 640k memory is more than anyone will ever need

 

[hat]

If you're doing a file transfer between two modern systems over a gigabit network, you won't see performance better than 113 MB/s because the network is saturated. Your average computer has a hard drive or SSD capable of double that these days. Home wired network speeds haven't improved in 15 years.

Is there really a need to improve beyond gigabit? Even 100mbit was fine until hard drives started getting fast and SSDs became common, outside of rare conditions. These days when everyone is streaming HD content everywhere, we have found a reason for the common user to have gigabit, or at least, fast WiFi beyond N. I think we need a good few more years before everything else catches up to present a need to go beyond gigabit.

 

[moproblems99]

If you're doing a file transfer between two modern systems over a gigabit network, you won't see performance better than 113 MB/s because the network is saturated. Your average computer has a hard drive or SSD capable of double that these days. Home wired network speeds haven't improved in 15 years.

Don't disagree but transferring 60gigs doesn't take that long and people aren't doing it every day. Some people sure but not enough to payoff the R&D, manufacturing, packaging, and marketing.

 

[Space Lynx]

light based hardware is arriving sooner than we thought

Nvidia's eyeing up ways to beef up its future AI chips with GPUs on top of GPUs and the power of light

That's Nvidia's prediction. Mine is more profits.

www.pcgamer.com

The two big innovations in this picture, insofar as they could be applied to AI accelerators, are silicon photonics and the vertical stacking for logic. The former uses photons (light) to transmit data to and from optical components, which is faster and uses less power for more bandwidth than traditional electrical data transmission.

 

[Space Lynx]

won't let me edit my last post, but it looks like everyone is investing in light based chips that will eventually replace silicon

Nvidia, AMD, and Intel invest in Ayar Labs to enable futuristic light-based chips

Ayar Labs is making waves by shrinking fiber optic data transmission technology to the chip scale. Its flagship product, the TeraPHY optical I/O chiplet, delivers a staggering...

www.techspot.com

probably is the future at this point

 

[wheresmycar]

If photonics has the potential to redefine computing, its going to be a long journey before it gains momentum, add a decade more for enterprise inception and then add another decade longer before consumer products reach the market. By that time, I’ll probably be 100 and gone. Progress doesn’t belong to some distant future tiktokiums, its now or never!!

 

[qxp]

I think this is better news for cheap networking equipment and interconnects than data processing. Like...maybe fiber optic-in-home networks can now become a reality.

Actually, it is not that expensive now - 10Gb hardware is downright cheap, 40Gb is cheap too if you buy used. The datacenters are switching to 200Gb networking, so there is fair amount of older gear on sale.
You can put in fiber optic cables yourself, the wall jacks are not that expensive either. I would suggest to go with single-mode fiber as it scales well from 10Gbit to 200Gbit, which eventually will become affordable too.

Not seeing a benefit for this. Is this a just because type of thing? Latency is already practically non-existent in a LAN (sub 1 mil). Gigabit bandwidth is already sufficient for most people but I suppose if most ISP ever get off their ass and make higher speed access more affordable then Gb will not be sufficient in many cases.

If you're doing a file transfer between two modern systems over a gigabit network, you won't see performance better than 113 MB/s because the network is saturated. Your average computer has a hard drive or SSD capable of double that these days. Home wired network speeds haven't improved in 15 years.

Indeed, modern PCIe SSD has bandwidth of several GB/s, which can saturate even 10Gbit connection. But you also win because optical fibers don't conduct electricity and reduce susceptibility to surges from lightning.

 

[KLiKzg]

Well, this gives so much opportunities. As the light can be in different lengths, so we can have simultaneous computing with same hardware.

But also, you do not have to have so much of the large processors, sorry LPU; as those apparatus would work not in GHz, but in several THz.

 

[Space Lynx]

If photonics has the potential to redefine computing, its going to be a long journey before it gains momentum, add a decade more for enterprise inception and then add another decade longer before consumer products reach the market. By that time, I’ll probably be 100 and gone. Progress doesn’t belong to some distant future tiktokiums, its now or never!!

Who knows what the future holds with so much intellectual property and young geniuses out there

 

[qxp]

Who knows what the future holds with so much intellectual property and young geniuses out there

Intellectual property actually slows things down - you get these big explosions of activity when some key patents expire and suddenly everyone can build.