now someone needs to sell a monitor that has a PCI x16 slot for a video card built into it, and we can start putting video cards into our monitors. Then it would accelerate whatever you plugged into it - laptop or desktop. And my desktop computers could be uber tiny mitx builds since the video card is the largest component.
When that 400W PSU is integrated into the monitor, it wouldn't necessarily be a problem. If you're putting a PCIe slot into the display, the concept of thin has been thrown out.
If you still want thin, there is the MXM card option. Not the highest performing solution but it'd simplify video signal routing. A true PCIe card would need a short cable to go from the card into the monitor.
Actually i'd be nicer to have a 500W unit so that you can provide 300W to the GPU, 100W for laptop charging and another 100W for the display + integrated USB Type A ports.
No good. I don't want to buy a monitor, i just want to play new demanding games on my laptop. A new monitor costs me at least 150€ (and that's a 21'' at 1080 at 60Hz with 2W speakers, so not much). We just need a case that has room enough for a standard PSU, GPU and maybe a 3.5'' hard drive. The IC board should come with the case and there should be a few extra USB ports for a mouse, controller and to charge your laptop or tablet (if you are using an external display and it has an x86 CPU that might work). When you get home it's all there and you just plug in the laptop and GAME ON!!
I'm hoping that the iPad Pros (or some other tablet any company would build) connect via Thunderbolt and you can use them as external monitors with their own graphics processors.
Almost seems too good to be true. Now if we can work on a new reversible large version. Like the size of USB Type-A, but look/feel of USB Type-C. That would be better for desktops I'd think.
The whole point of type-c was to get rid of the confusing array of many USB connectors that exists currently. There's no reason to have another version.
No, there should be only one connector. However, I would support specifying the dimensions of the plastic block behind the connector, so that desktops could have the port sunk in and use the plastic block to help provide rigidity to the connection. That would provide the same advantage that type A gets from being bigger, while still allowing the same cable to be connected to a phone that doesn't have room for a large connector.
I think type A's "advantage" is that it's old and badly designed. It was a big deal at the time but compared to other modern IO standards the "USB superposition principle" shows that it could be done better. (and in the case of FireWire-400, was done better :/)
So, to summarize: Thunderbolt 3 offers all the benefits the USB C/3.1 port would have given you anyways, but if you want more bandwidth or a longer cable, you can buy the same expensive active cables that made it niche in the first place.
With that said, the ability to run an external GPU is pretty cool. I'm glad they aren't against it now.
Yes, 10GBit Ethernet and external GPUs can be really cool features. On mainboards even 4 GPUs is stressing it. Air cooling becomes almost impossible or very loud. but if you could attach a bunch of GPUs in well-designed seperate chassis, that could open up a new world for number crunchers. The more GPUs per host the more efficient these system are (usually).
With 100 W delivered over the cable one might not even need seperate PSUs in those enclosures. But plenty of PCIe power connectors on the mainboard, right near the I/O panel.
Is it possible to due power pass through? For example, the ultrabook is being feed by a 100W Type-C power adapter but only needs ~20W. Can this excess power be routed to another Type-C port on the laptop to provide ~80W to an external device? With this setup an second AC adapter for the external device is avoided.
That should be possible already, all it would be determined by would be the power delivery circuitry. The CPU and other components would draw as much power as needed and any excess power available to draw would be given to other parts, although it might be something only possible while plugged in and might need special drivers for it.
It's more than just external GPUs. It supports the PCI Express protocol in general, which means it can run any type of desktop card. Firewire, Pro Audio, Pro Video, etc.
I imagine it was time constraints. If Intel is anything like NV then they are working on processors a few years before their expected release. There wasnt time to integrate thunderbolt with skylake. I bet we'll see it integrated passed cannonlake or whatever after skylake is called.
"Intel’s greatest argument in favor of the technology is docking, as the use of PCI-Express and now the addition of USB Power Delivery gives Thunderbolt a degree of flexibility and performance that USB Type-C alone doesn’t match."
How? A pure USB 3.1 type C dock can deliver 100W power for charging, additional usb ports for peripherals, and displayport video signal for the display. The only possible addition for thunderbolt would be for workstation docks that add large number of high throughput storage with gig ethernet and graphics which would overwhelm USB3.1 but that should be a minority of dockers. Most would probably be like the current thunderbolt display users who have networking, a mouse and keyboard connected to the display to the computer through one cable. USB is plenty for that scenario.
10 Gbit Ethernet is emulated but it'd be nice to have a 40 Gbit option for direct device-to-device transfers. On the Mac side, this is great for quick and dirty data transfer. Newer Macs have SSDs fast enough to saturate a 10 Gbit link.
WIth pure USB, you are limited to only two DisplayPort lanes and USB attached ethernet adapters which tend to underperform compared to their PCIe counterparts. Like you say though, it's probably not an issue for most users.
You are of course bandwidth-limited to 40 Gbps in each direction (and a maximum of 32 for PCIe alone). Of course if you try to attach too many high-bandwidth devices it won't work well. But who would want to run a 5k monitor off of internal graphics when they have an external GPU attached?
Indeed. The interesting thing here is that with an external GPU, Thunderbolt would be used for the initial data transfer and sending commands to the GPU, not raw pixels. With accelerated compositing, information in a window would need to be sent once to be cached in the GPU's memory. With video decode hardware being used, the compressed video stream is what would be passed over Thunderbolt, not the high bandwidth decompressed version that gets sent to the displays. Thus the bandwidth from the external GPU to the displays can actually exceed 40 Gbit and not be bottlenecked.
I'd definitely recommend getting a GPU with as much VRAM as you can afford though. This should limit the amount of data that needs to be sent over the bus, and maximize performance.
Something like Dell's 5K monitor would work with TB3. You'd just have an Alpine Ridge controller in the monitor splitting off TB3 into 2 DP 1.2 signals, instead of 2 incoming DP 1.2 cables as you have today.
How is the video data carried via Thunderbolt? Isn't it DP1.2 protocol? Because apparently Thunderbolt3 at 40Gbps doesn't transmit DisplayPort - it only transmits in 20Gbps mode. And, for 5k, you'd need 2x the bandwidth of DP1.2, which isn't enough at 20Gbps.
The only option I see is if there's a new PCI-Express protocol for frame buffer transmission over Thunderbolt, running at 40Gbps. A direct write to the monitor's frame buffer via PCI Express, instead of a video signal. This completely ignores the DisplayPort protocol.
I am TREMENDOUSLY EXCITED about the external GPU enclosures finally becoming available.
Finally, my dream of a superlight/portable ultrabook hooked up to a high-end enthusiast GPU to play games is on the horizon!
Honestly, I lay this at the feet of Alienware and MSI. These two pioneers got tired of waiting for Intel and just did it on their own, showing Intel that the market wanted eGPUs. Thanks!
External GPUs is something that should have been supported since day one with the first Thunderbolt release. On the OS X side of things, external GPU's can be enabled via driver hacks but they're not hot swapable. Hopefully by the time Thunderbolt 3 is shipping in some devices, the software side will have been worked out for external GPUs.
I'm using an external GTX 970 over Thunderbolt 2 as I type this. There are no software issues. This is external GPU year 3 for me as I upgraded my Lenovo S430 (Thunderbolt 1) to a Windows Macbook Pro 15 (Thunderbolt 2). The Magma eGPU enclosure was also upgraded from Thunderbolt 1 --> 2
Drivers come straight from nVidia and I've never had a game crash or anything. External GPUs are already here and working fine, you just haven't bought one.
Having said that, you're smart for waiting as Thunderbolt 3 looks to be perfect :)
TL;DR -- By us switching to USB-C ThunderBolt will inherit a whole bunch of advantages over our previous version of the product that virtually no-one used (except as a port to plug in their DisplayPort monitor).
So we now know that Thunderbolt is well and truly dead.
External graphics have been talked about every Thunderbolt generations. When are affordable, professionally done external TB GPU boxes coming? That's all I want. It shouldn't cost so much for just a small case, connectors, and a power supply.
By endusers, yes. This is the first time Intel sanctioned eGPUs. I too want to see pricepoints and specs.
I want to buy a thunderbolt enclosure with a pci-e 16x slot, a 350w PSU with at least 1 8-pin and 2 6-pin GPU plugs, 2 or more USB3.1 ports, a gigabit ethernet port, and both 1/8" and optical audio jacks for $300 or less. This enclosure should also charge my laptop over the same USB type-C cable.
At that price, with that feature set, they would sell millions of 'em.
I don't know the actual parts costs, but currently every thunderbolt port (or sometimes a pair of thunderbolt ports) on a peripheral adds about $100 to the cost.
Perhaps they can get this cost down with more volume, but it takes more than just connectors and a power supply.
There were rumors that Intel was refusing to certify external graphics solutions on previous version of TB. A few companies showed prototypes but they never made it to market. Now they're officially endorsing it, so there's a very good chance it actually happens.
I can see it now. Modded Mini ITX cases on both sides of a row of monitors. One case houses the main components (Minus gpu) and the other case has GPU, soundcard, and subwoofer with speakers on the outer edges of the setup. Symmetry
I'm a little confused about the USB 3.1 thing. I can see how having it built into the Alpine Ridge controller would allow the port to function as a full USB 3.1 port without requiring support from the chipset but I don't see how it allows you to have a USB 3.1 port on a monitor etc without still having extra hardware there. Thunderbolt 3 is using the alt mode of the USB C connector so it is taking over the pins that would be used for USB 3.1. There are still pins to natively provide usb 2 so you could have a USB 2 hub in the monitor no problem. You could also add a PCI-E usb3 controller in the monitor and provide USB3 ports over the thunderbolt connection. I don't see how Alpine Ridge avoids that?
NM I reread it and understand it now. They are saying that when you use Alpine Ridge in a monitor it gives you the TB connection back to the host and gives you a USB 3.1 port on the monitor.
in thunderbolt it hijacks ALL pins of the connector and runs a two stream thunderbolt protocol over it, signal wise this has *nothing* to do with usb3, it just uses the same copper wires and port design.
as far as i see it any thunderbolt chip (be it host or slave (the monitor)) has usb3.1 baked in, so: - if you plug a usb device into a pc with a thunderbolt port, it is attached to that chip - if you plug a usb device into a thunderbolt monitor, that device shows up under the thunderbolt chip in the monitor, whereas the one in the pc is transparent to the system.
Basically the monitor would have to have a thunderbolt chip in it that splits the usb 3.1 off from the thunderbolt signal and then pipes it out to ports on the monitor. Hopefully that can be done cheaply.
OK, so TB3 does 40 Gbps with an active cable (half that with a cheap passive cable), which is about equivalent to PCI-Express 1.0 on 16 lanes (32 Gbps). If we hang our graphics card off of that, what's the real-life framerate impact in games, compared to a semi- current PCIe 4.0 @16 lanes, which reaches 252 Gbps instead of TB3's 40 Gbps. I'm guessing the impact won't be that dramatic because most textures will be in the graphics card's RAM and 40Gbps for graphics command is fine, but still, textures need to be swapped in/out of GFX RAM and some time, maybe regularly. A PCIe 1.0 vs PCIe 4.0 test would probably be a good proxy (16 lanes for active, 8 for passive), pending actual TB3 hardware and assuming TB3 efficiency is abotu as good as PCIe ? Could you guys do that so we know where we stand ?
Well, first off, PCIe 4.0 isnt out yet, so you must be referring to 3.0. Since 2.0 is twice as fast as 1, and 3.0 is (almost) twice as fast again, a PCIe 3.0 x4 slot is about the same speed as a 1.0 16x slot. There have been tests showing a PCIe 3.0 x4 slot does limit GPU perf a bit but not too much. Anandtech did an article with a GTX 780 I think on PCIe 3.0.
Quick question - does having all these different protocols on top of USB 3/3.1 mean that we will need different versions of the ports? They may look physically similar - but are different internally to support the particular protocol - or any generic USB 3/3.1 port will support these?
Can I dream of a near future where all my devices use a single type of port for all I/O - Display/Keyboard/Printers/Charging/External Storage/External GPU? Is there is far away - I know theoretically you can use USB 3 to do that today?
Really excited about external GPU's - I will take this a step further - I want to use my Phone (& not Laptop) as my primary computing device - I would like to dock my phone into a larger screen & make it a tablet or a even larger screen with external GPU for Desktop gaming - Windows seems to be working towards that future.
Not every port will support everything. Beyond USB, everything else is optional. Stuff like DisplayPort only requires a few internal wires and a multiplexer, so it should be fairly common, but Thunderbolt, for instance, requires a dedicated controller.
Different ports will support different protocols, depending on the controller that's attached to the port.
All of the ports will support USB 3.1. Some ports will support DisplayPort, or Thunderbolt, or Power Delivery, or whatever else comes around via USB Alternate Mode.
And there will be little pictures above/beside each port that describe which protocols the port supports.
It's 3 steps forward, 2 steps back. :( We've eliminated the USB port chaos ... only to introduce which-protocol-is-supported-on-which-port chaos. :(
Thanks everyone - Agree its gonna be chaotic if ports don't support all protocols - its going to another round of fidgeting under the desk to find the right port - something I had hoped we eliminated
Is there anything that would prevent Apple from designing a USB-C to DP 1.3 adapter that runs TB3 in the 40 Gbps mode? The very short cable along along with active equalization in the adapter should allow the 40 Gbps mode, right?
The dearth of GPUs, scalers and Tcons that support DP 1.3 would make that a rather pointless exercise at the moment, but theoretically it could be done at some point in the future. I'm a little surprised that nobody came out with an OG Thunderbolt to DisplayPort 1.2 adapter, but I guess it would have been bloody expensive because it would have required a power supply, active cable, two 4C Thunderbolt controllers, and a DP concentrator like this one from ST: http://www.st.com/web/catalog/mmc/FM128/SS3/PF2552...
It would appear that Alpine Ridge controllers are finally going to support two DP source connections, so at least a Thunderbolt 3 to DP 1.3 adapter would only require a single 4C Thunderbolt 3 controller.
I'm rather hoping some of these features, or royalty-free equivalents, will be rolled up into a USB 3.2 spec. Not sure if they can increase the bandwidth without using active compoents though.
Interesting but not exciting, if Intel cared anything about Thunderbolt they would include a controller at least on their top level chipsets, server chipsets or even the Z series ones to get it in to the mainstream PC world and give it some traction.
USB 1 was in computers for quite a while before it got OS and peripheral support, whilst getting TB in to Apple products is a start it hardly makes it ubiquitous which is what is needed. It seems internal nonsense at Intel may be getting in the way of success for TB.
Windows 10 Continuum is an example of where TB and USB-C could make a case, a base station for a mobile phone that can drive a couple of 4k monitors and handle all the wired networking and peripherals would be a big potential sell in enterprise as a laptop replacement.
I think you may have a different definition of "success" for Thunderbolt than Intel does.
The die for the Z97 chipset is only about 52 mm^2, has a TDP of 4.1 W, and is priced around $48. The package is 22 mm x 23 mm though, because it has a ball count of 708 to support all of the various features. The DSL5520 Thunderbolt 2 controller has a die area of around 120 mm^2, a TDP of 2.8 W, and, at least according to ARK, is priced at $9.95. It also needs to be located very close to the physical ports, which is not necessarily where you want to put your PCH.
If the two were even being manufactured on the same process, integrating Thunderbolt 2 would increase the die area of the PCH by 230%, yet Intel would be hard pressed to raise the price even 20%. Sure the install base would swell, but it would be a terrible idea from a financial standpoint. Furthermore, they would still need to produce the discrete controllers for devices anyway, so unless the attach rate rises by 20 million extra devices due to the larger install base, volume will actually decrease for the discrete controllers. This could actually lead to upward pricing pressure for devices.
The bottom line is that Thunderbolt isn't meant to be ubiquitous or cheap. The use cases really are specialized, and the majority of PC users don't require it. Discrete GPUs probably represent the number one use of PCIe expansion slots by consumers, and fortunately it looks like Intel is finally allowing that to happen via Thunderbolt 3. Most folks can live without > 500 MB/s external storage solutions, single cable 5K monitor support, or 10 GbE / InfiniBand / Fibre Channel adapters for their Ultrabooks. Apple needs Thunderbolt because they no longer offer any Macs with user accessible PCIe slots. Every other OEM is still producing desktop PCs with standard PCIe expansion slots to serve customers that want to add specialized hardware to their machines, and piping PCIe over an insanely fast external serial link is never going to be as cheap as simply putting a card edge connector on a motherboard.
I've been looking for a site that would explain the USB-C, Thunderbolt, and USB-C connections. I am not computer knowledgeable, and my interest in all this will be regarded as rather provincial by the folks here: I want a port that will convey data to an external hard drive as fast as that data can be written to an internal hard drive. With such a set up, one could boot from an external drive as quickly as one can from the computer's internal drive. Does that require Thunderbolt 3? Will USB 3.1 do that?
So has this already made the USB-C implementation of the 12" MacBook legacy? Is that why 6 months after release, there still isn't a single-cable USB-C docking station worth purchasing? Is everyone waiting to go with Thunderbolt 3 as the new standard, like the upcoming Dell XPS implementations?
I guess being stuck at 5Gbps, plus the limitations of Broadwell Core M only being able to output 2560x1600 @ 60hz make the 12" MacBook, Chromebook Pixel 2, and all current forms of the Apple USB-C adapters stop-gap technology that will eventually be bargain bin trash?
Should we expect the 2016 successor of the 12" MacBook as well as the Pro line to feature Thunderbolt 3 and a whole new line of USB-C/TB3 compatible adapters/docking stations? Or will there be a market split for high end devices using Thunderbolt 3 and entry level devices using 10Gbps USB-C Gen 2?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
91 Comments
Back to Article
jwhannell - Tuesday, June 2, 2015 - link
now someone needs to sell a monitor that has a PCI x16 slot for a video card built into it, and we can start putting video cards into our monitors. Then it would accelerate whatever you plugged into it - laptop or desktop. And my desktop computers could be uber tiny mitx builds since the video card is the largest component.littlebitstrouds - Tuesday, June 2, 2015 - link
Eww no. My monitor purchases have outlasted every video card purchase by at least 2 fold.Raniz - Tuesday, June 2, 2015 - link
No problem as long as it's just a PCI x16 slot and not an integrated graphics card.wujj123456 - Wednesday, June 3, 2015 - link
You mean you want to have a 400W monitor power adapter?Kevin G - Wednesday, June 3, 2015 - link
When that 400W PSU is integrated into the monitor, it wouldn't necessarily be a problem. If you're putting a PCIe slot into the display, the concept of thin has been thrown out.If you still want thin, there is the MXM card option. Not the highest performing solution but it'd simplify video signal routing. A true PCIe card would need a short cable to go from the card into the monitor.
Actually i'd be nicer to have a 500W unit so that you can provide 300W to the GPU, 100W for laptop charging and another 100W for the display + integrated USB Type A ports.
SleepyFE - Wednesday, June 3, 2015 - link
No good. I don't want to buy a monitor, i just want to play new demanding games on my laptop. A new monitor costs me at least 150€ (and that's a 21'' at 1080 at 60Hz with 2W speakers, so not much). We just need a case that has room enough for a standard PSU, GPU and maybe a 3.5'' hard drive. The IC board should come with the case and there should be a few extra USB ports for a mouse, controller and to charge your laptop or tablet (if you are using an external display and it has an x86 CPU that might work). When you get home it's all there and you just plug in the laptop and GAME ON!!renstein - Thursday, June 4, 2015 - link
I'm hoping that the iPad Pros (or some other tablet any company would build) connect via Thunderbolt and you can use them as external monitors with their own graphics processors.wicketr - Tuesday, June 2, 2015 - link
Almost seems too good to be true. Now if we can work on a new reversible large version. Like the size of USB Type-A, but look/feel of USB Type-C. That would be better for desktops I'd think.SirKnobsworth - Tuesday, June 2, 2015 - link
The whole point of type-c was to get rid of the confusing array of many USB connectors that exists currently. There's no reason to have another version.sonicmerlin - Tuesday, June 2, 2015 - link
Are you kidding? You want *another* connector? You want to ruin the utter simplicity and beauty of having one connector that does everything?Are you the guy who made mini and micro USB a thing?
WorldWithoutMadness - Tuesday, June 2, 2015 - link
Maybe he has Type-A fetishes :pmkozakewich - Wednesday, June 3, 2015 - link
Oh man, think about it! We can go back to those big, beautiful, square Type B connectors!Telemachus13 - Tuesday, June 2, 2015 - link
No, there should be only one connector. However, I would support specifying the dimensions of the plastic block behind the connector, so that desktops could have the port sunk in and use the plastic block to help provide rigidity to the connection. That would provide the same advantage that type A gets from being bigger, while still allowing the same cable to be connected to a phone that doesn't have room for a large connector.BillyONeal - Tuesday, June 2, 2015 - link
I think type A's "advantage" is that it's old and badly designed. It was a big deal at the time but compared to other modern IO standards the "USB superposition principle" shows that it could be done better. (and in the case of FireWire-400, was done better :/)Timbrelaine - Tuesday, June 2, 2015 - link
So, to summarize: Thunderbolt 3 offers all the benefits the USB C/3.1 port would have given you anyways, but if you want more bandwidth or a longer cable, you can buy the same expensive active cables that made it niche in the first place.With that said, the ability to run an external GPU is pretty cool. I'm glad they aren't against it now.
MrSpadge - Tuesday, June 2, 2015 - link
Yes, 10GBit Ethernet and external GPUs can be really cool features. On mainboards even 4 GPUs is stressing it. Air cooling becomes almost impossible or very loud. but if you could attach a bunch of GPUs in well-designed seperate chassis, that could open up a new world for number crunchers. The more GPUs per host the more efficient these system are (usually).With 100 W delivered over the cable one might not even need seperate PSUs in those enclosures. But plenty of PCIe power connectors on the mainboard, right near the I/O panel.
sonicmerlin - Tuesday, June 2, 2015 - link
Imagine a sub 100W HBM mobile GPU running entirely off an ultrabook's USB port.Ryan Smith - Tuesday, June 2, 2015 - link
Don't expect to see an Ultrabook capable of delivering 100W, or even 50W for that matter.Kevin G - Wednesday, June 3, 2015 - link
Is it possible to due power pass through? For example, the ultrabook is being feed by a 100W Type-C power adapter but only needs ~20W. Can this excess power be routed to another Type-C port on the laptop to provide ~80W to an external device? With this setup an second AC adapter for the external device is avoided.jase240 - Sunday, June 7, 2015 - link
That should be possible already, all it would be determined by would be the power delivery circuitry. The CPU and other components would draw as much power as needed and any excess power available to draw would be given to other parts, although it might be something only possible while plugged in and might need special drivers for it.barleyguy - Tuesday, June 2, 2015 - link
It's more than just external GPUs. It supports the PCI Express protocol in general, which means it can run any type of desktop card. Firewire, Pro Audio, Pro Video, etc.BIG advantage over USB 3.1 IMO.
Mikemk - Tuesday, June 2, 2015 - link
Even the passive cable offers double bandwidth over USB 3.1Kevin G - Wednesday, June 3, 2015 - link
Not surprising. USB Type-C was designed with some head room for faster speeds. Thunderbolt is just using it before native USB does.jimmy$mitty - Tuesday, June 2, 2015 - link
Intel was never against it, there just was not a huge market since Thunderbolt was stuck to a niche port design and thus much more expensive.invinciblegod - Tuesday, June 2, 2015 - link
Why doesn't intel build in support for thunderbolt into their chipsets?Flunk - Tuesday, June 2, 2015 - link
Intel doesn't believe that Intel's Thunderbolt is worth supporting.jeffkibuule - Tuesday, June 2, 2015 - link
Cost.MrSpadge - Tuesday, June 2, 2015 - link
The controller is too large and expensive to add it everywhere, I suppose.Eidigean - Tuesday, June 2, 2015 - link
Because the Thunderbolt controller needs to be within a few centimeters of the connector. The south-bridge chipset is too far away from the backplane.barleyguy - Wednesday, June 3, 2015 - link
The solution to that is just to run a USB 3.1 cable from the chipset to the back of the case.beta_Cyth - Thursday, June 4, 2015 - link
I imagine it was time constraints. If Intel is anything like NV then they are working on processors a few years before their expected release. There wasnt time to integrate thunderbolt with skylake. I bet we'll see it integrated passed cannonlake or whatever after skylake is called.wicketr - Tuesday, June 2, 2015 - link
Also, how close are we to supporting good external graphics cards? What would the bandwidth need to be for something like a GTX 660?SirKnobsworth - Tuesday, June 2, 2015 - link
http://www.anandtech.com/show/7987/running-an-nvid...Apparently the x4 connection isn't a huge bottleneck, at least as far as gaming is concerned.
invinciblegod - Tuesday, June 2, 2015 - link
"Intel’s greatest argument in favor of the technology is docking, as the use of PCI-Express and now the addition of USB Power Delivery gives Thunderbolt a degree of flexibility and performance that USB Type-C alone doesn’t match."How? A pure USB 3.1 type C dock can deliver 100W power for charging, additional usb ports for peripherals, and displayport video signal for the display. The only possible addition for thunderbolt would be for workstation docks that add large number of high throughput storage with gig ethernet and graphics which would overwhelm USB3.1 but that should be a minority of dockers. Most would probably be like the current thunderbolt display users who have networking, a mouse and keyboard connected to the display to the computer through one cable. USB is plenty for that scenario.
nevcairiel - Tuesday, June 2, 2015 - link
I think the biggest selling point is PCI Express support through Thunderbolt.Senti - Tuesday, June 2, 2015 - link
Just 10Gb would be sweet – you won't be getting it in notebooks remotely soon any other way.Senti - Tuesday, June 2, 2015 - link
10Gb Ethernet*Kevin G - Wednesday, June 3, 2015 - link
10 Gbit Ethernet is emulated but it'd be nice to have a 40 Gbit option for direct device-to-device transfers. On the Mac side, this is great for quick and dirty data transfer. Newer Macs have SSDs fast enough to saturate a 10 Gbit link.SirKnobsworth - Tuesday, June 2, 2015 - link
WIth pure USB, you are limited to only two DisplayPort lanes and USB attached ethernet adapters which tend to underperform compared to their PCIe counterparts. Like you say though, it's probably not an issue for most users.DCide - Tuesday, June 2, 2015 - link
So a pure USB 3.1 type C dock can support a 5K monitor and daisy chain 5 more devices with 40Gb/s throughput? Does it support external GPUs as well?I may have missed something, but I'm sure it can't do ALL these things.
SirKnobsworth - Tuesday, June 2, 2015 - link
You are of course bandwidth-limited to 40 Gbps in each direction (and a maximum of 32 for PCIe alone). Of course if you try to attach too many high-bandwidth devices it won't work well. But who would want to run a 5k monitor off of internal graphics when they have an external GPU attached?Kevin G - Wednesday, June 3, 2015 - link
Indeed. The interesting thing here is that with an external GPU, Thunderbolt would be used for the initial data transfer and sending commands to the GPU, not raw pixels. With accelerated compositing, information in a window would need to be sent once to be cached in the GPU's memory. With video decode hardware being used, the compressed video stream is what would be passed over Thunderbolt, not the high bandwidth decompressed version that gets sent to the displays. Thus the bandwidth from the external GPU to the displays can actually exceed 40 Gbit and not be bottlenecked.Res1233 - Friday, September 4, 2015 - link
I'd definitely recommend getting a GPU with as much VRAM as you can afford though. This should limit the amount of data that needs to be sent over the bus, and maximize performance.Murloc - Tuesday, June 2, 2015 - link
99% of people can't tell the difference between physical layer and protocol.vFunct - Tuesday, June 2, 2015 - link
Lack of native support for 5k monitors is a damned shame. That's the standard new resolution now for the retina 27" displays.Or do are they supported via dual DIsplayPort1.2 channels in a single Thunderbolt 3 connector?
jeffkibuule - Tuesday, June 2, 2015 - link
The latter.BillyONeal - Tuesday, June 2, 2015 - link
Considering it isn't used outside of one manufacturer's self-contained PC I wouldn't exactly call it a "standard" at this point.vFunct - Tuesday, June 2, 2015 - link
You don't know about Dell's 5k display?Ryan Smith - Tuesday, June 2, 2015 - link
Something like Dell's 5K monitor would work with TB3. You'd just have an Alpine Ridge controller in the monitor splitting off TB3 into 2 DP 1.2 signals, instead of 2 incoming DP 1.2 cables as you have today.vFunct - Tuesday, June 2, 2015 - link
How is the video data carried via Thunderbolt? Isn't it DP1.2 protocol? Because apparently Thunderbolt3 at 40Gbps doesn't transmit DisplayPort - it only transmits in 20Gbps mode. And, for 5k, you'd need 2x the bandwidth of DP1.2, which isn't enough at 20Gbps.The only option I see is if there's a new PCI-Express protocol for frame buffer transmission over Thunderbolt, running at 40Gbps. A direct write to the monitor's frame buffer via PCI Express, instead of a video signal. This completely ignores the DisplayPort protocol.
Ryan Smith - Wednesday, June 3, 2015 - link
Video data in Thunderbolt is encapsulated in Thunderbolt's PCIe-like protocol and signaling. It's not using raw DisplayPort signaling.http://images.anandtech.com/doci/4194/Thunderbolt_...
jeffkibuule - Tuesday, June 2, 2015 - link
Wasn't the USB-IF initially against Intel using USB Type-A connector for Thunderbolt before it was released?DigitalFreak - Tuesday, June 2, 2015 - link
RTFAKevin G - Wednesday, June 3, 2015 - link
Correct. The initial proposal did modify the Type-A connector to include 4 optical cable for Tx/Rx. This may have been a contributing factor.Thunderbolt over Type-C however simply uses the aux channel pins already present in the design, no additional modification necessary.
schizoide - Tuesday, June 2, 2015 - link
I am TREMENDOUSLY EXCITED about the external GPU enclosures finally becoming available.Finally, my dream of a superlight/portable ultrabook hooked up to a high-end enthusiast GPU to play games is on the horizon!
Honestly, I lay this at the feet of Alienware and MSI. These two pioneers got tired of waiting for Intel and just did it on their own, showing Intel that the market wanted eGPUs. Thanks!
Kevin G - Wednesday, June 3, 2015 - link
External GPUs is something that should have been supported since day one with the first Thunderbolt release. On the OS X side of things, external GPU's can be enabled via driver hacks but they're not hot swapable. Hopefully by the time Thunderbolt 3 is shipping in some devices, the software side will have been worked out for external GPUs.MojaMonkey - Saturday, June 6, 2015 - link
I'm using an external GTX 970 over Thunderbolt 2 as I type this. There are no software issues. This is external GPU year 3 for me as I upgraded my Lenovo S430 (Thunderbolt 1) to a Windows Macbook Pro 15 (Thunderbolt 2). The Magma eGPU enclosure was also upgraded from Thunderbolt 1 --> 2Drivers come straight from nVidia and I've never had a game crash or anything. External GPUs are already here and working fine, you just haven't bought one.
Having said that, you're smart for waiting as Thunderbolt 3 looks to be perfect :)
RKCook - Tuesday, June 2, 2015 - link
TL;DR -- By us switching to USB-C ThunderBolt will inherit a whole bunch of advantages over our previous version of the product that virtually no-one used (except as a port to plug in their DisplayPort monitor).So we now know that Thunderbolt is well and truly dead.
tipoo - Tuesday, June 2, 2015 - link
External graphics have been talked about every Thunderbolt generations. When are affordable, professionally done external TB GPU boxes coming? That's all I want. It shouldn't cost so much for just a small case, connectors, and a power supply.schizoide - Tuesday, June 2, 2015 - link
By endusers, yes. This is the first time Intel sanctioned eGPUs. I too want to see pricepoints and specs.I want to buy a thunderbolt enclosure with a pci-e 16x slot, a 350w PSU with at least 1 8-pin and 2 6-pin GPU plugs, 2 or more USB3.1 ports, a gigabit ethernet port, and both 1/8" and optical audio jacks for $300 or less. This enclosure should also charge my laptop over the same USB type-C cable.
At that price, with that feature set, they would sell millions of 'em.
DCide - Tuesday, June 2, 2015 - link
I don't know the actual parts costs, but currently every thunderbolt port (or sometimes a pair of thunderbolt ports) on a peripheral adds about $100 to the cost.Perhaps they can get this cost down with more volume, but it takes more than just connectors and a power supply.
SirKnobsworth - Tuesday, June 2, 2015 - link
There were rumors that Intel was refusing to certify external graphics solutions on previous version of TB. A few companies showed prototypes but they never made it to market. Now they're officially endorsing it, so there's a very good chance it actually happens.beta_Cyth - Thursday, June 4, 2015 - link
I can see it now. Modded Mini ITX cases on both sides of a row of monitors. One case houses the main components (Minus gpu) and the other case has GPU, soundcard, and subwoofer with speakers on the outer edges of the setup. Symmetrykpb321 - Tuesday, June 2, 2015 - link
I'm a little confused about the USB 3.1 thing. I can see how having it built into the Alpine Ridge controller would allow the port to function as a full USB 3.1 port without requiring support from the chipset but I don't see how it allows you to have a USB 3.1 port on a monitor etc without still having extra hardware there. Thunderbolt 3 is using the alt mode of the USB C connector so it is taking over the pins that would be used for USB 3.1. There are still pins to natively provide usb 2 so you could have a USB 2 hub in the monitor no problem. You could also add a PCI-E usb3 controller in the monitor and provide USB3 ports over the thunderbolt connection. I don't see how Alpine Ridge avoids that?SirKnobsworth - Tuesday, June 2, 2015 - link
My understanding is that there is a USB 3.1 controller built into the Alpine Ridge TB controller. Upstream it's still PCIe.kpb321 - Tuesday, June 2, 2015 - link
NM I reread it and understand it now. They are saying that when you use Alpine Ridge in a monitor it gives you the TB connection back to the host and gives you a USB 3.1 port on the monitor.bernstein - Tuesday, June 2, 2015 - link
in thunderbolt it hijacks ALL pins of the connector and runs a two stream thunderbolt protocol over it, signal wise this has *nothing* to do with usb3, it just uses the same copper wires and port design.as far as i see it any thunderbolt chip (be it host or slave (the monitor)) has usb3.1 baked in, so:
- if you plug a usb device into a pc with a thunderbolt port, it is attached to that chip
- if you plug a usb device into a thunderbolt monitor, that device shows up under the thunderbolt chip in the monitor, whereas the one in the pc is transparent to the system.
extide - Tuesday, June 2, 2015 - link
Basically the monitor would have to have a thunderbolt chip in it that splits the usb 3.1 off from the thunderbolt signal and then pipes it out to ports on the monitor. Hopefully that can be done cheaply.StormyParis - Tuesday, June 2, 2015 - link
OK, so TB3 does 40 Gbps with an active cable (half that with a cheap passive cable), which is about equivalent to PCI-Express 1.0 on 16 lanes (32 Gbps). If we hang our graphics card off of that, what's the real-life framerate impact in games, compared to a semi- current PCIe 4.0 @16 lanes, which reaches 252 Gbps instead of TB3's 40 Gbps.I'm guessing the impact won't be that dramatic because most textures will be in the graphics card's RAM and 40Gbps for graphics command is fine, but still, textures need to be swapped in/out of GFX RAM and some time, maybe regularly. A PCIe 1.0 vs PCIe 4.0 test would probably be a good proxy (16 lanes for active, 8 for passive), pending actual TB3 hardware and assuming TB3 efficiency is abotu as good as PCIe ? Could you guys do that so we know where we stand ?
SirKnobsworth - Tuesday, June 2, 2015 - link
http://www.anandtech.com/show/7987/running-an-nvid...Tests were done a while back.
StormyParis - Tuesday, June 2, 2015 - link
Thanks !extide - Tuesday, June 2, 2015 - link
Well, first off, PCIe 4.0 isnt out yet, so you must be referring to 3.0. Since 2.0 is twice as fast as 1, and 3.0 is (almost) twice as fast again, a PCIe 3.0 x4 slot is about the same speed as a 1.0 16x slot. There have been tests showing a PCIe 3.0 x4 slot does limit GPU perf a bit but not too much. Anandtech did an article with a GTX 780 I think on PCIe 3.0.extide - Tuesday, June 2, 2015 - link
So now that it's twice the b/w, can we get a PCIe 2.0 x8 connection instead of an x4 across it? How exactly does that work ?extide - Tuesday, June 2, 2015 - link
Or would it be like a 3.0 x4 as its the same number of pins but faster.SirKnobsworth - Tuesday, June 2, 2015 - link
It would be 3.0x4.zarmfu - Tuesday, June 2, 2015 - link
Wow, external graphics card. I have that in a 3 year old Sony Vaio Z3.Sherlock - Tuesday, June 2, 2015 - link
Quick question - does having all these different protocols on top of USB 3/3.1 mean that we will need different versions of the ports? They may look physically similar - but are different internally to support the particular protocol - or any generic USB 3/3.1 port will support these?Can I dream of a near future where all my devices use a single type of port for all I/O - Display/Keyboard/Printers/Charging/External Storage/External GPU? Is there is far away - I know theoretically you can use USB 3 to do that today?
Really excited about external GPU's - I will take this a step further - I want to use my Phone (& not Laptop) as my primary computing device - I would like to dock my phone into a larger screen & make it a tablet or a even larger screen with external GPU for Desktop gaming - Windows seems to be working towards that future.
SirKnobsworth - Wednesday, June 3, 2015 - link
Not every port will support everything. Beyond USB, everything else is optional. Stuff like DisplayPort only requires a few internal wires and a multiplexer, so it should be fairly common, but Thunderbolt, for instance, requires a dedicated controller.phoenix_rizzen - Wednesday, June 3, 2015 - link
Different ports will support different protocols, depending on the controller that's attached to the port.All of the ports will support USB 3.1. Some ports will support DisplayPort, or Thunderbolt, or Power Delivery, or whatever else comes around via USB Alternate Mode.
And there will be little pictures above/beside each port that describe which protocols the port supports.
It's 3 steps forward, 2 steps back. :( We've eliminated the USB port chaos ... only to introduce which-protocol-is-supported-on-which-port chaos. :(
gw74 - Wednesday, June 3, 2015 - link
eGPU: Finally!"The company is initially partnering with AMD on this endeavor"
*groans*
barleyguy - Wednesday, June 3, 2015 - link
*rolls eyes*Sherlock - Wednesday, June 3, 2015 - link
Thanks everyone - Agree its gonna be chaotic if ports don't support all protocols - its going to another round of fidgeting under the desk to find the right port - something I had hoped we eliminatedtildeleb - Thursday, June 4, 2015 - link
Is there anything that would prevent Apple from designing a USB-C to DP 1.3 adapter that runs TB3 in the 40 Gbps mode? The very short cable along along with active equalization in the adapter should allow the 40 Gbps mode, right?repoman27 - Monday, June 8, 2015 - link
The dearth of GPUs, scalers and Tcons that support DP 1.3 would make that a rather pointless exercise at the moment, but theoretically it could be done at some point in the future. I'm a little surprised that nobody came out with an OG Thunderbolt to DisplayPort 1.2 adapter, but I guess it would have been bloody expensive because it would have required a power supply, active cable, two 4C Thunderbolt controllers, and a DP concentrator like this one from ST: http://www.st.com/web/catalog/mmc/FM128/SS3/PF2552...It would appear that Alpine Ridge controllers are finally going to support two DP source connections, so at least a Thunderbolt 3 to DP 1.3 adapter would only require a single 4C Thunderbolt 3 controller.
benzosaurus - Friday, June 5, 2015 - link
After reading this article, I believe the title of could effectively be changed to "Intel announces Apple's Sky Lake Macbook Pro redesign."stephenbrooks - Saturday, June 6, 2015 - link
I'm rather hoping some of these features, or royalty-free equivalents, will be rolled up into a USB 3.2 spec. Not sure if they can increase the bandwidth without using active compoents though.Stimpak_Addict - Sunday, June 7, 2015 - link
This is really incredible.lorribot - Sunday, June 7, 2015 - link
Interesting but not exciting, if Intel cared anything about Thunderbolt they would include a controller at least on their top level chipsets, server chipsets or even the Z series ones to get it in to the mainstream PC world and give it some traction.USB 1 was in computers for quite a while before it got OS and peripheral support, whilst getting TB in to Apple products is a start it hardly makes it ubiquitous which is what is needed. It seems internal nonsense at Intel may be getting in the way of success for TB.
Windows 10 Continuum is an example of where TB and USB-C could make a case, a base station for a mobile phone that can drive a couple of 4k monitors and handle all the wired networking and peripherals would be a big potential sell in enterprise as a laptop replacement.
repoman27 - Monday, June 8, 2015 - link
I think you may have a different definition of "success" for Thunderbolt than Intel does.The die for the Z97 chipset is only about 52 mm^2, has a TDP of 4.1 W, and is priced around $48. The package is 22 mm x 23 mm though, because it has a ball count of 708 to support all of the various features. The DSL5520 Thunderbolt 2 controller has a die area of around 120 mm^2, a TDP of 2.8 W, and, at least according to ARK, is priced at $9.95. It also needs to be located very close to the physical ports, which is not necessarily where you want to put your PCH.
If the two were even being manufactured on the same process, integrating Thunderbolt 2 would increase the die area of the PCH by 230%, yet Intel would be hard pressed to raise the price even 20%. Sure the install base would swell, but it would be a terrible idea from a financial standpoint. Furthermore, they would still need to produce the discrete controllers for devices anyway, so unless the attach rate rises by 20 million extra devices due to the larger install base, volume will actually decrease for the discrete controllers. This could actually lead to upward pricing pressure for devices.
The bottom line is that Thunderbolt isn't meant to be ubiquitous or cheap. The use cases really are specialized, and the majority of PC users don't require it. Discrete GPUs probably represent the number one use of PCIe expansion slots by consumers, and fortunately it looks like Intel is finally allowing that to happen via Thunderbolt 3. Most folks can live without > 500 MB/s external storage solutions, single cable 5K monitor support, or 10 GbE / InfiniBand / Fibre Channel adapters for their Ultrabooks. Apple needs Thunderbolt because they no longer offer any Macs with user accessible PCIe slots. Every other OEM is still producing desktop PCs with standard PCIe expansion slots to serve customers that want to add specialized hardware to their machines, and piping PCIe over an insanely fast external serial link is never going to be as cheap as simply putting a card edge connector on a motherboard.
Zootgeist - Thursday, July 9, 2015 - link
I've been looking for a site that would explain the USB-C, Thunderbolt, and USB-C connections. I am not computer knowledgeable, and my interest in all this will be regarded as rather provincial by the folks here: I want a port that will convey data to an external hard drive as fast as that data can be written to an internal hard drive. With such a set up, one could boot from an external drive as quickly as one can from the computer's internal drive. Does that require Thunderbolt 3? Will USB 3.1 do that?I4AT - Monday, December 21, 2015 - link
So has this already made the USB-C implementation of the 12" MacBook legacy? Is that why 6 months after release, there still isn't a single-cable USB-C docking station worth purchasing? Is everyone waiting to go with Thunderbolt 3 as the new standard, like the upcoming Dell XPS implementations?I guess being stuck at 5Gbps, plus the limitations of Broadwell Core M only being able to output 2560x1600 @ 60hz make the 12" MacBook, Chromebook Pixel 2, and all current forms of the Apple USB-C adapters stop-gap technology that will eventually be bargain bin trash?
Should we expect the 2016 successor of the 12" MacBook as well as the Pro line to feature Thunderbolt 3 and a whole new line of USB-C/TB3 compatible adapters/docking stations? Or will there be a market split for high end devices using Thunderbolt 3 and entry level devices using 10Gbps USB-C Gen 2?