News Rumour Control: Intel's Light Peak to ditch light

Intel's much-vaunted Light Peak technology will apparently be launching based on copper cabling.

http://www.bit-tech.net/news/hardware/2010/12/13/intel-light-peak-to-ditch-light/1

 

Well there goes backwards compatibility then!

 

I'm more worried about 10Gbsec^-1 internal storage devices, before external storage devices.

Give me a Light Peak SSD any day. Note "give", no way I'll pay the crazy high price tag for that!

 

ya well when I see it and it works well then I will be impressed.

 

"Initial" Versions?

This is probably to shorten the gap between USB3 and LightPeak, and because customers started calling. But how can you manage backwards compatibility when one uses electrical connections and the other is light?

LightPeak 2 anyone? Now with actual optics!


... on another note, was there ever supposed to be a power supply connection on the originally planned lightpeak so that the target device didn't need external power connectors?

Anyway I am simply inquisitive and skeptical.

 

I dont mean to make a '640k aught to....' type statement, but I fail to see exactly what intel intends to accomplish here? In the next decade, I can see the need for internal speeds of 10gbit/s (for SSDs), however I would think that SATA has one more revision in it, which should push it into the 1 Gigabyte/s range, which, when I stop to think about it, is pretty much the same speed as main memory (PC133) a decade ago.

External?
eSATA 1.5gbit/s is more than enough, USB3 will be nice if SSDs ever challenge HDDs on $/GB
I just finished getting all the devices in my house onto gigE ports. I'll look at 10gigE in 5 years or so.

Where does that leave lightpeak?

 

Honestly, I'd be more interested in lightpeak as a fast interconnect onboard between chips on a motherboard rather than external storage. Or perhaps even CPUs in a server environment.

But then again, no reason to link it to main memory, since work is being done to stack memory die on top of CPU die... making Chaches in the Gigabyte range.

Rambling now. Just odd really.

 

Presumable the speed they're intending to get is limited by something other than the medium used to shift the data, so someone pointed out they could just use copper wire and save piles of cash. That's my interpretation of this anyway.

 

There's nothing inherently expensive involved in LightPeak, whether or not it is a fibre-optic cable used.

The only cost to Intel is the R&D and they won't waste the work they've already done. If the non-optical version is launched then it just means a delay so that they can launch the optical version at a higher bit-rate. The plan was always to develop the standard far beyond the initial launch anyway. Launching a copper cable version will just mean that they can get the controllers and ports into products before they release an optical standard that they're happy with.

 

PCI express 3.0 is already 8 gbit/s for just 1x, never mind 16x.

infiband already has that covered

Interesting...

 

I just want Intel to stop stalling on USB3. It's becoming a joke!

 

It's going to be like Firewire. Theoretically better than USB, but never really goes anywhere.

 

I think a general user of a PC might be attracted to the optics of light peak if Intel market it properly, they'll see its completely different to USB and the speed gains that come with it as the reason to buy, if intel stick to copper they would probably relate to the USB name and buy that,the average user knows nothing about firewire but everybody knows about USB. perhaps they're talking about lightpeak using copper wire as the power source for the optics and somebodys put 2+2 together but isnt very good at maths.

 

Moving to fiber for 10gbit is just silly.

You can do that in copper just fine, and without needing any expensive transceivers. Latency is probably improved as well.

What tends to go over people's heads is that the huge advantage of fiber is very low attenuation and good resistance to interference, not speed or bandwidth. You can move a lot of data very quickly over copper. The only issues with copper arise when you get to long distances and have to fight attenuation and interference. At a couple inches, copper can move hundreds of gigabits per second, as seen in PCIe 3.0 speeds. At a hundred meters, you'll need good shielded cable to do 10 gigabits. At a kilometer, forget about it. Fiber won't even notice a kilometer. You need to get into the hundred km range before a significant amount of attenuation occurs.

But at the consumer level, dealing with a normal sized room, there's minimal benefit to choosing a fiber interconnect over a copper interconnect. It's just a whole lot of extra expense for little gain.

 

To an extent, yes. Apple shot themselves in the foot with Firewire, wanting a ridiculously high royalty ($1/port or so). Motherboard manufacturers didn't bite (aside from high-end boards targeted at bleeding-edgers and DAW folks), since the royalty on USB is $0 (<A HREF="http://www.usb.org/developers/usb20/faq20/">Clicky< /a>.

LightPeak may come on some enthusiast boards, but I doubt we'll ever see wide-spread adoption, especially after Intel pissed off every motherboard manufacturer on the planet by making them go off the farm for USB 3.0 chips. Intel was hoping to have everyone on board with their shiny new tech, but now that USB 3.0 is here (and backwards-compatible), there doesn't seem to be a lot of point to LightPeak, especially if it isn't released as an open standard. So far, I haven't seen any indication that Intel intends to do that.

 

Could this be thought as IDE to Sata 1 = USB3 to Peak?

How long has it taken to remove IDE ports from motherboards?

There's clearly room for one more and no one knows how much the royalty is. If Sony and Apple are interested then with them being THE two high end brands (expensive), maybe there is something in this Intel technology.

 

Think of it like going to a place that's a 200 miles away. You could drive there in 4 hours at 50mph. Or, you could drive a half hour to the airport, spend a half hour going through security, a half hour taxiing on the runway, an hour in the air at 200mph, a half hour getting off the plane, a half hour getting your baggage, and a half hour driving another car to the destination.

You think, well, airplanes are a lot faster and higher-tech than cars, so the second way is better. But it's not. It took just as long. The actual transmission was more efficient, but the amount of intermediate steps required was higher, so there was not a difference in overall time, and it was more expensive.

That's what using fiber for 10 meter or less connections like peripherals and stuff is like.

Now, let's change it up. What if your destination is 2,000 miles away? In that case, it would take 40 hours to drive there. Or, it would take 10 hours to fly there. The amount of time it takes to get to the airport and take off is the same regardless of how far you fly, 1.5 hours, as is the amount of time it takes to get out of the airport (1.5 hours), so your total time is 13 hours, it's much faster. Now flying is a much more appealing option. Even though the plane ticket is more expensive, the plane can fly nonstop. If you were to drive, you'd get tired and need to stop and rent a hotel room part of the way on the journey, which would be more expensive than just flying.

That's what using fiber for 100 meter or longer connections is like.

 

Very good post. I just want to know why mpg was mentioned?

 

LOL! I think he mean mph, and I think you knew that already?!

 

I would take the train. 200 miles @100mph = 2 hours.