41

At work, I want to plug all my USB devices into a single USB hub so I only need to plug one USB cable, external display, and the power cable into my laptop. I have some USB 3.0 devices and some slower USB 2.0/1.1 devices.

I'm aware that all the devices on a single USB Root Hub share the same bandwidth, but I'm curious whether plugging an older device into a hub causes the entire hub to fall back to a compatibility mode.

If I plug the slower devices into my USB 3.0 hub, will that hub and all its connected devices slow down to USB 2.0 speeds, or will the USB 3.0 devices continue to run at USB 3.0 speeds?

For example, suppose I have an USB 3.0 gigabit network adapter and an USB 2.0 keyboard. If I plug the keyboard into the same USB 3.0 hub that the network adapter is plugged into, will the network adapter's maximum theoretical throughput instantly drop to 480 Mbps or slower (USB 2.0's maximum throughput)?

rob
  • 14,148
  • 5
  • 52
  • 85
  • For connecting USB 1.1 devices to USB 2.0 hub, it depends on whether the hub has a single or multiple transaction translators. For USB 2.0 devices in USB 3.0 hub, there is apparently no such thing, though: https://en.wikipedia.org/wiki/USB_hub#Transaction_translator – endolith Feb 04 '16 at 15:11

2 Answers2

42

Short answer: No.

Long answer:

I stumbled across the answer to my question in a comment to an answer for a seemingly unrelated question. It turns out USB 2.0 and USB 3.0 are physically segregated in the wiring, so plugging USB 2.0 devices into an USB 3.0 hub won't affect the performance of USB 3.0 devices in the way that I was wondering.

Community
  • 1
rob
  • 14,148
  • 5
  • 52
  • 85
  • @Ben Can you clarify something? If I have two USB2 webcams, normally each would saturate a USB2 link. If I connect both to a USB3 hub, are you saying that they would each be able to use 480 Mbps since the hub would aggregate traffic and resend over the 5 Gbps USB3 link to host? – Nicu Stiurca Sep 30 '15 at 14:16
  • @SchighSchagh: Yes, barring some major design problem in the hub, that's exactly the advantage. (Well, neither device will reach 480 Mbps, but they will each receive as much bandwidth as if there were no other device attached) – Ben Voigt Sep 30 '15 at 14:37
  • @SchighSchagh Wikipedia says that doesn't happen: https://en.wikipedia.org/wiki/USB_hub#Transaction_translator – endolith Feb 04 '16 at 15:28
  • @endolith That article is mentioning USB 1.x devices plugged into USB 2.0 ports, *not* USB 2.0 devices plugged into USB 3.0 ports. – clarkcox3 May 06 '16 at 19:06
  • @clarkcox3 From the link "Note that USB 3.0 hubs do not currently perform transaction translation to super-speed for USB 2.0 devices." USB 2.0 devices have to share a USB 2.0 bandwidth. – endolith May 06 '16 at 20:19
  • @endolith Yes, but that says nothing about this question (i.e. even without the transaction translation described in the article, you could still go either way on the question of whether or not a USB 2.0 device will slow down a USB 3.0 hub) – clarkcox3 May 09 '16 at 16:10
  • 1
    @clarkcox3 I'm saying that BenVoigt's comment is incorrect; the USB 2.0 signals are not translated to USB 3.0 signals, so all the USB 2.0 devices must share a single 480 Mbit/s bandwidth. If you plug two USB 2.0 devices into a USB 3.0 hub, the USB 3.0 wires from the hub will have no data on them. – endolith May 09 '16 at 16:51
  • 1
    @endolith: You can say that, but the better USB 3.0 hubs do have transaction translators (to SuperSpeed). See the VL670 for example, dated 2014. – Ben Voigt Jan 28 '18 at 05:05
  • At least the manufacturer page says 2014, it's always possible that does not reflect reality. – Ben Voigt Jan 28 '18 at 05:16
  • 1
    I have a Lenovo E50-80 (model 80J2) - it features 2 USB 3.0 ports. I can use 2 1080p USB 2.0 webcams on these ports simultaneously even though each webcam draws about 15 megabytes/s alone and they will not work together when connected via a USB 2.0 hub. This means the USB 3.0 must do transaction translation from high-speed to super-speed. Surely such a transaction translation may not be a part of USB3.0 specs but in reality there seem to be devices that do perform this. I monitored the data traffic with usbtop utility. Anyone has similar experience? – Kozuch Jul 13 '18 at 12:37
  • @Kozuch it's also possible you just have two USB3 Busses. – HilarieAK Oct 03 '20 at 13:33
  • 1
    Transaction translation is not built into the USB 3 spec as it was for USB 2, and you can assume the *default* configuration for typical hubs is that USB 2 communication is *not* translated to USB 3 speeds and re-transmitted, but instead is routed via the traditional USB 2 bus. The interesting benefit of this is that USB 2 and earlier devices occupy zero bandwidth of the USB 3 bus, so they will not slow down or demote USB 3 communication, even by a small amount. But the drawback is all USB 1/2 devices are relegated to share a combined 480Mbps bus. – thomasrutter Apr 26 '22 at 07:45
-2

No.

Here is how to imagine speed in wires.

Think of it like a tunnel that transfers water.

Assume a big tunnel is split evenly into several similarly-sized tunnels (HUB). Next, you connect your 2.0 device (which is a smaller tunnel comparing to 3.0) to the hub.

What happens? Do the other tunnels get smaller? No, they do not. Your 2.0 device uses as much speed as it can.

This analogy really helped me to understand network problems.

Dennis Wurster
  • 283
  • 2
  • 10
UltraDEVV
  • 473
  • 6
  • 20
  • 10
    This analogy is not correct for a wide variety of network (really "bus") situations. For example, USB 1 devices on USB 2 will slow the bus down by occupying a disproportionate amount of time for the data transferred (blocking other, faster devices from using the bus). Same for 10mBit devices on Gigabit networks. RS-485, SATA, and other serial buses only go as fast as the slowest device. Not true for USB3 however, because the cable contains DIFFERENT WIRES for v.2 and v.3 of the protocol. It's the difference between putting minivans and Ferrari's in the same lane vs each getting their own lane. – DrFriedParts Jun 22 '16 at 11:11
  • 1
    @DrFriedParts: You're not correct for Gigabit Ethernet networks though. You _would_ have been correct if it was a 10/100 shared-medium Ethernet, but modern switched networks (which is required for Gigabit) no longer act like a "bus" of any sort. – u1686_grawity Feb 01 '17 at 06:55
  • @grawity -- You have to be more nuanced than that. Your statement is only true if the gigabit switch has enough lanes (i.e. most "managed" hubs) and if you can avoid the problem scenario typically involving simultaneous transfers from a single gigabit-equipped client to a mix of gigabit and Fast Ethernet clients. 802.3 flow control, switch fabric bandwidth, and other limits still affect gigabit switches. – DrFriedParts Feb 01 '17 at 09:38
  • @grawity -- there are no "Gigabit hub's" because GigE requires point-to-point links. What there are are "unmanaged switches" and these typically have insufficient lanes to allow all ports to communicate with all ports (because that very rarely happens). So when ports use the same lanes in the switch they must wait for the slower transfer to complete before the lane becomes available and you have exactly the same scenario as in USB1/2. This occurs rarely, but it does happen. It most commonly occurs with "a single gigabit-equipped client to a mix of gigabit and Fast Ethernet clients." – DrFriedParts Feb 07 '17 at 00:37
  • Interestingly, broadcast traffic through any network switch will tie up the switch and operate at the slowest speed of all connections. If one of those connections is a PC in sleep/off mode, that will often mean it maintains a 10Mbps connection on its network port, inadvertently limiting all broadcast traffic on the network to 10Mbps. Such traffic should not comprise a high proportion of traffic in normal situations, but there are some monitoring or media casting devices which *do* cause a lot of broadcast traffic in normal operation. There's a related problem in wifi too. – thomasrutter Apr 26 '22 at 08:02