Fiber Optic Thunderbolt Cables Begin Mass Production, Available Up to 30m in Length

[TMay]

Originally Posted by The Deepness

The 40-channel capability may be imposed by the TB technology. MMF is capable of handling up to 1700 or so channels simultaneously.[COLOR="#808080"]

I should have been more specific. Copper may not be practical for 40 channels.

Either way, Intel controls the silicon, and they may have decided that the given roadmap looks practical and economic to 40 channels/100Gb and 30m.

[torana355]

Originally Posted by RBR2

Keep reading, starting with the white paper. Putting your head in the sand may be comforting, but ignores reality. This is not an Apple specific problem. It is a problem across the board.

As to Apple products, different ones exhibit problems to varying degrees with a variety of products.

Which USB 3 manufacturer whose products are a problem do you work for. You certainly are behaving like a shill.

I own multitude of different usb3 devices and none of them have any of those issues. We use many USB3 devices at work and no issues with our wifi there either. Using properly shielded cables stops any interference. I work in the printing industry so i don't work for any usb3 manufacturer...

[deconstruct60]

Originally Posted by The Deepness

When installing or using fiber optic cables, you always have to go by the cable manufacturer's specs as to how tightly you can bend a cable,
....
It'd be interesting to learn if Sumitomo Electric Industries has developed a more durable fiber material, or new method of grading or stepping the refractive indices of the core and cladding, such that the cable can withstand such macrobending without significant attenuation due to dispersion.

The likely technology was covered in Intel demos back when this was code named 'Lightpeak'. Corning invented the fiber a while ago.

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

The increased bend performance has to do with the internal reflective casing the fiber is wrapped in.

Somewhere there is an Intel video demo where the cable is almost tied in a knot and still works. The factor is cost not whether it can be done in an affordable enough fashion for Thunderbolt purposes. It would not be used in very high bandwidth like telcom/network backbone transport ( which relatively Thunderbolt is not. It is only high in comparison to the legacy protocols it typically is deployed to transport. ).

[deconstruct60]

Originally Posted by RBR2

The second generation Thunderbolt is supposed to debut in the first half of 2013.

According to what source? As recently as a year ago Intel was saying this:

"... As a result, Intel told me that we won't see any increase in Thunderbolt speeds for the next two years. ... "
http://www.anandtech.com/show/5405/t...likely-in-2014

We may see a 'dog and pony' show at one of the 2013 Intel Developer forums. However, it is highly unlikely will see a bump in actually deployed Thunderbolt systems before 2014. That will be similar to the 'dog and pony' Lightpeak ( and fiber based tentative USB 3.0 ) demos that Intel years before release of Thunderbolt (and the converted back to copper USB 3.0).

There several huge problems with bumping Thunderbolt too quickly.

1. The current form barely has traction. If basically tell folks the current stuff is being superseded then more than a few folks will just skip it.


2. Bandwidth limitations. Most designs that computers are leveraging right now are limited to x4 PCI-e v2.0 lanes off the IOHUB chip. There is no PCI-e v3.0 to tap into with most Core i5/i7 designs. Until Intel's mainstream core reference design come up with more PCI-e v3.0 lanes there is not "excess" internal bandwidth to export to a faster Thunderbolt.

3. "Faster than PCI-e v3.0" is going to run into many of the same interference problems that inexpensive and botched USB 3.0 problems have (e.g., 'corner cutting' inexpensive cables kicking off gobs of noise , RF problems at the connectors , etc. )

It is supposed to be cheaper, and faster the the current Thuderbolt and use fiber optic cabling.

The current standard would be cheaper if the component deployments were scaled up more. Going to yet another generation at even lower scale than current Thunderbolt levels would far more likely increase costs; not lower them.

Likewise doing a die shrink on the current technology and incorporating a few more discrete parts into the core TB controller package will also reduce costs. Again going to new standard would likely run counter to higher integration and pushing die size down. Again reflective more so of minimally status quo costs ( if not a net increase).

Getting the cost of Thunderbolt down is imperative to its acceptance and success. At this time, USB 3 is experiencing major problems that provide the perfect opportunity for Thunderbolt 2 to overake it and simply replace it.

This is the delusion FUD that has been floating around since Lightpeak demos a couple of years ago. Thunderbolt was never intended to and never will displace/replace USB. Seriously, that notion isn't even being spun by Intel anymore ( corporate wide it never was. There were some 'Lightpeak' folks selling Kool-aid initially but even they have stopped drinking their own Kool-aid. )

Cost is the only advantage which USB 3 ever had over Thunderbolt

No. Compatibility is the major advantage. Cost was secondary to that.

Getting a specification established so that multiple vendors can get into production (and competition which will help with pricing) is an important step in the process, the importance of which can not be overestimated.

Does Intel really want an open, community standard? It isn't now. There is nothing to indicate there is a major spin-out Intel is gearing up for.

[deconstruct60]

Originally Posted by el-John-o

I would love to see thunderbolt realized as an ultra high performance network interface (sort of like Fibre Channel).
....
Gigabit ethernet is becoming outdated in some environments.
....
Oh the possibilities!

Possibilities that already exist. Both Fiber Channel and Gigabit Ethernet were put in the rear view mirror several years ago for true big data shops.

It is expensive but 10GbE and Infiniband solutions already exist and are deployed. Fiber Channel over Ethernet FCoE also is increasingly deployed.

There was a hope that somehow Thunderbolt (TB) would deliver 10GbE or 10+ Gb Infiniband bandwidth at a much more affordable price point but that really hasn't surfaced yet. If the prices for this optical cable made a revolutionary difference I suspect they would have been attached to the story.

More likely long fiber TB runs will have a bigger traction in deplacing extended Firewire runs than contemporary very high bandwidth alternatives. It may also help 10GbE switch (and per port) costs come down some finally to.

Of course with these long thunderbolt cables, several users sharing one computer already seems around the corner. 'Thin clients' have existed for a long time, but Thunderbolt opens the door for these types of computing systems to work with just one cable,

No it doesn't open the door. TB is a general networking protocol. It is flexible in deployment but it is still has much of the hierarchy elements that PCI-e ( and lessor extent USB ) exhibit.

These long TB cables bring no huge increase in Bandwidth. People seem to labor under the misconception that somewhat TB Bandwidth goes up with the number of ports. It doesn't. The more 'clients' on a TB daisy chain the more the bandwidth is shared/split between those clients.

Imagine a small photography or videography studio being able to buy just one loaded Mac Pro, and a few thunderbolt displays, and being all set with several powerful workstations at a fraction of the cost?

If all the studio needed was 2-3 2005-2007 era Mac Pros than yeah it is feasible to split up a 2013-2014 Mac Pro into slices to be shared. All you have "rediscovered" is the utility and efficiency of timesharing computers concept from the 70's (and late 60's). But if the workload demands go up to match the modern machines then would still need 3 new workstations over time.

Multiple VNC sessions to a Mac Pro is doable now with 1GbE switch and a multiple port card addition to the "server" Mac Pro. Thunderbolt doesn't add much there and TB displays are the 'tail wagging the dog' (just there to drag in Thunderbolt).

[KnightWRX]

Originally Posted by deconstruct60

If all the studio needed was 2-3 2005-2007 era Mac Pros than yeah it is feasible to split up a 2013-2014 Mac Pro into slices to be shared. All you have "rediscovered" is the utility and efficiency of timesharing computers concept from the 70's (and late 60's). But if the workload demands go up to match the modern machines then would still need 3 new workstations over time.

The problem is also that TB support is not written in such a way. Plugging 4-5 TB displays into a Mac Pro does not result in 4-5 user sessions, but rather 1 user session spread accross 4-5 monitors. A lot of people are truely misunderstanding Thunderbolt and what its actual purpose is. It's not a single cable thin client protocol, it's not a networking stack to hook multiple storage arrays to multiple client machines, it's not even a networking stack to hook up multiple nodes together.

It's a peripheral bus local to a single host. Anything else you want to make it do requires using that peripheral bus to hook up actual peripherals to it that perform the needed function. Need networking ? Hook up Ethernet to your Thunderbolt bus... etc...

[rmwebs]

Originally Posted by apple-win

<snip - bit long to quote the whole thing >

Charles just used the existing technology and made it work in the telecommunications market. He didn't invent it. Not that it detracts the great work he did, but nonetheless it would be wrong to give him credit for creating fiber optic technology.

http://en.wikipedia.org/wiki/Charles_K._Kao

pioneered in the development and use of fiber optics in telecommunications

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

Fiber optics, though used extensively in the modern world, is a fairly simple, and relatively old, technology. Guiding of light by refraction, the principle that makes fiber optics possible, was first demonstrated by Daniel Colladon and Jacques Babinet in Paris in the early 1840s. John Tyndall included a demonstration of it in his public lectures in London, 12 years later.

A number of people had used the technology long before Charles. Notable people include John Logy Baird, Alexander Graham Bell, Clarence Hansell to name a few.

Getting back to Charles though. When he did work with Fiber, it was under a British company, so my point about it not being US technology still stands.

[TMay]

Originally Posted by KnightWRX

The problem is also that TB support is not written in such a way. Plugging 4-5 TB displays into a Mac Pro does not result in 4-5 user sessions, but rather 1 user session spread accross 4-5 monitors. A lot of people are truely misunderstanding Thunderbolt and what its actual purpose is. It's not a single cable thin client protocol, it's not a networking stack to hook multiple storage arrays to multiple client machines, it's not even a networking stack to hook up multiple nodes together.

It's a peripheral bus local to a single host. Anything else you want to make it do requires using that peripheral bus to hook up actual peripherals to it that perform the needed function. Need networking ? Hook up Ethernet to your Thunderbolt bus... etc...

I suspect that people are looking at a one size fits all solution, whereas I consider TB, Displayport, USB 3.0, Wifi, BlueTooth and Ethernet as complementary for OSX. Similarly, USB 3.0, Lightning, Wifi, cellular networks, Bluetooth and soon enough, NFC, are complementary for iOS.

Most people here would find a mix of all or almost all of these Industry standards, excepting Apple's proprietary Lightning solution for iOS, in daily use of Apple/Windows/Android products.

TB, embraced early by Apple, will become more widely available for Windows users with Intel Haswell support, which should increase the number and kind of peripherals available for Apple users as well, leaving Lightning as the single proprietary connection solution from Apple.

[MadDawg2020]

Originally Posted by Drew017

perfect for connecting to that thunderbolt display in the next room

Sure if you dont mind paying more for the cable than the display...
AND if you happen to have one of the handful of Thunderbolt displays...

----------

Originally Posted by Godgem

Its good for conecting cameras for live streaming event for exemple...

Okay but the problem is but there is not a single camera made that supports Thunderbolt and if by chance there was you would have to take out a mortgage to get one.

There are already far too many connection standards for Video as it is.
BNC, Composite, Component, SCart, S-Video, VGA, RGB, HDMI, USB, iLink, Firewire to name a few... and most of these are in addition to audio and time sync connections. As video technician you need a trunk full of cables, adapters and accessories to connect things now, there will be a full out revolt if they add more!

[KnightWRX]

Originally Posted by TMay

TB, embraced early by Apple, will become more widely available for Windows users with Intel Haswell support, which should increase the number and kind of peripherals available for Apple users as well, leaving Lightning as the single proprietary connection solution from Apple.

Except outside Apple forums, even Intel doesn't consider TB is be a mass-market consumer application. Their own project managers see Thunderbolt as a completement, not a replacement, for USB 3.0. Both are to serve different purposes.

The solutions, albeit expensive, that we've seen is what Thunderbolt is being aimed at. Replacing things like eSATA, external SCSI, Firewire and allowing things like laptop docks and PCIe type expansion like you see on desktops only through wires rather than motherboard slots.

[cirus]

It would be nice to see them push to make thunderbolt widespread and cheaply available as opposed to this.

[AidenShaw]

Originally Posted by RBR2

Keep reading, starting with the white paper. Putting your head in the sand may be comforting, but ignores reality. This is not an Apple specific problem. It is a problem across the board.

We read the white paper - it acknowledges the possibility of a problem, and some reasonable things to do to avoid it (like shielding connectors and cables).

Chill on the usb3gate doomsday rhetoric - the rest of us seem to be doing just fine with our USB 3.0 peripherals.

One possible difference, though, is that I don't use Apple products, and use copper ports almost all of the time. WiFi sucks, so happy to have GbE ports on the docking stations for my laptops at work and at home. Bluetooth keyboards/mouses - never have been "snappy". The only time I turn on the WiFi on my laptop is if I'm in a conference room - which means that I probably didn't drag my USB 3.0 peripherals along to the meeting.

The only thing wireless in my typical day (besides my 4G Thunderbolt) is that the phone's Bluetooth is on for the hands-free in the car.

Originally Posted by RBR2

Which USB 3 manufacturer whose products are a problem do you work for.

Priceless - accusing anyone who disagrees with your "USB 3.0 nightmare" scenario to be a paid astroturfer.

[knewsom]

Dare I say this is what they've been waiting on to update the Mac Pro...?

[repoman27]

Originally Posted by SeattleMoose

So the Thunderbolt glacier has moved another inch....yawn.

I'll be fossilized by the time this "new" technology is actually mainstream and affordable.

Maybe that's because you can't accept that Thunderbolt is intended to be neither. It's an advanced interface for mobile and desktop machines that carries a premium due to its very nature. Some people are willing to pay more for something that is potentially faster/better. For everyone else, it continues to work as just a regular digital display output port at no additional cost.

Originally Posted by TMay

I should have been more specific. Copper may not be practical for 40 channels.

Either way, Intel controls the silicon, and they may have decided that the given roadmap looks practical and economic to 40 channels/100Gb and 30m.

Thunderbolt is 2 full-duplex channels per port. There is nothing in the roadmap for increasing it beyond that. Also when it was released, at 10.3125 GT/s per channel, it was about as fast as anything on the market, even in the datacenter or for the telecoms. A short list of standards that have faster channel speeds includes the likes of FDR InfiniBand at 14.0625 GT/s and OC-768/STM-256 at 39.81312 GT/s, but these aren't the types of things you tend to find on consumer notebooks and desktops, even at the high-end. Currently available VCSELs can only manage 25-40 Gbit/s per channel, regardless of how many channels you can squeeze down one fiber if you use WDM.

At this point, Thunderbolt is not in any way limited in bandwidth by using copper cables (or connectors, for that matter). The 4-channel Thunderbolt controllers that have been released thus far have the same bandwidth on the back end as the PCH on the systems they've been paired with (i.e. PCIe 2.0 x4 + DP 1.1a x2 = DMI 2.0 x4 + FDI).

Let's take a look at the total non-memory bandwidth of a Sandy Bridge LGA 1155 CPU:
PEG - PCI Express 2.0, 5.0 GT/s x16, 8b/10b encoding, 64 Gbit/s
eDP - DP 1.1a, 2.7 GT/s x4, 8b/10b encoding, 8.64 Gbit/s
FDI - DP 1.1a, 2.7 GT/s x8, 8b/10b encoding, 17.28 Gbit/s
DMI - DMI 2.0 (PCIe 2.0), 5.0 GT/s x4, 8b/10b encoding, 16.0 Gbit/s
Total = 105.92 Gbit/s

Ivy Bridge brings us PCIe 3.0 so we have:
PEG - PCI Express 3.0, 8.0 GT/s x16, 128b/130b encoding, 126.03 Gbit/s
eDP - DP 1.1a, 2.7 GT/s x4, 8b/10b encoding, 8.64 Gbit/s
FDI - DP 1.1a, 2.7 GT/s x8, 8b/10b encoding, 17.28 Gbit/s
DMI - DMI 2.0 (PCIe 2.0), 5.0 GT/s x4, 8b/10b encoding, 16.0 Gbit/s
Total = 167.95 Gbit/s

So in a 2-port scenario, 1st gen Thunderbolt already provides external I/O channels equivalent to 43% of the total bandwidth of the CPU in the case of SNB and 26% for IVB.

While a 10 Gbit/s channel may not be fast enough for external GPU applications or certain exotic storage systems, it is still very fast in the context of the platforms it is designed for.

Originally Posted by MadDawg2020

Sure if you dont mind paying more for the cable than the display...
AND if you happen to have one of the handful of Thunderbolt displays...

----------



Okay but the problem is but there is not a single camera made that supports Thunderbolt and if by chance there was you would have to take out a mortgage to get one.

There are already far too many connection standards for Video as it is.
BNC, Composite, Component, SCart, S-Video, VGA, RGB, HDMI, USB, iLink, Firewire to name a few... and most of these are in addition to audio and time sync connections. As video technician you need a trunk full of cables, adapters and accessories to connect things now, there will be a full out revolt if they add more!

The copper Thunderbolt cables also contain the transceivers in the connectors, and the cable itself is much more complex than a simple 4 strand fiber version would be. The only reason these cables would be priced considerably higher would be due to the size of the target audience and what it's willing to pay. Also you don't need to have a Thunderbolt display to use this cable; any two-port Thunderbolt device can be placed at the end and would allow the connection of any display with a resolution up to 2560x1600.

The primary market for an optical Thunderbolt cable is based on Thunderbolt's DisplayPort capabilities, not its PCIe bandwidth (although that certainly is an added bonus). Digital signage, displays used on stages and sets, remote monitors for video professionals, etc. You'll notice that Sumitomo showed off samples of these cables extensively at NAB.

And, BTW, the Blackmagic Cinema Camera is available in two models, albeit to a limited degree, for $2995. And the interfaces you mentioned have almost all gone the way of the dodo as far as video is concerned. Analog is completely legacy at this point. For digital video interfaces you have HDMI for consumer electronics, DisplayPort for PCs, and SDI for professional gear. For data syncing you used to have USB 2.0 and 1394a/b, now you have USB 3.0 and Thunderbolt.

[el-John-o]

Originally Posted by deconstruct60

Possibilities that already exist. Both Fiber Channel and Gigabit Ethernet were put in the rear view mirror several years ago for true big data shops.

It is expensive but 10GbE and Infiniband solutions already exist and are deployed. Fiber Channel over Ethernet FCoE also is increasingly deployed.

There was a hope that somehow Thunderbolt (TB) would deliver 10GbE or 10+ Gb Infiniband bandwidth at a much more affordable price point but that really hasn't surfaced yet. If the prices for this optical cable made a revolutionary difference I suspect they would have been attached to the story.

More likely long fiber TB runs will have a bigger traction in deplacing extended Firewire runs than contemporary very high bandwidth alternatives. It may also help 10GbE switch (and per port) costs come down some finally to.









No it doesn't open the door. TB is a general networking protocol. It is flexible in deployment but it is still has much of the hierarchy elements that PCI-e ( and lessor extent USB ) exhibit.

These long TB cables bring no huge increase in Bandwidth. People seem to labor under the misconception that somewhat TB Bandwidth goes up with the number of ports. It doesn't. The more 'clients' on a TB daisy chain the more the bandwidth is shared/split between those clients.




If all the studio needed was 2-3 2005-2007 era Mac Pros than yeah it is feasible to split up a 2013-2014 Mac Pro into slices to be shared. All you have "rediscovered" is the utility and efficiency of timesharing computers concept from the 70's (and late 60's). But if the workload demands go up to match the modern machines then would still need 3 new workstations over time.

Multiple VNC sessions to a Mac Pro is doable now with 1GbE switch and a multiple port card addition to the "server" Mac Pro. Thunderbolt doesn't add much there and TB displays are the 'tail wagging the dog' (just there to drag in Thunderbolt).

Well a lot of that was how thunderbolt, if realized, could make these same things EASIER.

GbE is more than capable of handling RDP tasks, and presently there are lots of people sharing a single high performance workstations with seperate video, some sort of USB hub, and whatever else is needed. Thunderbolt DOES give the ability for one capable to be routed to each workstation, which is great for simplifying situations where, in fact, a single high performance workstation is better than several consumer grade machines, and yet high performance machines at each workstations are not feasible.

Yes, it's existing technology. Heck, Steve Jobs demo'ed it with the original iMac a bazillion years ago. Thunderbolt could simple make it cleaner and easier. I'm not talking about next week, I'm talking about where thunderbolt could be realized in the future. Thunderbolt being less limited to the types of roles it can have is where the genius is. Multiple cables replaced with one is what it's all about, in this case.

As far as FibreChannel and such, I still think TB could, in a few years, be poised to replace it as a low cost similar performing alternative. The coolest thing about that, in my mind, is actually going to be in the small scale, not the large scale. I know of a few individuals who are in fields where they have a large storage array connected to several workstations. Using anything from GbE to FibreChannel. In theory, a thunderbolt 'network' could deliver as good or better performance at a lower cost. Heck, even if the rack-side equipment cost exactly the same, the ability to connect directly to a workstation (assuming Xeon support for thunderbolt and thunderbolt on future Mac Pros) without having to buy very expensive interface cards would save several hundred to well over a thousand bucks per client.

I'm not an enterprise level IT, I'm not an IT professional at all. Others like yourself obviously know a lot more about this stuff than I do, but nonetheless I am excited about what thunderbolt can become. Right now it's fantastic for what it is, enterprise level I/O performance on consumer grade machines. It's an odd 'trickle down'. When we were on 10BaseT and the servers ran 100BaseT, the servers upgraded to 1000BaseT and we got 10/100 LAN, then we got gigabit and the servers moved on to 10GbE, FiberChannel, etc. Now, we sort of kind of got FiberChannel/10Gbe, only in the form of a new I/O with similar performance. I think cost will go down in the future.

Or it could crash and burn, we'll see. But I hardly think FireWire was a failure either, as others seem to imply. It never took off in the mainstream, but why should it? Why does the typical user putting word documents on a thumb drive or an occasional song on an iPod need fast I/O with higher cost? USB makes more sense for most people. Before USB 3.0, FW800 was the De Facto standard for YEARS in many professional workstation environments. I've got a good friend who works in marketing/advertising and up until very recently when you step into that building you step into a sea of firewire cables and external firewire RAID arrays (now Thunderbolt and USB 3.0). It's a high speed I/O intended for individuals who needed to move large files quickly, and who had to do that often. Those in that market DID use FW800/400, and used it a lot.

[deconstruct60]

Originally Posted by el-John-o

Well a lot of that was how thunderbolt, if realized, could make these same things EASIER.

GbE is more than capable of handling RDP tasks, and presently there are lots of people sharing a single high performance workstations with seperate video, some sort of USB hub, and whatever else is needed.

GbE RDP isn't doing to full screen hi-res video very well. But yes for mainstream, 2D, desktop work it works fine as long as don't choke the network bandwidth path between server and client.

Multiple video card output per user isn't very common. The number of folks using RDP/"Terminal Services" is substantially higher. There is a corner case with Microsoft Multipoint Server Stations ( http://technet.microsoft.com/en-us/l...omponents.aspx ) that could be 'cleaned up' by combining the video and usb link onto one cable but that's a very narrow market.

There are two other MMServer set-ups: USB zero and RDP over LAN. USB zero will pick up much more deployment flexibility when it comes to video with the shift to USB 3.0. Rapidly changing video is likely still a problem but the working range is larger. RDP over LAN would simply work better on 10GbE than 1GbE. 10GbE will likely get cheaper in the future. 1Gb Wifi will certainly get cheaper in the future and that requires no wires/cables.

MMServer's Direct-video and USB zero modes do have a distance problem. Thunderbolt doesn't have exclusive distance benefits; quality optical fiber does. Fiber isn't exclusive to TB.

Should someone add Thunderbolt to MMServer's repetoire? Sure. Is that going to revolutionize Direct-video deployments in number and quantity? Probably not.

The huge disconnect here is that Apple doesn't have a MutliPoint Server solution like this. The closest is Lion/Mountain Lion multiple session VNC based solution.

http://appleinsider.com/articles/11/...screen_sharing

VNC is firmly rooted in Ethernet ( wired or wireless ) based deployment. A hack could be layered on top of Thunderbolt to leverage it, but Thunderbolt isn't really an "Direct video server" enabling solution.

Thunderbolt DOES give the ability for one capable to be routed to each workstation, which is great for simplifying situations where, in fact, a single high performance workstation is better than several consumer grade machines, and yet high performance machines at each workstations are not feasible.

It is doubtful Thunderbolt is going to route more than 3 ( limit is 2 now) video steams. Its Display Port stream size is extremely likely going to be limited to the number of screens the vast majority of folks attach to computers. That number is <3 now but they could bump to 3 for "power users", but it is unlikely to go past that.

Yes, it's existing technology. Heck, Steve Jobs demo'ed it with the original iMac a bazillion years ago. Thunderbolt could simple make it cleaner and easier. I'm not talking about next week, I'm talking about where thunderbolt could be realized in the future. Thunderbolt being less limited to the types of roles it can have is where the genius is. Multiple cables replaced with one is what it's all about, in this case.

But it isn't cleaner and easier. RDP/VNC is one cable (Ethernet). Previous generation's serial vt100/vt220/tn3270 links were a one cable.
There is nothing 'easier' here in terms of physical deployment.

It may be faster, more performant, in some circumstances. Or more cost effective (e.g., against 10GbE which has stubbornly stuck to higher prices even after the appearance of 40 and 100GbE. ). However, easier, in some new way, is highly questionable.

As far as FibreChannel and such, I still think TB could, in a few years, be poised to replace it as a low cost similar performing alternative.

Wrong tool for the wrong job. FibreChannel is used to build Storage Area Networks (SAN). Thunderbolt is oriented to Direct Attached Storage (DAS). A DAS solution isn't going replace a SAN one. It is largely a different tool.

Some may try to spin the notion that Thunderbolt is some sort of NAS, but it isn't. One computer is manipulating the SATA/USB/Firewire controllers spread out over the Thunderbolt 'network' is not NAS. It is largely one system with the parts distributed. It is not multiple consumers 'talking' to the storage controllers it is just one, with perhaps multiple streams, talking to many. A 1:N ration instead of a N:1 ratio. There is a very substantive difference between the two.


Additionally, Thunderbolts pragmatic 6 device limitation makes it far more a DAS solution than a NAS or SAN solution which are scalable to a completely different degree.

There is likely some hackery you can do with custom drivers layered on top of Thunderport's PCI-e layer that create a virtual Ethernet or FC layer. That coupled with attaching a local CPU to the perhiperal along with the storage controllers and similar virtual Ethernet drivers would get you NAS/SAN over Thunderbolt, but you are still stuck with the limited device network.

The coolest thing about that, in my mind, is actually going to be in the small scale, not the large scale. I know of a few individuals who are in fields where they have a large storage array connected to several workstations. Using anything from GbE to FibreChannel. In theory, a thunderbolt 'network' could deliver as good or better performance at a lower cost.

That is only if they limit themselves to old GbE and 4-8Gb FC. There are faster Ethernet (10GbE ) and FC ( 10GbFCoE ) solutions that should start to come down in price with arrival of 40 and 100 GbE which also can leverage fiber if need distance and low/lower RF interfence.

This is close to the same "theory" that Firewire 800 storage networks would displace limited 100MbE SAN networks more cheaply.

Heck, even if the rack-side equipment cost exactly the same, the ability to connect directly to a workstation (assuming Xeon support for thunderbolt and thunderbolt on future Mac Pros) without having to buy very expensive interface cards would save several hundred to well over a thousand bucks per client.

Thunderbolt doesn't obviate interface cards. It just moves their location outside the box. This is one of the core, deeply seated, misconceptions about Thunderbolt. It is just a transport for PCI-e (and Display Port). The need the functionality that formerly resided on a PCI-e card (that went into a PCI-e slot) is still there. All Thunderbolt does is move the pragmatic location of the card outside the box of the "personal computer". That's it. The interface card is still there. It is now embedded into the external peripheral.

So instead of buying the card and being able to use different peripherals that attach to that card, you now buy the card and peripheral bundled together. That is one reason the prices are higher.

Thunderbolt is a big win on devices that don't have PCI-e slots (i.e., everything except the Mac Pro).

It's an odd 'trickle down'. When we were on 10BaseT and the servers ran 100BaseT, the servers upgraded to 1000BaseT and we got 10/100 LAN, then we got gigabit and the servers moved on to 10GbE, FiberChannel, etc. Now, we sort of kind of got FiberChannel/10Gbe, only in the form of a new I/O with similar performance.

The "trickle down' just has been stalled for the last 4 years. There is now 40 and 100 GbE so 10GbE will start to trickle down now. IMHO FiberChannel is dead in term of separate cards and switches. FC over Ethernet with the same cards/switches as 10GbE is what is largely going to deploy. Thunderbolt isn't going to kill off FC as a distinct physical interconnect, that is already well underway by 10GbE. And higher bandwidths by Infiniband (which as lower latencies that both of those, but costs more.).


Frankly, with 1Gb WiFi coming wired 10GbE is going to have to get more pervasive and cheaper. The same reasons as outiined about when the when points crank up the speed the bandwidth to the edges has to go up too. Eventually wireless traffic has to transition onto wires. Thunderbolt is only an additional competitive reason that 10GbE needs to transition to much lower prices over next couple of years to survive.

But I hardly think FireWire was a failure either, as others seem to imply. It never took off in the mainstream, but why should it?

It did take off in the mainstream. For a while it was very widely used a broad array of video cameras. It is still present on non-as-low-priced-as-possible PCs. The broad based penetration though largely stopped at FW400 though.

The notion of "failure" is largely misplaced on the notion that there has to be "one port to rule them all". That's goofy. It is like proposing that there is one screwdriver ( or one hammer) to rule them all.

Thunderbolt is going to survive and have a viable market. Probably at least as widespread as FW400 eventually got to. Perhaps a bit more if the viability as standard docking station port is widely adopted. But it is not, and never, was a "USB killer". Or a 10GbE killer. Or a FiberChannel killer.

It is likely that the combination of TB and USB 3.0 will snuff off FW ports on PCs over time though.

[deconstruct60]

Originally Posted by KnightWRX

The problem is also that TB support is not written in such a way. Plugging 4-5 TB displays into a Mac Pro does not result in 4-5 user sessions, but rather 1 user session spread accross 4-5 monitors.

First, that is largely a matter of OS X not the interconnect. Using basically the same underlying hardware and interconnect protocols, Microsoft's Multipoint Server does do different sessions on different monitors (http://technet.microsoft.com/en-us/l...omponents.aspx ).


Second, you snipped out my comments about TB displays being very much a tangential "tag wag the door" issue. They are equally that same state in trying to unjustify TB and those arguments trying to justify TB.

Third, it is feasible to use the PCI-e connetion as a virtual Ethernet transport layer. In fact this is exactly what Intel is doing with their Xeon PHI PCI-e card which run a Linux image on the card and communicate to the "host" computer over virtual Ethernet connection to pass data to/from the host.

The same thing could be done with a modified Smart-and-Thin Thunderbolt display which got its own CPU just to run an enhanced VNC client. If trying to model a "thin client" distribution system then the nodes on the network ( this case Thunderbolt) would need to be "Thin" ( which means not completely lacking in a smarts (a ÇPU) .)


It is feasible to do, just not particularly likely. Or if done a widespread mode of deployment.

A lot of people are truely misunderstanding Thunderbolt and what its actual purpose is. It's not a single cable thin client protocol, it's not a networking stack to hook multiple storage arrays to multiple client machines, it's not even a networking stack to hook up multiple nodes together.

That's true. It is really merely a transport for PCI-e and Display Port traffic. That's it. Those two data streams are pragmatically hierarchical in deployments. There is a CPU (or GPU) on top of the food chain.

It's a peripheral bus local to a single host. Anything else you want to make it do requires using that peripheral bus to hook up actual peripherals to it that perform the needed function. Need networking ? Hook up Ethernet to your Thunderbolt bus... etc...

That is the normal mode but frankly those Ethernet ( or FC or whatever ) controllers are going to translate that external protocol into PCI-e traffic. As long as the perhiperal translate into PCI-e traffic (by whatever means necessary) that traffic can be fed back to the host computer.

So you could have an external device that has no physical Ethernet connector on it feed back Ethernet like data at 10GbE-like speeds if that is what is pumped onto the Thunderbolt bus. A Thin client could do that. That a dumb docking station can't (e.g., the TB Display) is beside the point.

[Garsun]

Originally Posted by saxofunk

You would not need the adapter if you're inserting another TB device into the chain.

Computer > RAID/HD/SSD > Display

Computer > Audio/Video interface > Display(s)

Computer > Breakout box > Display

I'm sure there are more options...

This is true, however I think the more likely scenario would be:

Location 1: (Back room or Cupboards on other side of office ):
MacMini and
Thunderbolt disk array.

Location 2: (Sitting on the desk or perhaps floating on a VESA articulated arm)
Apple thunderbolt display
USB peripherals

It would defeat the purpose to have a noisy disk drive as a Thunderbolt adapter.

That said, I am intrigued about your A/V Interface, what did you have in mind ?

[KnightWRX]

Originally Posted by deconstruct60

First, that is largely a matter of OS X not the interconnect. Using basically the same underlying hardware and interconnect protocols, Microsoft's Multipoint Server does do different sessions on different monitors (http://technet.microsoft.com/en-us/l...omponents.aspx ).

The point is, TB is a PCIe bus. There is no logic in there aside from that of a PCIe bus. Which makes the rest of your post basically redundant. Of course you can build PCIe type adapter cards into peripherals and make them mimick things like Ethernet or whatever (that Linux host on a card thing actually comes from a Sun idea that used to ship x86 PCs on a card for Sparc boxes to run Windows NT since the SPARC port never was officially shipped).

I did not deny this, I just said TB by itself could not do it.

Originally Posted by deconstruct60

Second, you snipped out my comments about TB displays being very much a tangential "tag wag the door" issue. They are equally that same state in trying to unjustify TB and those arguments trying to justify TB.

I snipped your rant because it had little to do with what I was responding to. People right now are having wet dreams and imagining things about TB that just are ridiculously untrue. People are seeing it as more than what it is, a simple external PCIe bus.

It's the same reason I will now snip the rest of your post. Good day.

[repoman27]

Originally Posted by KnightWRX

The point is, TB is a PCIe bus. There is no logic in there aside from that of a PCIe bus...

...People right now are having wet dreams and imagining things about TB that just are ridiculously untrue. People are seeing it as more than what it is, a simple external PCIe bus...

While I agree with most of your sentiment, at least half of the Thunderbolt equation is DisplayPort; it's not just PCIe. Adding Thunderbolt to a PC basically allows you to hang anything you would normally find attached to the PCH outside the box with no loss of performance. That's huge. It means you can externalize the vast majority of the components found in a typical consumer notebook or PC without a performance penalty. As the market shifts to UltraBooks, all-in-ones and USFF PCs, this type of technology was necessary to move forward without having to make massive trade-offs.

Thunderbolt is really not ideal for external GPUs or for that matter any PCIe application that requires a PCIe 2.0 x8 or wider connection. However, it is perfect for replicating some or all of the functionality of the southbridge outside of the box.

[AidenShaw]

Originally Posted by repoman27

While I agree with most of your sentiment, at least half of the Thunderbolt equation is DisplayPort; it's not just PCIe.

The biggest T-Bolt mistake of all....

Had T-Bolt been simply two PCIe 1.0 x4 channels, it would have had a better chance. Instead, it's a murky melange of IO and graphics.

Can you sell a T-Bolt port that doesn't implement graphics? What if you want 8 T-Bolt ports on your "future" Mac Pro - does every port have to carry DisplayPort video, or can all the bandwidth be dedicated to IO?

Using the mDP connector and bundling video was a bad, bad design choice. (Note that I say "and", it's an OK connector - it's the bundle that sucks.)

[tbrinkma]

Originally Posted by iVoid

"For those who already own a Thunderbolt-enabled Mac, Intel notes that the existing Thunderbolt ports will be compatible with both copper and fiber optic cables, ensuring cross-compatibility once the new cables arrive."

Does that mean there are optical transceivers already in the Macs with TB ports? Didn't think they did.

Or is the copper to optical hardware inside the optical TB cables?

Either way, that would be great not to have to have new hardware for it.


THIS may let me buy a 27" iMac for my home. I keep my Mac, printer, external drives, etc in my office and run a long DVI cable to a monitor in my living room. Keeps the noise level down.

With this I could keep an iMac in my living room and run the optical TB cable to my office and be in good shape. I dred to think how much a 50' optical TB cable might be though... ::shudder::

It means the 'copper to optical' hardware (and 'optical to copper' hardware as well) is inside the cable. It would have been possible to make TB that had an optical connector that the cable plugged into, but that connector wouldn't have functioned with a copper cable. Having the optical<->electrical transceivers built into the cable means that you can use *either* a copper or an optical cable, depending upon your needs. Going the other way around would have required only allowing a single cable type, or having Thunderbolt 1.0 and Thunderbolt 2.0 (now with optical!) require different, incompatible, cables.

Given that most of the expense in a copper Thunderbolt cable comes from the chips required at both ends of the cable the price differential between the optical and copper cables shouldn't be *that* huge. (Excepting typical vendor mark-up practices, of course. )

[tbrinkma]

Originally Posted by KnightWRX

And the quoted 100$ per connector in the article. Doesn't seem to have phased people, this thing is going to cost 200$ before the actual optical fiber, just the connectors.

Ouch. I do believe I missed that, and underestimated the costs of the optical transceiver in an earlier post.

Oh, well. Not like I was in the market for an optical TB cable anyway.

----------

Originally Posted by Morshu9001

But FireWire can act as an ethernet cable effectively, and I thought TB could too.

I don't doubt that it *could*, but someone would need to write drivers, and possibly create the equivalent of a TB *switch* to make it generally useful in that regard. (And then you'd run into the issue of having N computers' PCI-e buses directly connected to one another. )

Probably not worth the hassle in anything but the most unique of circumstances.

[tbrinkma]

Originally Posted by AidenShaw

USB can use hubs, which is far superior to the horrors involved with daisy-chaining.

In theory, sure. But, as always, theory and practice are only the same in theory.

In practice, hubs cause different issues than they solve. Take a look at the first step of troubleshooting *any* USB device (including hubs!). It's always, "plug the device directly into the computer, not into a hub". (Believe it or not, I've actually got a USB 2.0 4-port hub that only works properly plugged into a hub which is plugged directly into the computer. And, gods help me, a device that I rarely need to use (thankfully) which only seems to work properly plugged into *that* hub. Just to show the world of difference between theory and practice.)

Done right, daisy chaining means you only ever need a cable and a new device to add something to the chain. Not that everyone does it right, some nimrods only put a single port for daisy chain capable connector on a device so they can save a few cents in manufacturing.

On the other hand, even done right (which we've already established it isn't always), hubs run into the issue of needing more hubs to expand when you run out of ports.

Firewire gave the best of both worlds in that it could both daisy chain *and* support hubs. (Though, I think from that perspective, the daisy chaining on a Firewire device was essentially a 1-port hub.)

----------

Originally Posted by goobot

They said it themselves, it's just called that to go along with the thunderbolt name.

I was actually expecting "Dock 2", or the new "Dock" connector. Thunderbolt and Lightning do make a catchy pair, though.

[Renzatic]

If you're having so much trouble out of your USB devices, TBrink, I'd be more concerned about my computer than I would the reliability of USB in general. Seriously. I've been using USB stuff for years and years now, and the only problem I ever run into is occasionally having to replug in a device to get the OS to pick it up.

I'm wondering why you're still on insistent on extolling the wonders of Thunderbolt vs. the horrible pains of USB. We've long since established that any device you'd normally use a USB port for (printers, mice, keyboards, ect.) would be a complete waste of a link on a TB chain.

They're not comparable in the least. USB is for little devices that don't require huge amounts of bandwidth. TB is for linking external RAID clusters together and other pieces of hardware you'd normally expect to pop into a fullsized PC tower.

In other words, you're allowed to like USB again. They're not competing technologies.