Apple Still Considering USB 3.0 in Addition to Thunderbolt?

[TMay]

still considering? Why is it taking so long? It is clear that USB 3.0 will become standard because people are familiar with USB. And I'm sorry, but Thunderbolt will not take off, I understand the investment they made into it, but they have to realize that thunderbolt is dead, it was since day 1. Outside of tech geeks, average consumer doesn't even know what thunderbolt is, but they do know USB. Thunderbolt is firewire all over again.

Even if Apple were to be the only user of TB, it would still be worth the companies efforts to pursue it for competitive advantages within its pro market and even within its consumer market. One of the reasons that Apple is so successful is that they have numerous retail stores that are manned by people who can explain TB advantages to the average consumer.

Firewire hasn't died btw. AFAIK, it isn't even sleeping.

 

[tbrinkma]

imo, usb 3 is better than thunderbolt- I'm not a pro user, so speed like that crazy don't really matter, and usb 3 is backwards compatible... so all in all, it seems like a better option!

Don't get me wrong, there's nothing wrong with USB 3.0, but it's most certainly *not* 'better' than thunderbolt.

For one, USB 3.0 is *slower* on it's single channel, than thunderbolt is on *either* of it's two channels. Add to that the fact that Thunderbolt is essentially PCI-E over a cable, and you're left with Thunderbolt being capable of doing *everything* USB is capable of and more.

In fact, you can create a Thunderbolt peripheral which provides a computer with a USB 3.0 controller and multiple ports. There's no way to provide a USB 3.0 device which gives a computer even a single working Thunderbolt port.

 

[tbrinkma]

Actually, USB 3.0 supports 127 devices per port, just like USB 2.0.

Which means it can support 8, which TBolt cannot.

Actually, you've got to be careful when making claims like that. Technically, USB 3.0 (like USB 1.0, 1.1, and 2.0) can support up to 128 devices per controller. Most computers, regardless of how many ports they have run them off of a single controller. Not only that, but since they run multiple ports off a single controller, there's actually already a device between the controller and the port. (Specifically, a hub.) That's why you can string up to 127 devices off a single port, but that also assumes you string all of the devices (including hubs) off a single port.

In practice, however, USB (at least with 1.0, 1.1, and 2.0) stop usefully supporting additional devices after about a dozen, simply due to lack of bandwidth to share between them. (And I'm factoring low-bandwidth devices like mice, and printers into that number.)

Disclaimer: Some computers have more than one USB controller in them spread between the available ports. This is most commonly true when someone has added ports to their computer using an expansion card. Even then, you'd need 4 USB 3.0 controllers in a computer to theoretically match the available bandwidth of a single Thunderbolt port. (Since the USB protocol has so much overhead, you'd actually need somewhere in the neighborhood of 6-8 USB 3.0 controllers to match the actual throughput of a single Thunderbolt port.)

As for the question of how many devices can be 'hung' off a single Thunderbolt port, the answer is more than off a single USB port in theory (because we can hang USB ports off a Thunderbolt port), and almost as many in practice even without doing that.

I missed Thunderbolt by a single generation of MacBook Pro. I expect that by the time I'm ready to upgrade again, I'll be buying a Thunderbolt port based docking station along side my new laptop, and that said docking station will provide a number of USB 3.0, ExpressCard, and Firewire ports. Honestly, I can't figure out why they're not already available, given the obvious nature of them, and the fact that they'd require exactly zero special driver support.

 

[AidenShaw]

In fact, you can create a Thunderbolt peripheral which provides a computer with a USB 3.0 controller and multiple ports. There's no way to provide a USB 3.0 device which gives a computer even a single working Thunderbolt port.

Are you sure about that?

USB 3.0 just another serial bus - in theory the PCIe serial protocols could be encapsulated on the full-duplex USB 3.0 serial stream. Just like how the PCIe serial stream is encapsulated on the TBolt serial stream.

USB 3.0's raw data rate equals a PCIe 2.0 x1 port, and is double a PCIe 1.0 x1 port. You couldn't match the throughput of native TBolt or PCIe 2.0 x4 - but it might be possible. I admit not practical, but possible.

To paraphrase the funny old chicken sandwich commercial - "bits is bits".

----------

Actually, you've got to be careful when making claims like that. Technically, USB 3.0 (like USB 1.0, 1.1, and 2.0) can support up to 128 devices per controller. Most computers, regardless of how many ports they have run them off of a single controller.

Thank you for the correction....

 

[xgman]

Don't get me wrong, there's nothing wrong with USB 3.0, but it's most certainly *not* 'better' than thunderbolt.

well . . . except "better" selection of available and affordable peripherals.

 

[tbrinkma]

Are you sure about that?

USB 3.0 just another serial bus - in theory the PCIe serial protocols could be encapsulated on the full-duplex USB 3.0 serial stream. Just like how the PCIe serial stream is encapsulated on the TBolt serial stream.

USB 3.0's raw data rate equals a PCIe 2.0 x1 port, and is double a PCIe 1.0 x1 port. You couldn't match the throughput of native TBolt or PCIe 2.0 x4 - but it might be possible. I admit not practical, but possible.

In this case, yes, I am sure of that.

Just the raw bandwidth requirements of Thunderbolt are beyond those of a USB port. There's no way to provide a pair of full-duplex 10Gbps channels over a single port capable of (at most) 5Gbps. You would have to string together 6-8 USB 3.0 controllers to match the real-world, usable bandwidth of the Thunderbolt port, and then you'd have to deal with the overhead of mimicking PCI-E on a bus which was never designed to do so.

In the end, you could theoretically manage it with about 10-12 USB 3.0 controllers being tied together by a proprietary driver, but that's most definitely not the same thing as creating a device which gives you Thunderbolt on a USB 3.0 port. (Not only that, but it would certainly be more expensive than just adding a Thunderbolt port to the computer in the first place.)

Restricting it to a single USB 3.0 port, would probably result in a 'Thunderbolt port' with speeds comparable to USB 2.0.

And all of that ignores the DisplayPort compatibility aspect.

 

[deconstruct60]

Are you sure about that?

USB 3.0 just another serial bus - in theory the PCIe serial protocols could be encapsulated on the full-duplex USB 3.0 serial stream.
......

"bits is bits".

But if the bits are sent too fast to handle you won't be able to implement it correctly. In general yes, you can mimic the transport layer on top of another. The problem is that if speed is part of the specification then you'll miss that part.

The real point is that USB 3.0 and TB have too different objectives. USB 3.0 is to be "fast enough" for most uses and backwards compatible. TB is to be a legacy data aggregation (a couple to several older serial streams and consolidate them onto one 'wire' ).

Neither one is going to be a good replacement for the other due to the huge mismatch in objectives. There is no "one port to rule them all". Never will be.

 

[2IS]

Even if Apple were to be the only user of TB, it would still be worth the companies efforts to pursue it for competitive advantages within its pro market and even within its consumer market. One of the reasons that Apple is so successful is that they have numerous retail stores that are manned by people who can explain TB advantages to the average consumer.

Firewire hasn't died btw. AFAIK, it isn't even sleeping.

Short of an official memo from IEEE saying it's dead what would need to happen for you to concede it's dead?

 

[deconstruct60]

Umm... Have you ever bought a camera?

that should be have you bought a camera in the last 2-3 years.

Firewire (or some 1394 variant) was on lots of Video cameras at one point. That is one reason there was an uptake on inclusion into Windows PC designs through 2001-2008 (or so). For the last several years though video cameras are largely gone USB 2.0 only. There are some holdouts in to the high end range (where many folks are more conservative and relish in hooking up to 8 year old legacy equipment. ) You'll find FW there but that too will disappear over time.

It is still slightly premature for Apple to dump FW, but it will be caught in between a rock and a hard place with TB and USB 3.0. USB 3.0 will be more ubiquitous (the USB 2.0 cameras will turn into USB 3.0 cameras over next two years. So will iPods.... same reason why iPods flipped to USB from FW ... more widely adopted. ) and will attack from the "low end". TB has a good chance to be become popular enough to be much more common than FW (quasi standard docking port, AV PCI-e extenter format, etc.). All the more so when FW TB dongles get cheap enough so that folks with legacy equipment are just pointed to one of those.

USB 3.0 is "fast enough" to do isochronous work in most modern situations (with an abundance of cores). Again for high end work TB is going to work better ( just like adding PCI-e cards has over last 5-6 years. )

Again this highlights that USB 3.0 and TB are two halves to solving a wide scope of problems. Not one solution to choke off another.

Apple has dump many other legacy standards when they thought they though that standard had "dead ended". On mainstream consumer devices, FW has "dead ended" ( no mainstream system vendors are adopting anything past FW800 and many didn't move past FW400). It is only a matter of time before Apple drops the axe.

Inertia may carry FW through the 2012 Mac design cycle because TB will still be in immature stage. But once USB 3.0 is mature and TB has visible multivendor traction... Apple will likely drop the axe. [ IF TB never gets multivendor traction then that's a reprieve. ]

 

[deconstruct60]

Short of an official memo from IEEE saying it's dead what would need to happen for you to concede it's dead?

Firewire isn't dead. It just has very low momentum.

There is still going to be a large pool of legacy devices for many years to come that folks will either plug into a computer or a dongle/docking-station.

Over time though the popular A/V usages will disappear into USB 3.0 or TB. The storage usage is already obsolete as SSD become more mainstream and there are numerous alternatives with deeper traction in more widespread standards ( eSATA [SATA] , USB 3.0 [ USB] , and now TB [ PCI-e] ).


Likewise, there are going to be non consumer electronics that use the faster implementations FW3200 (e.g., control system interconnect for automotive/aerospace) . That means "someone" will be using it, so "dead" is the wrong adjective. It won't be widespread usage though.

eSATA isn't going to die overnight either. But its momemtum also will get seriously blunted by the USB 3.0 + TB combo.

 

[hstewart]

still considering? Why is it taking so long? It is clear that USB 3.0 will become standard because people are familiar with USB. And I'm sorry, but Thunderbolt will not take off, I understand the investment they made into it, but they have to realize that thunderbolt is dead, it was since day 1. Outside of tech geeks, average consumer doesn't even know what thunderbolt is, but they do know USB. Thunderbolt is firewire all over again.

You got to give it some time, it only about 6 months old and at first computers come and then devices. Thunderbolt is no way dead and Firewire still has some usage.

Thunderbolt has one thing that USB x can never do - externalize PCI-xpress.

Even if storage and PCI-E stuff never takes off - it always has display capabilities - but what is really exiting is external PCI-express stuff

 

[racketeer71]

Even if Apple were to be the only user of TB, it would still be worth the companies efforts to pursue it for competitive advantages within its pro market and even within its consumer market. One of the reasons that Apple is so successful is that they have numerous retail stores that are manned by people who can explain TB advantages to the average consumer.

Firewire hasn't died btw. AFAIK, it isn't even sleeping.

But what is this pro-market you're talking about?

And what is it, that pros need from TB, that they cannot get from existing, proven, technology?

Remember, most pros isn't first-movers. They have deadlines and stuff to do, so they want something they're pretty sure just works.

It often seems on these forums, that a lot of people have a clear idea what the "pro market" wants, though it seldom is what I see that the pros I know want (I'm no pro, so I listen more to people I know are pros, that people on random forums that says they're pro). And I'm pretty sure a real pro, doesn't need the assistance of a random dude in a retail store...

 

[AidenShaw]

except that...bits-is-bits

There's no way to provide a pair of full-duplex 10Gbps channels over a single port capable of (at most) 5Gbps.Restricting it to a single USB 3.0 port, would probably result in a 'Thunderbolt port' with speeds comparable to USB 2.0.

And all of that ignores the DisplayPort compatibility aspect.

But if the bits are sent too fast to handle you won't be able to implement it correctly. In general yes, you can mimic the transport layer on top of another. The problem is that if speed is part of the specification then you'll miss that part.

Except that by definition TBolt allows oversubscription. Six devices in a daisy chain can't all run at full speed, so there has to be a mechanism for bandwidth throttling. For example, packets may arrive at 10 Gbps, but the inter-packet gap lengthens so that each device is effectively throttled.

A USB 3.0 <=> TBolt bridge would just throttle sooner....

Again, I'm not saying that it would be practical (for example, run at full native TBolt throughput) or make any sense, but none of us are both USB and TBolt engineers who can prove that the PCIe part of TBolt (or even TBolt itself) can't be encapsulated on USB 3.0. Both are serial packet protocols.

Anyway, enough of the thought experiment - I doubt that anybody's going to build one.

 

[nylonsteel]

re original article

oh cmon aapl - give us both usb3 and thurnderbolt

 

[TMay]

Short of an official memo from IEEE saying it's dead what would need to happen for you to concede it's dead?

http://www.mrv.com/datasheets/OP/PDF300/MRV-OP-SFP1394_HI.pdf

It's moved on to different markets, the above being aviation.

 

[manu chao]

If you're referring to USB 3.0, Intel *is* shipping USB 3.0 ports on most of its Sandy Bridge motherboards. They're using the tiny NEC controller, like most USB 3.0 vendors.

What Intel didn't do was fit native USB 3.0 onto the Sandy Bridge support chipsets themselves. This was likely due to engineering scheduling concerns with adding an emerging new technology to a critical piece of the Sandy Bridge rollout - not due to "resisting advances".

You might very well be right but the press did spin this differently. There were enough headlines saying Intel delays USB 3 to bolster TB.

I think Thunderbolt is more of a crapshoot technology than USB 3.0... not because USB 3.0 is better. Just because USB has established a firm hand in the market. Look at Firewire, Firewire was by most definitions superior technology to USB, but USB became the standard and Firewire has slowly but surely been phased out in most markets.

Yes, but I think it was only USB 3 that really pushed FW into a corner.

 

[TMay]

But what is this pro-market you're talking about?

And what is it, that pros need from TB, that they cannot get from existing, proven, technology?

Remember, most pros isn't first-movers. They have deadlines and stuff to do, so they want something they're pretty sure just works.

It often seems on these forums, that a lot of people have a clear idea what the "pro market" wants, though it seldom is what I see that the pros I know want (I'm no pro, so I listen more to people I know are pros, that people on random forums that says they're pro). And I'm pretty sure a real pro, doesn't need the assistance of a random dude in a retail store...

I'll pass this on from Intel:

"For time-sensitive data, such as video and audio during creation
and playback, data transfers can be critical to the success of the
work. Thunderbolt technology was specifically designed with
video and audio applications in mind with inherently low latency
and highly accurate time synchronization capabilities.

For some power users, optimal workflows can be had with workstation
performance and expandability while using a thin and light laptop.
Thunderbolt technology enables using the thinnest and lightest laptops,
connected, with “in the box” performance over a single external cable, to
high-performance external media drives, HD displays, HD media capture
and editing systems, as well as legacy I/O hubs and devices, for the
utmost in performance, simplicity and flexibility."

http://www.intel.com/content/www/us...thunderbolt/thunderbolt-technology-brief.html

 

[gnasher729]

I think in the end, the success of Thunderbolt comes down to whether Microsoft will add Thunderbolt port support to Windows 7 the beginning of 2012 (through a driver update) and support it natively in Windows 8.

Why would you need any driver support for Thunderbolt? Thunderbolt is PCI-Express on a cable. You don't need drivers for PCI-Express. You need drivers for the things connected to it, but that would be exactly the same drivers as if a device was plugged into a PCI-Express slot.

 

[munkery]

I hope you enjoy your thunderbolt products. ( 4 of them )

Currently Thurderbolt is a disaster. Is has such a limited choice of products .

The products in the last two links make up for the deficit in the number of items .

Can you connect 8 devices to a TBolt port?

Advantage -> USB 3.0

The limit is six TBolt devices. Period. No hubs or switches exist.

Oh no!

I just connected to a USB hub via TBolt and just totally negated your argument!

What to do now?

I see 5 TBolt devices in a nearly full daisy-chain.

The other post said He wanted USB 3, which supported up to 8 devices. And you responded with this. Did you read closely

I see 3x8 (24 individual drives) connected via raid devices as well as other peripherals.

Can USB 3.0 do that?

Also,

Oh no!

I just connected to a USB hub via TBolt and just totally negated your argument!

What to do now?

You think living in the US of A is the same thing as living in China? Over investment and trade debt issues?

That has to be the most asinine thing I've ever read on this board.

I said nothing about where one lives.

It was a statement related to economics. Given the current economic climate and the projected transition of China to being the strongest global economy, the opportunities and benefits of living in the US that existed in the past may not be present in the future.

The US can't compete with other countries in terms of international sales of final product goods, such as cars and etc. This is due to many factors but primarily it is due to labor costs.

The US doesn't have a tremendous amount of raw materials to maintain its economy via selling raw materials to the countries that are more successful in the realm of international trade.

The US economy is imploding and it's taking the global economy with it because the US was an investment anchor within the global financial system. Most other countries were buying US bonds as a major mean of investment due to the continuing growth in the US.

Once the US is no longer the strongest global economy, those countries will invest in the countries that are the global leaders. There will be a reorganization of the global financial system and the US will no longer be as much of a major player in the game anymore.

Subsequently, many of those aforementioned opportunities and benefits will begin to be reduced.

 

[AidenShaw]

You might very well be right but the press did spin this differently. There were enough headlines saying Intel delays USB 3 to bolster TB.

Many of the Apple blogs and bloggers certainly did spin it that way....

Of course, the fact that Intel roadmaps from long before Light Peak was metalized said the same story wasn't mentioned.

 

[TMay]

Except that by definition TBolt allows oversubscription. Six devices in a daisy chain can't all run at full speed, so there has to be a mechanism for bandwidth throttling. For example, packets may arrive at 10 Gbps, but the inter-packet gap lengthens so that each device is effectively throttled.

A USB 3.0 <=> TBolt bridge would just throttle sooner....

Again, I'm not saying that it would be practical (for example, run at full native TBolt throughput) or make any sense, but none of us are both USB and TBolt engineers who can prove that the PCIe part of TBolt (or even TBolt itself) can't be encapsulated on USB 3.0. Both are serial packet protocols.

Anyway, enough of the thought experiment - I doubt that anybody's going to build one.

It's a nice thought experiment. The reality is that TB has a roadmap for more channels and fiber for distance, albeit lacking a spec for power over fiber, and this gives it at least a short term edge over USB org, who are probably rolling the USB 4.0 spec with more channels to match TB as we speak.

Truth be told, I'll bet Apple is quite happy to have any standard that supports multi-protocols and gives them the ability to strip out the bulk and weight of performance hardware from the MBA/MBP and make it available in a pseudo dock consisting of a Magsafe power connector , TB connector and a TB display.

So far, and inspite of a current lack of any variety of TB peripheral, I'd say that Apple should be pleased with the results.
As an aside, I have some ancient GPIB/HPIB gear from HP that I would love to see get the TB interface treatment.

 

[AidenShaw]

Oh no!

I just connected to a USB hub via TBolt and just totally negated your argument!

Yes, a USB hub, not a TBolt hub.

No more than six TBolt devices per port. Period.

If your TBolt device is a USB controller, you could have 127 USB devices. If your TBolt device is a SATA controller, you can have as many SATA devices as it supports.

But no more than six TBolt devices.

 

[munkery]

Yes, a USB hub, not a TBolt hub.

No more than six TBolt devices per port. Period.

If your TBolt device is a USB controller, you could have 127 USB devices. If your TBolt device is a SATA controller, you can have as many SATA devices as it supports.

But no more than six TBolt devices.

Given that TBolt can support USB such that it can support as many devices as USB plus other TBolt devices, I don't think your argument makes any sense.

----------

BTW, this was your original argument:

Can you connect 8 devices to a TBolt port?

It does not define any limit to only TBolt devices.

Also, here is you negating your own argument:

If your TBolt device is a USB controller, you could have 127 USB devices. If your TBolt device is a SATA controller, you can have as many SATA devices as it supports.

 

[deconstruct60]

Truth be told, I'll bet Apple is quite happy to have any standard that supports multi-protocols and gives them the ability to strip out the bulk and weight of performance hardware from the MBA/MBP and make it available in a pseudo dock consisting of a Magsafe power connector , TB connector and a TB display.

A theory that has already failed to get large traction in the market place. The remodeled and modernized MBA now has two USB , not one and certainly not zero. Apple tried to get squeeze out USB and failed. I doubt they are going to repeat that mistake any time soon.

Given Apple doesn't put modems in the MBA and the fact that folks have devices like mice/USB drives/keys/card readers/etc to plug in means you need USB on the box.

This whole notion of TB only sockets computers is fantasy land.

 

[deconstruct60]

Except that by definition TBolt allows oversubscription. Six devices in a daisy chain can't all run at full speed, so there has to be a mechanism for bandwidth throttling.

Unless the standard is crap it is the aggregate bandwidth that is required to meet the standard. You can't just set up a node where it just lies about some isochronous reservation of bandwidth just to choke all the other nodes down to lower rates that it can handle.

PCI-e doesn't work well where the latencies get too far out of whack. (otherwise there would be few distance limitations on the spec. You'd just stick repeaters on the lines to keep extending the length). There has to be a cap on the how much you and overload the TB network.

I don't any hack like this would pass certification tests. It is a hack you make kind-of sort-of work in a lab, but it is just as hack.