Hacker News new | past | comments | ask | show | jobs | submit login
USB 4.0 [pdf] (usb.org)
197 points by kalium-xyz on Feb 1, 2020 | hide | past | favorite | 77 comments

This needs a 2019 in the title.

This is a little more than USB 3.2 2x2 + TB3 but there's one crucial difference: the new, still optional USB 4 Gen 3 (what TB3 was) now carries USB packets as well. Previously TB3 only carried PCI Express and DisplayPort packets and the Intel hubs had USB root hubs in them. This solution worked out so well some PCIe enclosures opted to use two TB3 controllers just to get a problem free USB experience...

Also, PCIe was nerfed by Intel in TB3 for unknown reasons to 22Gbps, there's hope it'll be 32Gbps this time for real. You can still run a 3440 x 1440 @ 60 Hz monitor on the remaining bandwidth (or to spew marketing BS, 4k @ 30 Hz).

>Also, PCIe was nerfed by Intel in TB3 for unknown reasons to 22Gbps, there's hope it'll be 32Gbps this time for real. You can still run a 3440 x 1440 @ 60 Hz monitor on the remaining bandwidth (or to spew marketing BS, 4k @ 30 Hz).

I'd just like to clarify, and add some info:

- you're not running the video over pcie

- ...unless you use an egpu (which has been a mostly problem free experience for me but wow YOUR MILEAGE MAY VARY)

- the "problem" with pcie-via-tb3 is not bandwidth, but latency. You can run "normal" GPUs on 4x with minimal penalty for most games.

- problem free USB experience... if you're using titan ridge with modern drivers - some laptop and enclosure manufacturers ship and include in their built-in updating software old drivers. Get the latest from intel and hope for the best. Alpine ridge is still buggy, and also quite limited in many ways. Fortunately, most laptops sold in 2019 shipped titan ridge. The only way to tell if you have alpine ridge is to count how many tb3 capable usbc ports you have, if the laptop you bought maxed out the usbc ports - which it might not have on an ultraportable. insane.

- you can only get the highest speeds (>20Gbps) over a 6inch cable or a USB Type-C Active Cable. Both sides of the cable include a chip which performs clock, sync, and some power negotiation of the link


My biggest issues with my eGPU have been with Windows' weird limitations on addressing space for devices, so if you have 3+ graphics cards (builtin primary, builtin secondary, eGPU card) you end up having to endlessly screw around with disabling and reenabling devices and device buses to change the order they're initialized in, messing with bootloaders to change which PCIe lanes get loaded at startup, etc etc. It's a nightmare.

By contrast, all my experiences on Mac have been either "it just works" or "after one reboot it just works". Still not ideal, but not anything close to the 4+ hours it took me to get it working on Windows for the first time (and after which it immediately stopped working again the first time I disconnected the Thunderbolt cable).

At work I have a late model Dell laptop that works well with a Dell dock. My year older Alienware (also Dell) has struggled to work with various docks, but I have built a network of single purpose USB gadgets under my desk (Hub, Ethernet, SD card, XBox dongle) that is rock solid that plugs into a USB-A 3 and leaves the USB-C for the monitor.

My biggest issue with the eGPU, on an Intel Skull Canyon NUC so Intel from top to bottom including Thunderbolt technology, it was very flakey with the eGPU and the computer frequently getting locked into a dysfunctional state. Sometimes I'd spend 20 - 30 minutes cycling through powering on and off before it all worked.

Now I use it with a Macbook Pro and it works very smoothly, even using a TB2 adapter. I can only use it in bootcamp because of Apple's policy towards nVidia, but it works every time and disconnects every time.

Even with the troubled experience, being able to move a GPU between computers is really just awesome, I wouldn't buy another machine without Thunderbolt or USB 4 eventually.

>if you have 3+ graphics cards (builtin primary, builtin secondary, eGPU card)

I have not had any of these problems. Intel GPU, Nvidia MX150, AMD RX 580. I wonder if there's some weird driver interaction stuff?

But yeah, the technology is clearly not at perfect maturity. All I can offer are condolences :(

My problems haven't been with drivers, but rather with Windows being unable to allocate enough PCI resource lanes even though the hardware can physically support everything.

> USB 4 Gen 3

Are they going to do that thing again where they rebrand the existing standards from USB 3 Gen X to USB 4 Gen X?

Yes they are, at least for now.

I would actually like a Type C sized USB port that is magnetic.

Kind of like the magnetic Apple charger for the older MacBook.

Why magnetic? So that I can quickly attach and detach a flash drive to it, without actually plugging it in. And then have an external mechanical locking pin or mechanism, to hold the device or cable in place, if you need to better secure it.

Why would this be superior?

* You eliminate the actual wear and tear of the USB metal housing.

* The external locking mechanism can be easily repaired or replaced when it wears out or breaks.

* It would probably help make your computer more waterproof.

A couple possible issues, mostly inherent to magnetic connectors like this:

* attracts everything magnetic and mettalic, not just the connector - need to be careful no to attract to many metal turnings, washers, etc. * the number of contact points & the quality of contact might not be enough for high speed communication * the contact pads/pins might be more exposed than normal connectors, so more likely to be damaged or accumulate dirt

Apple avoided this by making it an "intelligent" port. There is a bit of handshake that occurs when connecting the magport charger. Until this handshake happens, the power does not turn on. When first connected the LED turns amber to indicate the connection has been made, but the electrons are not yet flowing. Once the handshake occurs, the electrons start flowing and the LED turns green.

It would be cool to go one step further and have the handshake enable an electromagnet.

That would be kind of pointless though, wouldn't it? The handshake only takes place after the cable has been physically connected to the port. You would need to manually guide the connector to where it needs to be and then keep it stationary until the handshake is ready (which might not be a very long time in human terms, of course).

The part I liked about the magnetic connection was that yanking the cable wasnt catastrophic. The secondary benefit was that the connector had basically zero depth.

what about accidentally unplugging and corrupting the data? it's fine for power cable, but not sure how much for data. It would need some sort of locking in place.

There's some magnetic USB-c adapters you can get on Amazon. They kinda look like pogo pins

I was under the impression these only work for usb c power distribution; do they also work for data, display, etc?

Yes. Some have only 2 leads, but the newer ones have all of them.

Any recommendations for a good one? I know the Type C market is rife with poor quality stuff.

Knaive KD3 [1] has worked well with my MacbookPro. It supports 100watts of USB-PD and USB 3.1 (10gbit). I've had no problems doing 87watts, 1080p @ 60hz HDMI and ethernet/USB over a single cable.

USB 3.1 only has bandwidth for 4K @ 60hz w/o USB; so with non-thunderbolt cables (like these magnetic adaptors) it's either Power+4K@60Hz or Power+USB+4K@30Hz.

[1]: https://amzn.to/38ZXPVa

We've been using these on our Macbooks at home: https://www.amazon.co.uk/gp/product/B07GHZC8FN

I've been using this one: https://www.amazon.com/gp/product/B07MMKZ8XD

All 24 pins are connected, so video, data, and power all work as expected.

>I would actually like a Type C sized USB port that is magnetic.

I would love one to. Except there wont be another port once EU pass the USB-C law.

This is from Sep 2019, HN discussed it here: https://news.ycombinator.com/item?id=20870326

Does anyone knowledgeable know if a magnetic coupling is coming for USB?

I have used volta cables for a while (I preferred the 1.0 to the 2.0) and I'm surprised that this design idea isn't more widely popular, given how user friendly it is.

USB-IF has no plans to include magnetic retention in the spec for USB. The onus is on manufacturers to implement this on their devices and cables if they wish for this to become a recurring feature on the market. There might be a more detailed reason as to why and if you want, I can seek an official response from the USB-IF team and their association management company, VTM Group.

Thank you for the offer. It would be interesting to hear their response. It’s seems there are a lot of people who like magnetic retention. I miss MagSafe :)

That would be awesome. If I were to say anything, it would be to emphasize that given a) the overwhelming consensus of interest from technical niches for magnetic attachment AND b) Apple's apparent acceptance of the current general status quo, it will not be long before general customer demand for magnetic coupling reaches a tipping point - while completely un-standardized.

One only needs to glance at USB 3.0, 3.1 and 3.2, the rise of 4.0, ThunderBolt, the current cable disaster, etc, to appreciate why adequate standardization would be a net positive for the market (less friction, better adoption - benefiting consumers _and_ vendors), and provide for smoother industry adoption going forward.

I understand that USB-IF standardized all the plug formats. This was critical to get right in 1999, otherwise nothing would have interconnected properly and USB would never have caught on. If we are in a similar situation today, standardization of magnetic coupling is no less critical.

Beyond the technical factors, standardizing magnetic attachment critically prevents vendor lock-in by eliminating the value proposition for introducing unique and deliberately incompatible designs. Without standardization, there is currently unlimited potential behind establishing vertically integrations (walled gardens) around custom designs. A great example of where this can benefit vendors while fragmenting the market and creating terrible user experience is the charger industry, which the EU recently began to enforce stronger restrictions on to try to eliminate e-waste (https://news.ycombinator.com/item?id=22204174). USB-IF is likely familiar with this case.

Given the huge convenience factor and the practically infinite ways magnetic coupling attachments could be designed, it's plausible that USB might end up in a world of standardized plug-based connectors next to some n>1 quanity of vendor-specific approaches to magnetic attachment. (This is already happening, albeit as USB-IF ratified plugs next to arbitrarily-designed magnetic adapters built into cables.)

While I'm not especially familar with the technical details, I get the impression the market may carefully be circling Apple and their use of the MagSafe connector, and perhaps also related patents that may not yet have expired. I wonder if it's possible nobody wants to be the first to ask the question, for fear the answer capitalizes on the situation in a way that does not unilaterally benefit all. The only solution may be careful timing - waiting for successful market penetration, then quickly working to standardize before vendors get orientated.

(Clearance: feel free to quote and/or use for parts if/as you wish)

Please do. Putting the onus on manufacturers basically ensures the connector will not be "Universal", which is the most important feature of USB.

Please, do ask them. Also remind them that giving this much freedom to the market inevitably leads to attempted vendor lock-in.

Standardise as much as possible.

This a hundred times over. Type C is better than micro, but a magnetic port would make much more sense for most mobile consumer devices like laptops and smartphones.

Type C magnetic cable exist, I have two of them.

My understanding is that none of the Type-C magnetic cables support TB3 or high power options. So they can't be used for connecting a laptop to a docking station, for instance.

Amazon has a whole bunch of those cables/adapters that claim to support 100W/5A charging. Something about being "20 pin" instead of "24 pin" makes even the ones that support high speed links not support thunderbolt, but I'm having a hard time figuring out exactly which pins aren't connected. (It might be the two that are never connected, plus the sideband pins? But that would need awkward wiring, so maybe it's sideband and CC? But that would break more things, I think. So maybe some of it is power pins? But there's 8 of those by default, so why would dropping 4 pins cause it to lose thunderbolt support?)


• Two-lane operation using existing USB Type-C® cables and up to 40Gbps operation over 40Gbps certified cables

• Multiple data and display protocols that efficiently share the maximum aggregate bandwidth

• Backward compatibility with USB 3.2, USB 2.0 and Thunderbolt

Thank you. I appreciate the summary.

God I miss the days when I could tell what a port was capable of by looking at it.

I've been soldering, crimping, and plugging things in for forty years, and I struggle to think of a connector for which there is exactly one configuration, pinout, cable specification, capability and equipment compatibility for all its use cases seen in the real world.

In my household, USB Type-C has in practice been a great simplifier. Less so at work, but still not more complicated than the multifarious uses of, say, the modular connector family, or the DIN series.

I am less experienced, but my USB C experience has been less positive. Both my partner and I have USB-C phones, from different manufacturers. Both use different "fast charging" standards, and neither of the fast charging standards appears to work as well as with their respective manufacturers chargers (which as a side note are incredibly difficult to find replacements for). I also have a Nintendo switch, which doesn't charge with my MacBook pro's charger, and my switch charger doesn't power my MacBook properly. My car has Android auto, however the cable I was using would cut out if I played audio and used navigation, so I had to buy another cable from a different manufacturer (that cost the same as he first one) for it to work.

I've gone from multiple chargers with different ports, to multiple chargers with the same port, and specific cables with tags on them, and having to explain to my partner that you can't plug x into y, even though they all have the same port.

* Devices are a 2015 MacBook pro, an external GPU, a Huawei phone, a pixel 3, a Nintendo switch, two sets of wireless earbud. For those, I have 4 specific chargers, and two specific cables (3 if you count the moulded plug to the Nintendo switch).

It's quite late here and the Huawei charging "standard" has really thrown me, but:

* Pixel - 9V/2A or 5V/3A by the looks of it. A decent USB C charger will do this

* Nintendo Switch - just don't use anything other than the Switch charger or a power bank for this. It's 15V/3A, but finicky.

* Macbook Pro - not familiar with this, didn't realise they went USB C all the way back in 2015, but it probably needs 60W+ and an e-marked cable.

* Huawei phone - if you give me the model number I can probably work this out. It's very hard to work out what that charging spec is, but it may also work with Qualcomm QuickCharge, which apparently makes the USB-IF very sad when delivered over the Type C cable.

I use a charger similar to https://www.kogan.com/au/buy/mbeat-gorillapower-80w-5-port-u..., and I also highly recommend something like https://satechi.net/products/satechi-type-c-power-meter-for-....

I appreciate the response. It's a p20 pro. I realize that I could find a charger that's compatible with all of the devices, but the fact is that all of them claim to be USB-C, and I'm still not sure if I buy a third party charger from a trusted brand (e.g. Anker) if it will work with all devices.

Now imagine explaining your comment to a lay person!

> it probably needs 60W+ and an e-marked cable.

I wasn't even aware e marked cables were a thing. I guess I need something similar for correct Android auto support with my phone and my car?

Honestly, your comment, while incredibly helpful, explains how unclear and how much of a mess it is.

Half of the problem is the USB implementers forum seems to be disconnected from reality and the other half is manufacturers not wanting to use standards

P20 Pro: Uses "Huawei Supercharge" which is not Huawei Fastcharge or VOOC, nor is it compatible with USB PD or Qualcomm QuickCharge. Jesus. I'll cross that brand off my list completely

re: Android Auto - I have no idea, sorry. Double check the cable works correctly to transfer data to a PC (or for tethering etc) - the last time I looked, AA was VNC over IP over USB.

Agreed that the problem is the implementors, but unfortunately that's all that matters. At least when the cables didn't fit I could easily not buy the wrong cable, but now I'm left scouring reddit or asking for help on HN!

> * Nintendo Switch - just don't use anything other than the Switch charger or a power bank for this. It's 15V/3A, but finicky.

The dock can be fussy, but I think the switch itself is fine with anything that doesn't horrifically overvolt pins.

That's kind of the point though, right? You've got a dock which takes a standard port, same as the console, except _neither_ of them actually respect the standard. It would probably more convenient if they just used the wii u power brick instead

In what way does the switch itself not respect the standard?

Using a generic power brick for the dock wouldn't have much impact. But changing the console itself would be a big loss toward charging it on the go.

Ive found mentuons that certain USB c cables won't charge the device itself, e.g. https://www.reddit.com/r/oneplus/comments/5xexiz/fyi_the_one...

It's very confusing to say to a consumer - hey, the cable that fits your phone won't actually charge your switch, even though if you switch cables, it will work with both. Except of course, your phone won't fast charge.

I don't think dash cables are compliant. That's not really something you can blame USB for.

Of course you can't blame USB, they didn't make the cable. But if you have a one plus phone, a Huawei phone and a Nintendo switch (3 very common devices) that all claim to use USB, and yet none of the devices chargers are compatible with each other, who do we blame?

At least when my 3ds, Nokia and razr phones had different connectors, I wasn't going to be under the impression they would work with each other

It doesn't like special cables, though. For example, it won't charge off a Dash/Warp charge cable like Oneplus uses.

Would you rather each of those use a custom port and none of those things be compatible at all?

They already are incompatible. Just in a much more annoying fashion than ever before because the incompatibilities don't show themselves until you actually try it.

I much prefer to just look at the cable and go 'nope, that clearly doesn't work with that' instead of playing cable roulette.

Honesrly, yes. It would be easier to explain to my partner which cable/device/charger combo she can use, and would be easier to find a replacement that I know will work.

For all intents and purposes, they _are_ incompatible already, they're just pretending to be.

The innergie 60C Charger supports a lot of different voltages. And is super small

There has just never been a large enough variety of connectors for each application to have an ideally specialized standard.

Especially the greater number of pins, the wider variety of possible uses.

But USB is only 4 pins.

IIRC when USB first appeared on consumer devices, Radio Shack was still going, but USB was the first consumer connector that they did NOT have a do-it-yourself solderable or crimpable cable terminal for, nor any truly appropriate cable. And they never did appear as years went by.

You had to cut your cables in half and splice them to appropriate extensions if you wanted to take advantage of the full 92 foot cable-length maximum under USB 1.

What ports are you thinking of? PS/2, parallel, serial, VGA, HDMI, DisplayPort, Ethernet, USB since 2001... they can all vary in capability. DVI might not? But you have to look really closely to make sure you have the right kind of DVI.

DVI is a pain! Is it analog only? Digital only? Dual link capable?

So glad that HDMI superceded that one.

Yeah, about HDMI... why did DisplayPort get invented?

higher performance, simpler, can be used internally and royalty free

Hmm? All DVI is digital (except for the analog vsync line).

DVI-A is an analog-only signal that runs over a DVI connector. https://en.wikipedia.org/wiki/Digital_Visual_Interface#/medi...

Yeah but with those you still vaguely know what the port is used for. I have devices with a type c and a micro USB port that only use the micro port for charging. I have devices with a type c port and a separate power connector that can charge over the type c port but only at 500mA. I have devices that can output video through their type c port but only when using a Display Link adapter which on paper needs USB 3 but the device only supports USB 2 data throughput.

> I have devices with a type c port and a separate power connector that can charge over the type c port but only at 500mA.

What a weird design, especially for not even being 900.

> I have devices that can output video through their type c port but only when using a Display Link adapter which on paper needs USB 3 but the device only supports USB 2 data throughput.

They did nonstandard video output on all kinds of previous phone ports too.

But what part says you need to support USB 3 to support an alternate mode with video?

I'm gonna go out on a limb and suggest that most of those devices are awkward eBay crapware?

Nope. The first is an Asus laptop, the second is an HP tablet, and the third is my Oneplus 6.

Serial ports (RS-232) were one of the most variable ports ever used. Each device that you wanted to connect to a computer required a differently wired cable. Wires mapped to different pins, pins soldered together. It was a nightmare.

DVI had the plug-of-death syndrome. the cross-of-metal was mechanically about the worst design choice they could have made. (the wierd connector Sun used, which had three vaguely analog coax connectors was maybe worse)

ok. truthfully, the sliding latch on the Ethernet AUI cut my fingers more. I hated that.

Ha. I hated that damn AUI connector latch. I remember giving David Boggs grief over that c. 2000? He said he thought it was a terrible mistake as well and I think the reason given was it was something they found/was the easy quick choice.

DB13W3? All the old Unix workstations had that video connector, my silicon graphics boxes need a sync on green compatible monitor, so just adapting that connector isn't enough.

Yep. thats the one. love that cable (not)

This particular document doesn't communicate value to the consumer. I have much love for USB 1, 2, 3 and if were not for those USB-C connectors that don't seem well connected to the circuit board I'd be satisfied.

I always get a little surprised when I see the new USB logo on things. It just looks so stylistically different from the old trident + "USB" logo that it's jarring.

I want to start seeing these USB4 ports in laptops though to free us up from Intel+Thunderbolt.

Huh? Intel invented Thunderbolt. USB4 is Thunderbolt 3.

> USB4 is Thunderbolt 3

No, USB4 is a successor of Thunderbolt 3 (but similar enough that it can be backwards compatible with Thunderbolt 3). For instance, Thunderbolt 3 tunnels only PCIe and DisplayPort, while USB4 also tunnels USB 3.x.

I agree with this analysis.

USB4 is more of a Thunderbolt 4. Not only can it tunnel USB natively, it seems likely to be supporting a newer higher speed higher than TB3's 40gbps (but if it don't, I'm still happy with 40).

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact