Will Intel Thunderbolt kill SuperSpeed USB?

[PorscheDan]

http://www.semiwiki.com/forum/content/390-will-thunderbolt-kill-superspeed-usb.html

Personally I do not think so, for business reasons. I do respect the technology though.

PD

 

[996GT2]

It will not.

Why? Because USB 3.0 is backwards compatible with USB 2.0. This is a huge advantage in favor of USB 3.0.

I like to think of Thunderbolt as being similar to Firewire 800 back in the day. Was it faster than USB 2.0? Definitely. Was it nearly as widely adopted? No.

 

[BoT]

+1 .. mark his words

 

[Broheim]

thunderbolt? probably not
but Light peak will... Copper won't be able to compete with fiber, and if Intel can realise their goal of every widely used protocol running over light peak then it's a sure winner...

 

[PorscheDan]

thunderbolt? probably not
but Light peak will... Copper won't be able to compete with fiber, and if Intel can realise their goal of every widely used protocol running over light peak then it's a sure winner...

I thought Light peak and Thunderbolt were the same thing?

http://en.wikipedia.org/wiki/Light_Peak

PD

 

[khon]

I thought Light peak and Thunderbolt were the same thing?

http://en.wikipedia.org/wiki/Light_Peak

PD

People make the distinction because light peak was originally intended to be based on optical fibers, while thunderbolt (at least initially) uses copper cables.

 

[wuliheron]

Light peak was originally intended for use with fiber optics and the new thunder bolt version uses metal wire. It's a cheaper and more durable implementation of the technology meant for consumer use where the ability to download a movie in about 30 seconds is fast enough. Using fiber optics it could go as high as 100gb/s which is more useful for things like servers where you don't have to worry about a three year old breaking a fiber optic by tying it into knots.

 

[Broheim]

People make the distinction because light peak was originally intended to be based on optical fibers, while thunderbolt (at least initially) uses copper cables.

yeah exactly, I don't know if intel will change the name back to Light peak once they use fiber, but considering they renamed it to thunderbolt because they went with copper they probably will..

 

[ViRGE]

Besides backwards compatibility, USB is also much cheaper. TB requires 2 controllers in the host: a TB-to-PCIe bridge, and then a PCIe-to-<whatever> controller for the device in question. Two chip solutions rarely win.

 

[itsmydamnation]

Light peak was originally intended for use with fiber optics and the new thunder bolt version uses metal wire. It's a cheaper and more durable implementation of the technology meant for consumer use where the ability to download a movie in about 30 seconds is fast enough. Using fiber optics it could go as high as 100gb/s which is more useful for things like servers where you don't have to worry about a three year old breaking a fiber optic by tying it into knots.

its pretty useless for current Datacentre architextures as well, anyone can make a fast point to point link, making a fast lossless swtiching fabric on the other hand .

 

[WhoBeDaPlaya]

What wavelength does Lightpeak operate at again?
I'm thinking bending radius losses might be a problem for consumers.

 

[Arkadrel]

Will Intel Thunderbolt kill SuperSpeed USB?

No... why? because most plug-in devices dont require/benefit from the speed increase of thunderbolt.

Say you have a mouse... you want to plug it in... it could work on a thunderbolt port, but.... it ll work just as well on a usb one.

the only thing that really benefits is external hard drives, and those arnt fast enough yet to gain much/anything from useing thunderbolt.

So... until there are devices that clearly benefit hugely from it... it wont take off.


also...

Usb is backwards compatable, most people own alot of usb devices they want to be able to use, so motherboards would most likely still need usb ports.

When USB4 comes out it ll probably be faster than thunderbolt.... but then again maybe we ll see thunderbolt 2.0 version thats even faster! which is great... except we have no devices we need it for.


1) thing I believe could change it... is if thunderbolt is cheaper than useing USB, unless that happends and we all end up with alot of external HDD that need thunderbolt speeds... I dont think thunderbolt will ever amount to much.

Besides backwards compatibility, USB is also much cheaper. TB requires 2 controllers in the host: a TB-to-PCIe bridge, and then a PCIe-to-<whatever> controller for the device in question. Two chip solutions rarely win.

Intel will need to figour out how to many these chips tiny and cheap(er than usb) for TB to take off.

 

[khon]

What wavelength does Lightpeak operate at again?
I'm thinking bending radius losses might be a problem for consumers.

The wavelength is mid-infrared like most optical communication (~1500nm).

Anyway I don't think bending radius will be much of an issue. You won't be able to do sharp 90 degree turns, but you will be able to bend it quite a bit.

 

[Broheim]

one of the points of light peak is to have one unified interface with enough bandwidth to serve every need, instead of buying X number of 3rd party controllers, manufacturers (motherboard especially) could save money on just buying the one...

another advantage would be that optical fiber doesn't lose signal strength at the same rate as copper, so you aren't limited to, as an example, 3 metres as usb is.

also, as it's intel's own technology they don't have to deal with licensing fuss like firewire had to back in the day.

 

[Idontcare]

What wavelength does Lightpeak operate at again?
I'm thinking bending radius losses might be a problem for consumers.

Intel was quoted in one of those EETimes articles as saying the situation with lightpeak is not one of practical or engineering concerns but simply cost.

To implement lightpeak in its current form (not thunderbolt, which is like the backup plan) would add about $40 to the BOM...which means at the consumer wallet level it would add about $80 to the pricetag.

They need adoption rates to go up fast if they have any hope of getting out ahead of USB3. That won't happen if it makes an $80 pricetag difference to your $700 desktop.

So for now they've gone with plan B, seed the market with thunderbolt interfaces and hope the adoption rate is acceptably high to justify rolling out lightpeak itself in 3-4yrs time.

In so many ways this is all rather reminiscent of Intel's attitude and market approach when they wanted to go with Rambus over DDR.

 

[Dribble]

imo usb is likely to kill thunderbolt. Usb 3 (+ whatever comes after) is likely to be fast enough for pretty well everything but monitors and hard drives. Being as both of them have cheaper perfectly good standards with global support (display port and esata) thunderbolt is unlikely to win there.

The only use I can think of is an external pcie connector for laptops which is nice but it's not exactly a huge market.

 

[Phynaz]

Too expensive.

It's too bad, because it has great benefits. We've eliminated serial and parallel ports, and to some extent PS/2 ports. Problem is we have replaced them with a plethora of other ports - DVI, HDMI, lots of USB, 1394, eSATA.

I'd love to to have the promise of one or two ports on my machine that everything is daisy chained off. I just counted, and I have seven wires besides power connected to my PC. What a mess.

Replace that with one connection?
Yes, please.

 

[Arkadrel]

I kinda like that they look differnt...

but yeah... lets remove mouse/keyboard ports, and go all USB.
lets remove vga,dvi,hdmi and figour out which one to use and drop the rest.
lets just drop 1394 altogether.

whats wrong with sata cables?

 

[Cerb]

No... why? because most plug-in devices dont require/benefit from the speed increase of thunderbolt.

Say you have a mouse... you want to plug it in... it could work on a thunderbolt port, but.... it ll work just as well on a usb one.

the only thing that really benefits is external hard drives, and those arnt fast enough yet to gain much/anything from useing thunderbolt.

Have yo seen USB display adapters? Hello, Thunderbolt!

There really is no need to replace USB w/ TB. As long as TB does not suffer the royalties problem of FW, it should be quite capable of finding and holding on to niches, and may find itself integrated into PCI-e chips of various kinds. Being able to externalize PCI-e, with decent speed (10GB/s is enough for almost anything but performance video cards, before too much daisy-chaining), makes it quite a wildcard.

 

[Arkadrel]

Have yo seen USB display adapters? Hello, Thunderbolt!

Honestly before you mentioned it and I googled it no... apperntly there are many of them exsample below:

There really is no need to replace USB w/ TB.

Agreed, if there are no devices that benefit from its faster speeds.. .and it costs more, and you can run the same devices on both. Then thunderbolt is more expensive too? Why bother?

Problem is we have relaced them with a plethora of other ports - DVI, HDMI, lots of USB, 1394, eSATA.

Yes we dont need 1 more... sadly I see thunderbolt just ending up as yet one more of those ports (and the others will still be there too).

 

[Golgatha]

From a typical user perspective, USB 3.0 is many times faster than it needs to be for current external HDDs, it's backwards compatible with USB 2.0 devices, it is hot swappable, and it is cheap to implement.

Now let's compare the alternatives.

Lightpeak - faster but the speed isn't currently needed and probably won't be for quite awhile. It's interface isn't backwards compatible with almost a decade worth of computers currently in the wild as well (i.e. USB 2.0).

IEEE 1394 - Royalties and not ubiquitous.

eSATA - Pretty ubiquitous on newer machines, but you have to reboot your computer to connect an external drive. No thanks when there is a performance solution that's hot pluggable.

 

[Cerb]

eSATA - Pretty ubiquitous on newer machines, but you have to reboot your computer to connect an external drive. No thanks when there is a performance solution that's hot pluggable.

False. Both internal and eSATA ports will hotplug just fine.

eSATA's weakness is that power is not supplied with the data connector.

 

[Golgatha]

False. Both internal and eSATA ports will hotplug just fine.

eSATA's weakness is that power is not supplied with the data connector.

Yes hot pluggability is in the specification, but I have never seen it work in practice on a desktop board.

 

[Maximilian]

Only thing that will kill USB 3.0 is USB 4.0, i cant see this thunderbolt taking off due to USB's huge backwards compatibility advantage.

 

[Cerb]

Yes hot pluggability is in the specification, but I have never seen it work in practice on a desktop board.

What boards hasn't it worked on? I have yet to see it not work on desktop boards, across the following:
P35, P45, G33, Q35(or is that one Q33?), G41, 760G, 785G, 870, and on-board Jmicron, with Windows Vista, Windows 7, or Linux 2.6.31+.

Note that AHCI must be enabled before installing the OS (not strictly true, but it makes things easier).