It was said in the WWDC 2011 keynote. I'm tired of playing "let met Google that for you" with people around here. Look it up yourself.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Leaked Info on Third-Generation Thunderbolt Points to 40Gbps Transfer Speeds
- Thread starter MacRumors
- Start date
- Sort by reaction score
Replies
This is great news, and it's great to see the development continue. The problem with most of the replies I'm reading is that they all fall prey to the "if I don't need it, no one needs it" mode of thinking.
It's a professional connectivity type that is well needed and well used by pro level consumers for a number of devices. It's trickled down to consumer level devices (including external hard drives) and that's a great thing too. If you don't need it, great; save some dough and get the USB3 externals. I use those too.
Any advance in tech is welcome, and I cannot understand the naysayers. It works, it works great, and there's always room for improvement.
This is great news, and it's great to see the development continue. The problem with most of the replies I'm reading is that they all fall prey to the "if I don't need it, no one needs it" mode of thinking.
It's a professional connectivity type that is well needed and well used by pro level consumers for a number of devices. It's trickled down to consumer level devices (including external hard drives) and that's a great thing too. If you don't need it, great; save some dough and get the USB3 externals. I use those too.
Any advance in tech is welcome, and I cannot understand the naysayers. It works, it works great, and there's always room for improvement.
I bought into Thunderbolt when it was initially released. There was so much Promise... intel and Apple hinted at a very wide array of peripherals that would leverage the technology giving "unparalleled" expandability
that has never come.
We were showcased eGPU breakout boxes... Intel nixed anyone from developing anything along those lines.
We were showcased docking stations to turn your low end lapotp like the Macbook Air into a fairly competent workstation while docked... That has never come.
We were showcased so much promise (not the company, the concept).
What have we gotten?
A few raid boxes that cost several dozen percentage points more than their rivals using other connection technology.
We have seen one or two docking stations, that cost nearly as much as a cheap desktop computer to do the same thing.
We've seen none of this promise. The Thunderbolt standard, while extremely impressive has been nothing but an overcostly, underutilized and forgotten technology by the masses.
I give Apple a 'fistbump' for attempting to help push it to the masses with their computers, but unfortunately, between intel's draconic restrictions on developers and the overall cost of ownership of the technology, it is pretty much a non-starter
that has never come.
We were showcased eGPU breakout boxes... Intel nixed anyone from developing anything along those lines.
We were showcased docking stations to turn your low end lapotp like the Macbook Air into a fairly competent workstation while docked... That has never come.
We were showcased so much promise (not the company, the concept).
What have we gotten?
A few raid boxes that cost several dozen percentage points more than their rivals using other connection technology.
We have seen one or two docking stations, that cost nearly as much as a cheap desktop computer to do the same thing.
We've seen none of this promise. The Thunderbolt standard, while extremely impressive has been nothing but an overcostly, underutilized and forgotten technology by the masses.
I give Apple a 'fistbump' for attempting to help push it to the masses with their computers, but unfortunately, between intel's draconic restrictions on developers and the overall cost of ownership of the technology, it is pretty much a non-starter
There are at least some people who won't warm up to TB until it is 100% of PCIe speeds. It will be interesting to see if we ever reach that point.
The mac pro and every other machine has always had the same issue whenever there has been an update to the PCIe standard (or worse, a total switch from an earlier PCI or other slot format).
Are you actually doing something that is being back by having "only" TB1 or TB2 speeds?
The mac pro and every other machine has always had the same issue whenever there has been an update to the PCIe standard (or worse, a total switch from an earlier PCI or other slot format).
Are you actually doing something that is being back by having "only" TB1 or TB2 speeds?
You seem pretty sure of your "1%" figure. Thanks for your insights into what "Apple cares" about. I'm sure you know better than the rest of us.
Ivy Bridge i7 on down was 2012, Ivy Bridge xeon E5 was last fall. Same goes for Haswell, the i7 version was recent but the xeon version is still a ways off.
And you're insisting what that 99% with no evidence to back it up.
Last I checked, iPhones don't have wireless charging, correct? So every last iPhone user still uses wires, 100% of them. OK, maybe a few users are using third party adapters to get wireless charging but as you said, Apple doesn't care about the fringe 1%.
Considering wifi sync was ANNOUNCED at that event, it seems unlikely that Apple also claimed that 99% of users had stopped using wired sync at that same event. But good try weaseling out of backing up your claim with "I'm tired of looking it up..."
Really? I wonder who the other two are. Actually, wireless-everything is not appropriate in some environments...
If hardware is going to follow the trend that the new Mac Pro is setting, then Thunderbolt has to iterate frequently until it can match what an internal PCIe slot can do. It has catching up to do. Maybe at some point the iteration cycle can slow down.
I'm hoping to see things like external GPUs become the norm (even if only for pros). If Intel has been trying to stop eGPUs from happening for whatever reason, they can't prevent it from happening forever, especially if TB is fully capable of the bandwidth required.
Changing the connector might be annoying, but if there are inexpensive adapters then it wouldn't be a big deal. For example, going from FW400 to FW800 wasn't a big deal (to me at least). And now is the time to do it, when TB adoption is low.
I'm hoping to see things like external GPUs become the norm (even if only for pros). If Intel has been trying to stop eGPUs from happening for whatever reason, they can't prevent it from happening forever, especially if TB is fully capable of the bandwidth required.
Changing the connector might be annoying, but if there are inexpensive adapters then it wouldn't be a big deal. For example, going from FW400 to FW800 wasn't a big deal (to me at least). And now is the time to do it, when TB adoption is low.
I think TB 3 is going to be a game changer.
The smaller connector can be placed on iDevices to support fast data transfers and charging. Your MacBook Pro can power your external display which contains additional TB ports to connect additional devices. TB 3 can power external drives.
TB 3 is setting the standard for Thunder Bolt. I think the last two revisions were more 'beta' to get an idea of what will/won't work.
I see TB 3 more widely adopted and I will wait for TB 3 and Skylark CPU before upgrading.
----------
Exactly - it's very low right now. Having iterations to improve it to what it should be is best now than later.
The smaller connector can be placed on iDevices to support fast data transfers and charging. Your MacBook Pro can power your external display which contains additional TB ports to connect additional devices. TB 3 can power external drives.
TB 3 is setting the standard for Thunder Bolt. I think the last two revisions were more 'beta' to get an idea of what will/won't work.
I see TB 3 more widely adopted and I will wait for TB 3 and Skylark CPU before upgrading.
----------
Exactly - it's very low right now. Having iterations to improve it to what it should be is best now than later.
I know. Funny, isn't it? And, USB 2.0 was slow the day it came out. It is odd the way people point to USB 2 as some kind of paragon of virtue. Always slow, always CPU-intensive. And, it still is. USB 3 was the first USB fast enough to drive a modern disk.
Too bad they can't just come out with the TB3 connector now. Because I still don't have any TB peripherals-- I'm still using FW800, and wishing that FW1600, FW3200, and FW6400 had come out.
I hope they go to 160 Gbps so they have headroom to drive a 100 Gbps Ethernet at full speed.
Here I have to disagree. USB 2 is an "ecosystem"-- a slow, high-overhead ecosystem -- and, depending on how things are connected, it drags everything down to its level. Everything should just be USB 3, even devices that don't need that speed.
This is the root of the problem. Sorry, marketing people out there, but, this is a royal screw-up on the marketing side-- a perfect example of how marketing people can mess things up. This kind of "differentiation" makes perfect sense in a conference room, but, out in the real world, it is keeping TB from going anywhere, to no one's advantage, including Intel. If you are in marketing-- don't do this. It is a dumb strategy.
Oh, so many of you will never understand Thunderbolt.
Let me make a quick list of things it is great for, that USB (including USB 3) are not as ideal for:
Displays (yeah, it does that)
desktop RAIDs and high-performance desktop SSDs (which can actually use the extra bandwidth)
GigE and 10GigE NICs
Fibre channel HBAs
Video IO devices
Red Rocket cards and other professional PCIe devices in external Thunderbolt enclosures -- Pro Tools cards, audio interfaces, all that jazz
People who call Thunderbolt a failure, or compare it to USB -- SIMPLY DO NOT GET IT.
So, where is all that stuff, and, for the few things out there, are they cheap enough for the prosumer market?
Apple should have just stuck with a FW800/DP/USB3 combination until Intel was ready to mass-produce TB for everyone.
Sounds good in theory. Too bad TB is expensive and not mass-produced.
Apple should have known better than to listen to Intel's marketing on this one.
No special bridge likely because it is being passed through like the DisplayPort is now. Thunderbolt controller has two (dual) modes now when directly connected. One in which Thunderbolt is active and another where it is not active at all on a port that is in pass-through ( "backwards" compatible with display port) mode.
Not sure if it is poor translation but the "additional" modes could mean that with a USB 3.0 physical adapter and USB 3.0 signals being feed into the TB controller than could have a "non TB" mode where can use the port(s) for USB 3.0 ( and similar with HDMI input another HDMI 2.0 pass through mode + adapter).
That would decidedly different than encoding USB 3.0 ( or HDMI 2.0) into TB protocol to be transported somewhere else. In other words if no TB perhiperal to talk to then can still use the port for other uses ( USB 3.0 , DP 1.2 , HDMI 2.... each probably with a very low cost adapter since the port is physically different than all three. )
They could add the complexity to encode and transport USB 3.0 over daisy chains but it doesn't particularly make alot of sense or add lots of value. What primarily would be doing in that case is just adding more 'load' to the central USB 3.0 chipset versus adding another USB 3.0 controller. If really trying to add bandwidth then yet another controller (or an even faster USB 3.1 one) is needed. Otherwise, it is simply just diluting a USB controller even more. Not sure why need a 10, let alone 40, Gb/s conduit to do that.
Similar with HDMI 2.0. Perhaps a 'free' DP 1.2 to HDMI 2.0 converter so that could take transported DP 1.2 signals can convert them to HDMI 2.0 downstream... but yet another highly overlapping protocol.... why??? HDMI 2.0 just covers the ground that DP 1.2 already covers in terms of bandwidth. I can see next generation (or two) TB controllers cutting down on the external (to TB controller) DP ( and HDMI) chips need to buy to fully implement a solution ... but another long distant protocol handler... why? One reason why TB solution are expensive is the additional DP mode handling infrastructure that is required to be present. (e.g., two port TB solution tend to be substantially more expensive because can plug DP only devices into them). Integrating some of that into a more affordable chipset would lead to more affordable solutions. That is a major problem that TB has. Not lacking a HDMI connector.
There is overlap. So there is limited competition but it is much closers to say they are direct replacements. Pragmatically Firewire was/is a peripheral bus. It isn't a rigidly centralized as USB but the overwhelming vast majority of devices sold were peripherals.
They all require some CPU time. The CPU has to ask for something, disk don't have clairvoyant/mind-reader controllers that spontaneously ship data before you need it. Nursemaiding and communicating are two different things.
But wait USB 3.1 delivers 2 Amps, is reversible and doesn't have a built in supercomputer in each end of the cable and is therefore very cheap!
I think people should stop just looking at the speed of things and put more thought into usability, price, connector size, power output and things like reversibility. Thunderbolt will soon have to shrink if Apple makes thinner computers, and changing the connector shape of a cable hardly anyone uses yet is silly.
I think people should stop just looking at the speed of things and put more thought into usability, price, connector size, power output and things like reversibility. Thunderbolt will soon have to shrink if Apple makes thinner computers, and changing the connector shape of a cable hardly anyone uses yet is silly.
TB is not based on PCIe. Thunderbolt transports PCIe data. It isn't PCIe. The routing, config/distance, and admin demands are different.
So yes. It does need to be faster and is. PCIe x1 is slower than a Thunderbolt channel. Sure PCIe lanes can be bundled bigger than just one, but there is only a x4 on/off ramp to/from Thunderbolt. Thunderbolt just has to be faster than a bundle of x4 and it is. That is exactly why don't induce major latency glitches at the other end of a 5 device daisy chain.
Thunderbolt appears to parts of the rest of the system to just be a PCIe switch, but that doesn't mean that is what Thunderbolt is. That appearance at the "edge" of the Thunderbolt network is for ease of interface to mainstream design purposes, not driven by protocol implementation.
No. Thunderbolt 3 will have faster on/off ramps onto TB3 via PCIe v3.0. That doesn't mean Thunderbolt is PCIe v3.0.
Thunderbolt protocol is not "external PCIe". There already was/is an "external PCIe" standard and TB isn't it. It is different.
Why should Thunderbolt devices be mass produced when there is a limited market for people who want them and use them to there full potential?!?
I'm guessing when you say mass produced, you mean cheaper? The type of devices for Thunderbolt tend to be high performance and do more then a conventional consumer device.
Trying to get professional equipment at budget prices usually does not work in the real world without effecting quality and cutting corners somewhere.
The average consumer does not need a rack mounted enclosure that holds several Mac Minis with PCIe slots all connected by Thunderbolt for example. And certainly not cheap in any case.
Price is also irrelevant if you work in a job that has a higher rate of return based on the cost of the equipment you need to invest. Often the faster you can get your product out, the more money you can make.
If the job you work in does not have such a high rate of return, USB 3.0 should be sufficient and perhaps they don't need Thunderbolt to begin with.
Last edited:
I can't tell if you are joking or not. I'll assume not.
USB 2.0 stuck around WAY too long. Over 10 years hard drives got bigger and bigger and bigger, but USB 2.0 didn't get any faster. I remember at one point, copying my HTPC's drives to an external drive over USB 2.0 and it took over half a day.
Besides, your USB 2.0 devices work just fine in USB 3.0 ports using the same cables they came with, so what's so bad about progress?
This is just completely wrong. All of the current flagship smartphones have NAND storage that offers sequential read speeds at least 3x faster than typical USB 2.0 transfer speeds. The iPhone 4s had sequential write speeds of around 21 MB/s and sequential read of around 54 MB/s. The 5s can supposedly hit 164 MB/s when doing large sequential reads. USB 2.0 is only good for about 26 MB/s sequential write and 34 MB/s sequential read on a Mac, and can approach 40 MB/s under other OS's, but only if you use drivers which allow for higher than normal BOT packet sizes.
The problem with storage performance in mobile devices is not so much with the NAND itself, but with the optimization points it's targeting. In a smartphone or tablet, you tend to want a single package solution that is extremely power efficient. Most consumer SSDs utilize stand-alone, 8-channel controllers with external DRAMs that are designed for raw performance and thus can draw as much power as an entire iPad when under full load. They use less expensive NAND because the market is incredibly competitive, but can build reliability back in through techniques such as ECC, storing parity data, and RAIN. The majority of USB flash drives and a healthy chunk of the SD Card market are driven entirely by cost and use the cheapest, barely functional NAND that is essentially unfit for use anywhere else. SD Cards for the enthusiast / pro segment tend to focus on sequential write speeds because that impacts camera frame rates the most. Mobile devices put much more emphasis on small random read performance because it matters a lot for application and OS load times. Also, the NAND in mobile devices is actually probably quite thoroughly binned and tested, since the failure of a single die would result in the warranty replacement of a $229-$929 device.
USB 3.0 is uncommon on mobile devices at this point because it requires significant tradeoffs in terms of power consumption and die area for the SoC (and potential RF interference issues) without a ton of end-user upside. I'm sure all the handset OEMs would love to tick the SuperSpeed USB box on the spec sheet, but we're just not quite there yet.
Edit: And as for comparisons to spinning disks, the eMMC modules in the current flagship smartphones boast 4K random read performance in the range of 9 to 17 MB/s, vs. 0.55 MB/s for a 1TB WD Velociraptor.
Last edited:
There is no supercomputer inside of Thunderbolt cables. There is a relatively straightforward transciever that is tuned to the characteristics of the cable's length and materials to maximum signal bandwidth and lower noise. That doesn't meet any normal description of a supercomputer. It also isn't all that abnormal for 10+ Gb/s communications. 10+ Gb/s Ethernet, FC , Infiniband, etc all tend to have transceivers.
Facts are USB 3.1 still can't pragmatically transmit end user data at the same throughput bandwidth of TB v1; let alone v2 or v3. If slower than don't particularly need the help the transceivers provide. Similarly, time will tell if common USB 3.1 implementations don't have the same, or worse, RF interference problems that USB 3.0 had (e.g., Bluetooth RF interference problems of more than a few implementations).
Reversible is largely just thinner plug. String the pins out in a linear line and if connect to 'top' or 'bottom' of pin it doesn't particularly make a difference.
Thunderbolt doesn't have to ape exactly what USB is doing. It is largely trying to solve a different class of problems. Reversible is probably not as big of an issue as a remote GPU or storage connection coming unplugged.
The world's biggest buyer of Thunderbolt controllers is Apple. So thinner connectors is merely following where the biggest customer is going. Hardly silly.
This "thinner + additional modality" smells alot like Apple driving the design. Make Apple systems too thin for Ethernet .... hand folks a TB->Ethernet dongle. Make them too thin for USB 3.0 ... it wouldn't be surprising to hand them a USB 3.0 dongle. If TB v3.0 controller can simply do USB 3.0 signal pass through all they need put in small "clip on" adapter is the USB 3.0 PHYS chip and the power pass through. They could put no mainstream sockets at all on the system and folks would live off of either "clip on" dongles or TB docking stations/dongles.
Ditto on minimalistic HDMI and/or Display Port dongle where it is largely pin remapping and larger size/shape that the adapter is doing.
Kind of back to the original Macbook Air concept where most connectivity is oriented to being though the "air" (wireless), but a small number of ports for when have to deal with things that plug in.
I won't be too shocked if this turns out about a thin and flat as the lightning connector. It would be nice for critical connection duty if the connector was bit more secure ( snap in and higher connect/detach duty cycle).
Hey, thanks for this! If I understand what you're saying correctly, DisplayPort isn't wrapped by Thunderbolt, but rather a Thunderbolt port becomes a DisplayPort port when Thunderbolt is "disengaged", so there isn't any special hardware needed to unwrap the protocol. The adapter is, in effect, bare wires.
The implication in the slide then is that USB 3 will now be overlaid in the same way. This will please everyone who's complaining that they don't have enough USB ports and have no need for Thunderbolt. They can use a bare wire adapter to plug in.
Hopefully Apple will continue to provide a native USB port or two on their MacBooks though so I don't have to be sure to carry the adapter everywhere as a separate piece.
I think some people don't realize that TB3 is really the first True upgrade to TB. TB1 and TB2 for the most part are both 20GB. Both with 20GB speeds in both direction. TB1 did it with dual 10GB channel while TB2 uses a single 20GB channel. TB3 actually gives you more bandwidth.
From wiki:
"In June 2013, Intel announced that the next generation of Thunderbolt, based on the controller codenamed "Falcon Ridge" (running at 20 Gbit/s), is officially named "Thunderbolt 2" and slated to begin production before the end of 2013. The data-rate of 20 Gbit/s is made possible by joining the two existing 10 Gbit/s-channels. This does not change the maximum bandwidth itself but makes using it more flexible. Thunderbolt 2 was announced by Apple in June 2013 on their developer-conference WWDC to be shipped in the next generation of Mac Pro. Thunderbolt 2 is shipping in the 2013 MacBook Pro, released on October 22, 2013"
From wiki:
"In June 2013, Intel announced that the next generation of Thunderbolt, based on the controller codenamed "Falcon Ridge" (running at 20 Gbit/s), is officially named "Thunderbolt 2" and slated to begin production before the end of 2013. The data-rate of 20 Gbit/s is made possible by joining the two existing 10 Gbit/s-channels. This does not change the maximum bandwidth itself but makes using it more flexible. Thunderbolt 2 was announced by Apple in June 2013 on their developer-conference WWDC to be shipped in the next generation of Mac Pro. Thunderbolt 2 is shipping in the 2013 MacBook Pro, released on October 22, 2013"
here http://www.ispazio.net/474465/nuovi...c-2014-con-thunderbolt-ultraveloce-da-40-gbps
are saying that next 2014 Mac Book will sport it..
while here i read maybe in 2015..
i mean Broadwell ( hd6000? )MacBook Retina and TB3 ....shut up and take my money
are saying that next 2014 Mac Book will sport it..
while here i read maybe in 2015..
i mean Broadwell ( hd6000? )MacBook Retina and TB3 ....shut up and take my money
That was an extremely informative post, but I'd just like to point out that you started off by calling me wrong, yet nothing you've said corrected/contradicted anything I've said.
You simply moved to other metrics like random write or sequencial read while that's not what I was talking about. This was about sequential write and my 25 MB/s max estimation is still on point for modern smartphones:
We were discussing the topic of using a faster cable to sync your music. What's the point of fast read speed in that context? I had explicitely said:
This is how a good standard is ruined before it actually gets adopted. Keep stirring the pot.