I could see this used for livestreams where you have a lot of cameras to connect. The 3 independent video outputs also make this great for a multi monitor setup.
I develop hardware. At any given time I have 3 programmers, various charger cables, usb to serial cables, logic analyzers, a mouse, a spacemouse (for CAD), flash drives, 2 x label printer (strip and shipping) and various other stuff plugged into my 12 port hub. And I still have to switch them out due to lack of ports. And hubs don't play well with high speed USB3 based logic analyzers.
Well, that would depend on how much power you're willing to pump through on PD. PD doesn't in and of itself mean the downstream facing port support the full 100W (20V/5A). You could conceivably add a PD billboard and pump out the minimum supported wattage, which I believe is 150mA @ 5V, so 0.75W?
[edit] Looks like PD2.0/3.0 supports 100mA @ 5V minimum. [1]
Now what you could do that would be really neat is have a set power budget for all your ports, say 250W, and allocate that power on a first-come first-served basis. Provide a minimum of 5W per port, and then the rest is allocated in accordance with the power delivery negotiation protocol.
Say you plug in a laptop, it requests 85W, and you give it 85W. This leaves a power budget of roughly (250W - (minimum)(20*5 = 100W) - 85W = 65W).
Mac Pros had interchangeable CPUs and Logic Boards IIRC, so as they may as well call it Brainzz however, it won't change the function of the board itself. :)
The explanation I got for logic board was from the classic all-in-one Macs. They had a high voltage “analog board” for the CRT, and the low voltage digital “logic board” for the CPU.
Windows won’t have enough logical resources even they did away with IRQs and stuff. I’ve run into this at around a dozen USB devices—got to find the article on this.
Not sure, if your experience is more dictated by USB hub limitations, rather than hitting a Windows limit that soon. Here’s an interesting page describing how to best get to high USB device counts[0]
I can definitely see this happening if you need to run a farm of printers (be they of the 2D or 3D variety) and Ethernet ain't an option. I don't know of any such printers that can even use USB3, though (let alone require it), but still.
Indeed, a buddy of mine has at least 6 or so going at once for his 3D printing business, and we're considering a similar sort of strategy (w/ e.g. some box running Slic3r and Octoprint) so he ain't constantly running USB sticks back and forth.
That's a pretty bad idea, the usb virtual serial port used by most 3d printers, because of it's low throughput, can worsen print quality, it's is much better to just upload to the sd via octoprint. For 2d printers I haven't seen any without network connectivity in a decade, and I can't imagine anyone needs 20 2d printers that are so old or cheap they don't network.
> it's is much better to just upload to the sd via octoprint
That's... pretty much exactly the plan, yes. The printers we're using all (to my knowledge) run some version of Marlin w/ SD support, so we'd be doing this via the M28/M29 GCode commands over that same USB interface.
It ain't clear that OctoPrint supports this directly, but as long as it's able to shove the exact GCode we tell it to shove we should be able to make that happen. Since Marlin's supposed to wait until it gets an M29 command before it starts printing, it seems to me like low throughput would therefore be a non-issue (aside from, you know, the amount of time it'd take to push that data across, but that'd really only be an issue for the first time we're printing a new file).
That's ok. People have different needs. For me it is mostly many hard drives, exercise equipment, multiple cameras, some stuff I am testing and lots of smaller things.
I’ve got a bunch of audio related USB kit. Audio interfaces (sound cards), midi controllers, samplers, external drives, control interfaces, synths, etc. Fills up the better part of two 8 port hubs.
But in the book "How We Test Software at Microsoft", they mentions Microsoft test USB implementation with "USB Cart of Death", which including 10 of 8 port usb hub:
It might be true if you have old devices, though. If you have USB 1.0/1.1 full speed HID devices you can overcommit the available bandwidth for a human interface device. HID devices generally require that they be serviced every 1ms with a 64 byte packet. You might overrun that if you put enough old devices together on an old hub. It's quite unlikely for USB 2.0+ with anything High Speed.
The only issue with new devices is that USB 3.0+ requires twice the endpoints of USB 2.0. The problem is that the Intel controller chips allocate the same amount of memory for USB 3.0 as they do USB 2.0. So if you have a system with a lot of devices (96+), you can hit the memory limit of the controller.
We hit that limit with a 192 device system. It took a while to figure out what the heck was wrong.
Reading the comments I'm surprised so many people think 20 usb ports is so out of this world. For me this would be great! The list below isn't 20, but its still 14 and honestly I could find more things to keep plugged in w/o much effort either.
My understanding was the second plug powered the LED lights in the keyboard but now I'm not 100% sure if its for that or just for the passthrough USB port on the keyboard itself.. maybe both?
Sometimes USB keyboards with PS/2 backward compatibility have 2 plugs, one for they keyboard you can use with a USB to PS/2 passive adapter, and one that supports a built-in hub or some other secondary high-power feature such as RGB.
I suspect this has more to do with PS/2 compatibility than it does requiring a second port's worth of power.
cable to connect TI calculators (I have a USB version of this too, though, so I don't need to use a USB-to-serial converter with my old serial cable based one.)
And that's just what comes to mind right now, I'm sure there are more.
Apple didn't move to USB-C only; they moved to the vastly-superior Thunderbolt 3. (Which, yes, is a superset of USB-C.)
Furthermore, that's just on laptops, since USB-A ports no longer fit in Apple's increasingly-thin designs. The iMac still has four USB-A ports, right now, today. The Mac mini still has two USB-A ports. The Mac Pro can have dozens of USB-A ports, if you want them.
False. That proves you have not looked at the current MB Pro design. You are also blithely assuming there is space inside the case to just randomly add backwards-compatible shit that we all need to move on from in order to enable and promote more capable interfaces.
If you have "tens of devices" you need to plug into your computer at once, then stop buying laptops or get a USB hub. This isn't complex. No laptop ever designed, in history, would allow you to plug in that much USB-A stuff at once without buying adapters or a hub, etc.
So stop bitching at Apple about irrelevant stuff that Apple isn't to blame for.
On a second thought, this would be the dream industrial process, control or monitoring board.
A lot of monitoring hardware runs on legacy protocols (ModBUS comes to mind) which are encapsulated over and over newer protocols. I'm sure most of these controllers also come in USB flavors. Moreover, USB 3.0 is full duplex and supports more power than 2.0 This means better communication in real time monitoring applications using more complex controllers/adapters.
So, it may not be a speed-first but connectivity-first motherboard which is optimized for many small data bursts over many USB ports. Three displays will also help showing the information obtained from these devices in a nice videowall-ish view.
i just want to point out that a USB4 connection is good for 40Gbps, which is 2.5x PCIe 4.0 lanes of throughput. 20 ports of USB4 would take 50 PCIe 4.0 lanes to saturate.
i'm not sure how many actual pcie lanes USB4 can transport? 2x?
66 comments
[ 3.4 ms ] story [ 141 ms ] threadThis machine is clearly designed to straddle the DMZ in an infosec army. And/or Goldman Sachs, which is the same thing.
Phone farms? 20 random Android phones as physical emulators. More with hubs.
Testing USB flash drives? Flash drive factories use standard Windows boxes to test new drives and pre-load with bloatwares.
You mean they do this manually by plugging each drive? I would have thought the "bloatware" was already written on flash.
I've re-programmed a handful of flash drives of my own to mark damaged cells and in some case restore a flash storage map to return the drive to life.
Transcend has a one-click utility which does everything automagically to restore malfunctioning drives to best possible condition.
Flash controllers are much more intelligent than they look. They're almost black-magic like.
With active USB3 cables you can attach 20x 4K terminals, up to 10m away from the computer, easily.
[edit] Looks like PD2.0/3.0 supports 100mA @ 5V minimum. [1]
Now what you could do that would be really neat is have a set power budget for all your ports, say 250W, and allocate that power on a first-come first-served basis. Provide a minimum of 5W per port, and then the rest is allocated in accordance with the power delivery negotiation protocol.
Say you plug in a laptop, it requests 85W, and you give it 85W. This leaves a power budget of roughly (250W - (minimum)(20*5 = 100W) - 85W = 65W).
[1] https://en.wikipedia.org/wiki/USB_hardware#USB_Power_Deliver...
Apple calls motherboards as logic boards making non-apple devices look illogical piles of electronic components. :)
https://dortania.github.io/OpenCore-Post-Install/usb/#macos-...
[0] https://www.yoctopuce.com/EN/article/how-many-usb-devices-ca...
Indeed, a buddy of mine has at least 6 or so going at once for his 3D printing business, and we're considering a similar sort of strategy (w/ e.g. some box running Slic3r and Octoprint) so he ain't constantly running USB sticks back and forth.
That's... pretty much exactly the plan, yes. The printers we're using all (to my knowledge) run some version of Marlin w/ SD support, so we'd be doing this via the M28/M29 GCode commands over that same USB interface.
It ain't clear that OctoPrint supports this directly, but as long as it's able to shove the exact GCode we tell it to shove we should be able to make that happen. Since Marlin's supposed to wait until it gets an M29 command before it starts printing, it seems to me like low throughput would therefore be a non-issue (aside from, you know, the amount of time it'd take to push that data across, but that'd really only be an issue for the first time we're printing a new file).
That's ok. People have different needs. For me it is mostly many hard drives, exercise equipment, multiple cameras, some stuff I am testing and lots of smaller things.
https://books.google.com.tw/books?id=X5lCAwAAQBAJ&pg=PT414&l...
Edit: Maybe it is no issue with USB hub, but not work with too many usb controller (USB port to MB directly)?
It might be true if you have old devices, though. If you have USB 1.0/1.1 full speed HID devices you can overcommit the available bandwidth for a human interface device. HID devices generally require that they be serviced every 1ms with a 64 byte packet. You might overrun that if you put enough old devices together on an old hub. It's quite unlikely for USB 2.0+ with anything High Speed.
The only issue with new devices is that USB 3.0+ requires twice the endpoints of USB 2.0. The problem is that the Intel controller chips allocate the same amount of memory for USB 3.0 as they do USB 2.0. So if you have a system with a lot of devices (96+), you can hit the memory limit of the controller.
We hit that limit with a 192 device system. It took a while to figure out what the heck was wrong.
- mouse
- keyboard
- keyboard extra power
- audio
- streamdeck
- webcam
- printer
- photo scanner
- card reader
- phone
- microphone
- digital camcorder
- controller
- usb drive
Come again?
I suspect this has more to do with PS/2 compatibility than it does requiring a second port's worth of power.
- yubikey
And those of us working on electronics routinely connect another bunch of devices:
- logic analyzers
- oscilloscopes
- JTAG interfaces
- USB microscope
- dev board interfaces
- android embedded devices
- multimeters
That's why I'm so annoyed by manufacturers (hi Apple) reducing port counts and moving to USB-C only.
Apple //e
TRS-80 Model 100
Vt330 terminal
Plusdeck 2c cassette tape deck
Lego serial IR tower for RCX
Serial port for self-built cheap arduino clone (an NKC electronics freeduino serial) ( http://web.archive.org/web/20100131051737/http://mcukits.com... )
cable to connect TI calculators (I have a USB version of this too, though, so I don't need to use a USB-to-serial converter with my old serial cable based one.)
And that's just what comes to mind right now, I'm sure there are more.
Furthermore, that's just on laptops, since USB-A ports no longer fit in Apple's increasingly-thin designs. The iMac still has four USB-A ports, right now, today. The Mac mini still has two USB-A ports. The Mac Pro can have dozens of USB-A ports, if you want them.
Also, I don't care about Thunderbolt 3. I don't have a single peripheral that supports it. What I do have is tens of devices that use USB, mostly 2.0.
> USB-A ports no longer fit in Apple's increasingly-thin designs
That is not true for all designs. Apple would have no problem fitting USB-A ports in MacBook Pros.
If you have "tens of devices" you need to plug into your computer at once, then stop buying laptops or get a USB hub. This isn't complex. No laptop ever designed, in history, would allow you to plug in that much USB-A stuff at once without buying adapters or a hub, etc.
So stop bitching at Apple about irrelevant stuff that Apple isn't to blame for.
A lot of monitoring hardware runs on legacy protocols (ModBUS comes to mind) which are encapsulated over and over newer protocols. I'm sure most of these controllers also come in USB flavors. Moreover, USB 3.0 is full duplex and supports more power than 2.0 This means better communication in real time monitoring applications using more complex controllers/adapters.
So, it may not be a speed-first but connectivity-first motherboard which is optimized for many small data bursts over many USB ports. Three displays will also help showing the information obtained from these devices in a nice videowall-ish view.
Fun.
i'm not sure how many actual pcie lanes USB4 can transport? 2x?