Intel's fastest connector to date uses light to transfer data at 1.6Tbps | Hacker News Reader

Intel's fastest connector to date uses light to transfer data at 1.6Tbps

recaply.com

48 pointsjlarkin35323 comments

Why does Intel even discuss these things? Can I buy this? Can I go out and buy products that use this standard? Can I buy PCs that use AMD CPUs that use this? What about ARMs? What about Infiniband-like networks?

It makes me angry that a large corporation set out on a Tuesday morning with the sole purpose to waste people's time like this. If it is not going to be on store shelves within the next few months and if it doesn't have broad industry support: NO ONE CARES.

This is why Thunderbolt has been a failure: they talked about it for over two years and the final version is Intel-only, effectively Mac only (very few non-Apple PCs have it, and it seems most of them were bought just to be Hackintoshes), and not nearly as awesome as what they originally described; and AMD, Texas Instruments, VIA, and several other VESA members added a USB3-over-DisplayPort feature called DockPort because Intel refuses to license and standardize Thunderbolt.

You can downvote me if you want, but I imagine at least half of us have been thinking this and someone had to just come out and say it: advertise your product when I can buy it, otherwise it just looks like you're trying to pimp your stock to investors.

If ^this^ is not FUD, I don't know what FUD is. Textbook example.

Seriously, all this post is about is: let's add a layer of confusion and fear on a technological advance. "waste people's time" - haha, nobody forced you to spent time on this. "not being on AMD CPUs, only on Intel Macbooks " - This is a press release about something working in the lab, let's not have anybody announce anything they got in the lab because I can't buy it.

Driving confusion much, are we now ?

He's right though.

Unless this is going to result in Intel products which actually have some broad market acceptance, who cares?

Because nobody cares about Thunderbolt. We were promised "fast enough to use external graphics cards" - you know, if you can accept max 4x PCI-e, and are willing to pay as much as a new GPU and computer to drive it for a suitable enclosure.

High speed data over fibre has been in the lab for years, and at these speeds and higher. Implementing it in a lab is easy. Getting it into a product people can use, and expect to use with others, is evidently harder.

Actually, if any would've been bothered to read the article, this is not consumer-grade technology. This is targeted at very special uses inside datacenters.

So we're talking apples and oranges here.

This is a replacement for something like infiniband. It's not likely to be consumer-level tech.

> because Intel refuses to license and standardize Thunderbolt.

Openness is definitely a big issue, and technical advantages may not be able to overcome proprietariness and cost. It's similar to what happened with Firewire (1394) and USB. The former has many technical advantages, but the official standards are not free. The latter is simpler and slower, with freely available standards, and as a result it has become ubiquitous.

Intel has a habit of putting out PR that mixes stuff that is available now, stuff that will be shipping this year, and stuff that may or may not be shipping five years from now. The press loves it but it's annoying and confusing to us technical people. (Whatever you do, do not pay any attention to "rack scale architecture".)

When it comes to silicon photonics specifically, Intel has been talking about it for years and still hasn't shipped anything.

Yeah what happened to that? It looked interesting and completely vanished off the face of the Earth.

Making silicon photonics cheaper than VCSELs must be really hard because it seems like nobody is shipping them. Or maybe silicon photonics are only cost-effective for parallel fiber optics which are a pretty small fraction of the market.

First, the headline number comes from 64 25gbps fibers in parallel.

I found more interesting this little bit buried in the bottom quarter of the article…

Intel is proposing an Optical PCI-Express (OPCI) protocol for use on optical wires.

I cant find this protocol, or any clarification if it is an open protocol. The article kind of feels like it is protocol intended for widespread adoption.

Cost effective, space effective optical connections would be nice. (No pricing in article. Not terribly encouraging hints.) A clean, open, PCIe protocol for fiber would be huge.

I presume that over the next decade(s), Cu may be phased out and replaced by Fibre on a large scale for information transfer purposes.

I do wonder about the effects on our environment though, namely wether making copper (metal) or Fibre cables (plastics) produces more waste, uses more or less energy and which one's easier to recycle, and to what degree.

Most of the high speed connection (10GigE+) for the last 3+ years have already been fiber, and anything longer than 100m is always fiber - it's pretty rare to see copper used for anything faster than 1 Gig connections.

They still use copper on 10gbit in some places for rack-local connections through both cat6 and passive sfp+ cables

Yup, and, ironically, my last 10Gig deployment (2009) was 100% copper from the server to the switches - but it was a pain in the butt trying to get it to work, and I don't think I would ever try doing anything but fiber between my network elements for 10 Gig.

Naïvely, glass is made from sand and copper is mined. And I can already recycle glass from my house, but not copper. So it looks like an environmental win to me. There may be something I'm missing.

Copper is pretty trivially recyclable, which is why you get reports of drug addicts stealing copper piping from people's houses to sell on the black market.

Pipes, wires, metal decorations, air conditioning condensers, old electronics...

And bridges!!! http://www.huffingtonpost.com/2012/04/30/bridge-stolen-in-cz...

Don't get hung up on the "green" characteristics of just the active component. You have to look at the whole picture, including packaging and manufacturing. It's not like the choice is between melting a little sand or an old bottle and digging a big hole in the ground to extract copper.

- There's a lot of plastic involved, for either medium

- optical stuff is pretty exacting to manufacture. It's easily possible that it's more energy-intensive to refine the materials, make the lasers, draw the fibers and so on than it is to make the copper equivalent

While it's definitely exacting, AFAIK, they still use plain ol' gravity to draw the individual glass fibers.

I think the real account is in the phenomenal energy costs to convert the mines material to a final form. The glass is going to be mined as well, as even the cheap soda lime glass we see everywhere has many ingredients, and I would be surprised if the typical fiber was made of such cheap material (went and checked, and the optical fibers are likely to be doped and tuned to the application). A furnace consumes huge amounts of energy, for both refineries and glass plants (I worked at a glass plant, and the size of the natural gas line running to it 24/7 full open was shocking). Recycling helps (since the material stores some of the original energy), but there still needs to be a process to remove impurities.

As a sibling comment noted, green energy probably isn't what we would hope it to be.

Copper wires are usually isolated with plastic :)

The ability to take my memory and stash it a rack away, optical can enable that

In optical fiber, at a minimum, every 3.3cm is another clock cycle of delay. (210^8 m/s / 3GHz / 2) (speed of propagation in optical fiber / clock rate / round trip factor).

That being said, being able to have hard drives* in other racks would be great, if nothing else just for vibration concerns.

Reply on news.ycombinator.com