How Many USBs Can You Plug In At Once?

Get 3DEXPERIENCE SOLIDWORKS for Makers (20% Off $99USD/year!) at lmg.gg/3DSW or learn more at lmg.gg/f03c8
Save 10% and Free Shipping at Ridge by using offer code LINUS at www.ridge.com/LINUS
How many USB devices can be plugged into a PC at one time? It's a reasonable question, but the answer is more complicated than you might expect. We took 127 USB devices and plugged them in one by one until something broke. Then we took those same devices and reconfigured a few things to see if we can break it a different way. Just how broken do things get? Which limits get hit? How can you calculate YOUR maximum USB amount?
Discuss on the forum: linustechtips.com/topic/15253...
Buy a SABRENT 16 Port Powered USB 3.0 Hub: geni.us/nTjGG
Buy a UGREEN USB Card Reader Hub: geni.us/PoSn7IZ
Buy a StarTech 7 Port PCI Express USB 3.0 Card: geni.us/VZ8cC
Buy a StarTech 4 Port PCI Express USB 3.0 Card: geni.us/T868
Purchases made through some store links may provide some compensation to Linus Media Group.
► GET MERCH: lttstore.com
► GET EXCLUSIVE CONTENT ON FLOATPLANE: lmg.gg/lttfloatplane
► SPONSORS, AFFILIATES, AND PARTNERS: lmg.gg/partners
► EQUIPMENT WE USE TO FILM LTT: lmg.gg/LTTEquipment
► OUR WAN PODCAST GEAR: lmg.gg/wanset
FOLLOW US
---------------------------------------------------
Twitter: / linustech
Facebook: / linustech
Instagram: / linustech
TikTok: / linustech
Twitch: / linustech
MUSIC CREDIT
---------------------------------------------------
Intro: Laszlo - Supernova
Video Link: • [Electro] - Laszlo - S...
iTunes Download Link: itunes.apple.com/us/album/sup...
Artist Link: / laszlomusic
Outro: Approaching Nirvana - Sugar High
Video Link: • Sugar High - Approachi...
Listen on Spotify: spoti.fi/UxWkUw
Artist Link: / approachingnirvana
Intro animation by MBarek Abdelwassaa / mbarek_abdel
Monitor And Keyboard by vadimmihalkevich / CC BY 4.0 geni.us/PgGWp
Mechanical RGB Keyboard by BigBrotherECE / CC BY 4.0 geni.us/mj6pHk4
Mouse Gamer free Model By Oscar Creativo / CC BY 4.0 geni.us/Ps3XfE
CHAPTERS
---------------------------------------------------
0:00 Intro
1:06 Limits to push it to
2:21 THE HUB
5:31 Can we break THE HUB?
7:51 Big Hub Energy
9:18 Going Deep (into Bus Topology and DFPs)
10:13 AMD ruins our good time
16:00 Endpoints!
17:40 The USBinator!
19:23 4 ports is better than 7
20:36 USBinator powers on!
23:13 Is...Is this the limit?
24:00 Conclusion

[@SirSicCrusader]

The real question is how many USB devices can Linus drop at once?

[@loganiushere]

Agreed

[@kkshinichi]

Yes

[@Jonathanlaster20]

Take the answer from this video and add 12.

[@X2yt]

More than 4 at any given time.

[@sudoertor2009]

All of them

[@CanIHasThisName]

This actually makes me appreciate the USB standard. Not only the fact you can just plug anything into it and expect it to work 99,9% of the time without any additional input, but also the fact that everyone just agreed that this is something we're going to be using on just about everything.

[@Deses]

When they called it "universal" they really were onto something.

[@BritishBeachcomber]

But in the early days, USB was called *Plug and Pray* because it barely worked for anything but keyboards and mice.

[@billyeveryteen7328]

@@BritishBeachcomber It was a godsend for printers, especially in a house where you didn't need to have the printer plugged in all the time. I don't miss parallel ports at all. And most keyboards and mouses were still PS/2 in the earlier days, at least around 1999/2000. I feel like you had to go out of your way to find a USB keyboard or mouse, and there wasn't much point to doing so, since nearly all mobos still had the two necessary PS/2 ports.

[@Kiritomens]

@@billyeveryteen7328 yea parallel ports are something alright. I have barely used it ever outside of VM, and emulation purposes. It's still annoying to use even in that scenario.

[@Heizenberg32]

There may be one or two extra 9's in there... but it's impressive nonetheless!

[@burgerbait]

Imagine how many times Linus had to turn USB connectors to get them to actually plug in. Truly a dedicated man.

[@williamlafrance2514]

3 sides minimum per drive, must have taken a while...

[@arriswaasdorp4111]

There is a trick, on each side of the usb stick are 2 squares. If they are filled with plastic or something it’s the bottom side, if they are empty its the top side

[@Validifyed]

​@@arriswaasdorp4111that trick usually gets me from 5 attempts to just 3

[@gre18]

he got the ones that dont have a switch and give power either way

[@oldspice105]

Its Atleast 3 per plug. I've never done it in less..... It takes me 2 tries for usb c.

[@robertstockman2103]

I love how unscripted this episode is, and I do mean this 10/10. Please do more vids like this!

[@derkeksinator17]

The heated mouse got me.

[@the_undead]

It was scripted but then the writer happened to grab an AMD machine which reacted completely differently to this experiment than the Intel machine he did all his testing on. Most ltt videos are basically scripted down to the word. Some are partially scripted and partially off the cuff and then there's ones like this where it's scripted to some amount and then something unexpected happens and it's an adventure where the viewer knows about as much as Linus and the writer do about what's going to happen

[@Mxlted]

These limit finding videos are great for understanding our pcs more without having to go out and figure it out every single time.

[@thecanmanification]

Yeah, this was just a framing device for Linus to teach us about how the usb standard actually works

[@debbiebernhardt5406]

What is the limit of can(controlled area network) buses to usb adapter limit for modern computers?

[@TheSchokoxx]

well no shit

[@Motthias]

stackexchange is for this too.

[@PTNLemay]

Legit this is stuff you would cover in computer engineering classes.

[@caleballen1330]

Linus and his team doing something completely unreasonable but still somewhat educational is my favourite type of content

[@Acre00]

I subscribe exclusively for the completely unreasonable but somewhat educational content

[@danosdotnl]

*Clickfarm enters the chat*

[@larsradtke4097]

Not unreasonable. If I buy cheap Chinese 1time USB sticks, I need to check every stick for forgery with f3 - fight Flash Fraud. With Linux I use 12 port hubs and 50 sticks a limit. Depending on the PC I cannot use that many. With an i3 the USB3 windows device limit was 8! With a XEON 50 worked slowly. With my 4600h 50 no issue

[@3w3Ch00B]

This and the janky stuff, especially related to cooling

[@Stealthrellik]

This actually helped me out with figuring out how to deal with my valve index and many other USB devices fighting over USB resources lol

[@TBonerism]

I think the fact that you could plug in more than 128 devices is a relatively recent thing potentially starting at the 3000 series CPUs. I was running into USB issues on first gen Ryzen and I wasn't anywhere close to those numbers and I was having many of my peripherals turn off and then turn back on like they were reset on the addressing end. I added an internal USB controller card and that completely solved my issue.

[@Thewaterspirit57]

And that’s like…. One of the things they actually fixed lol. I couldn’t imagine having a USB device not working because of the parts I have. At least when USB was introduced, stuff still worked and they most likely shipped with a floppy disc for the driver

[@roots4x]

@@Thewaterspirit57 Damn USB is old. Physical media is so pointless and such a waste of resources.

[@Thewaterspirit57]

@@roots4x Dude no. Physical media exists to preserve things. If companies stopped making physical media including drivers, then there would be most lost media and also more older hardware thats useless cuz the driver was never archived or was deleted.

[@roots4x]

@@Thewaterspirit57 it’s almost assuredly easier to find drivers online or from your personal archives than from a floppy disk somewhere. You can easily dedicate a$5 USB drive to driver storage, or have a backup for another $5. It’s a huge waste of resources to ship a floppy drive to every customer in the age of the internet.

[@CsBence98]

Linus did a good job explaining the different limitations imposed by the USB spec, but there is one more he didn't talk about: time. More specifically, scheduling of the USB request blocks inside a USB frame that's sent out, at most, once every millisecond (in the case of USB 2.0 at least, don't know if they raised it in later versions, but I don't think so). He mentioned the four types of endpoints: ones for configuration (Control), rapid but short message transfer (Interrupt), large data (Bulk) and continuous transfer (Isochronous). What he didn't talk about is that you can't just mix'n'match them willy-nilly, because there's a limit of how much of the frame each of these types of endpoints IN TOTAL are allowed to occupy. For instance, IIRC the limit for Interrupt types is 10%. This endpoint type is used by keyboards, mice, game controllers etc., so if you only have these devices in your system, you will effectively only be able to use at most 10% of the full bandwidth of USB.

[@steelerfaninperu]

For those who don't know, USB Tree View is the OP GOAT program for this stuff. It shows you all connected devices and their level so that you can diagnose issues. If you've ever had the annoying reconnecting USB device that keeps chirping its connection sound on and off, this is the tool to use. It'll highlight funky devices just as it did in this video.

[@Codeaholic1]

lsusb enters the chat

[@jGRite]

Pin this wonderful comment above us.

[@user-fe7bo5mm1o]

​@@Codeaholic1sus

[@trebuchette633]

Oh, thank you! I was wondering what software they were using!

[@JacksonDevices]

​@@foofoodogWait, this isn't NetPositive!

[@nissanGold]

As a USB4 System Validation Engineer at Intel I found this video super interesting ,entertaining and educational. Great job Linus 👏

[@erebostd]

@@sophiophiletheory and reality. If you don’t know that’s there’s a huge difference you are not working in development, no matter what…

[@DarkStar14n]

@@sophiophile Spec is one thing, reality is another.

[@arnabbiswasalsodeep]

@@sophiophile Google what a validation engineer does, different than verification, emulation & testing. So there's 4 different job roles to ensure something is working.

[@CMDRSweeper]

The bad news is now... You will be plugging in a lot of USB devices at work from here on out to solve and fix issues as the limits must go up! :D

[@velo1337]

can you please share some knowledge you have gained over the years working on usb. like troubleshooting and stuff..

[@Luo-xue]

Some powered hubs also had this rather dangerous flaw in that it connects directly to the VBUS pin to the host, so it can backfeed the power to the computer, which can damages the controller and sometimes the entire computer if it was a laptops or all in ones. BigClive has a video about this that explains it pretty well.

[@raymondpavlov7624]

I’ve had issues with plugging in too many hubs rather than too many devices which is interesting. Having the same amount of devices plugged into two hubs seems to work better than having them plugged into 4 different hubs. Also some hubs work almost flawlessly whereas others just keep dropping out when I connect thumb drives to them but peripherals work fine… USB stability, even more than the instantaneous number of devices you can have plugged in, is a big issue and it would be worthwhile to investigate stability in a seperate video. I will say that having a seperate usb controller via pcie definitely improves stability! I’ve noticed that. But when I plug in multiple external hard drives I have significant problems with them just disconnecting intermittently even if they are on seperate controllers (this happens more often under high bandwidth usage but also seems to happen even if I’m just using the drive a little bit??). I’ve noticed this issue on different devices; my old Dell intel 7th gen i7 laptop and on my X99 5820k PC. 4 port powered hubs seem more stable than 7 port powered hubs and 7 is more stable than 13 port hubs; but I have another specific 8 port hub that isn’t sold anymore and for some reason is super stable. Maybe stability varies on vendor or newness of technology like Linus noticed on AMD vs Intel. I wonder if there’s a measurable difference on USB stability and % uptime on Windows vs Mac or from vendor to vendor

[@spongerobert]

There's definitely a difference between the chips in different brands of hubs. We used to run in the same issue doing backups to multiple external drives and we didn't try and copy something to all the external drives at the sane time but rather one after the other and we figured out that some powered hubs just are better than others and actually motherboards as well. We eventually also fixed it with a board with a bunch of PCIe slots and USB cards as opposed to USB hubs. Laptops on the other hand are all over the show and some laptops (modern and old) just really don't like having a bunch of USB devices at the same time.

[@ViViDxUK]

(22:46) Linus - "Oh, crap, something is depressed." Tanner - "Uh, sorry." Gotta love Tanner

[@ChrisLincoln]

Every so often I get re-amazed by USB. I started using computers near the end of the 80s on a DOS machine and the hot-swappable, plug and play nature of USB really changed the game.

[@lukethmpsn]

And the fact that almost every device uses the same connector is so damn useful.

[@axelxan]

​@@lukethmpsn For real. I still remember days when you bought new phone and almost every one of them had different power supply plug, you could buy those junky power supplies with full palette of plugs that usually broke after couple uses. Those were the times

[@strilight]

Not only data but power too. Not too long ago every small low voltage/amperage device used a little barrel jack connector or some proprietary thing for cell phones. If you were lucky you'd be able to find that one wall wart from some device you may or may not have had years ago. And even if you found one that matched the voltage and jack size you'd still have to squint and see if the jack polarity was correct. All small battery chargers used to require a wall outlet and low voltage hobby stuff like small wind/solar was a hassle because everybody did it a different way so wrangling your charge controllers was much more of a pain. Now it's all 5v over USB. If it's small and cheap it'll probably take USB. Even if USB charges slower than some of the older methods because of massively improved modern lithium batteries you'll still probably be charging it less. Things are so much easier now.

[@mestrinimaster3602]

​@@axelxanI still have boxes full of Nokia cables that came with the unlocking box 😅 It wasn't the only one making a different cable for each but man, it was the most

[@briant2100]

ya when usb first came i laughed... i was like ya thats gonna work

[@danagoyette7932]

At one of my previous jobs, we had devices with test harnesses, and the test harness PC would often run out of USB endpoints, so we had to run the harness machine with the Intel XHCI disabled, so the ports went to the two EHCI controllers instead. But then Intel removed the EHCI controllers from later chipsets. It seemed like Intel's limit was 64 or so, or maybe even less than that.

[@alexcrouse]

USB does a lot of polling, so when there is a ton of devices, each gets a chance to speak, which slows things down. Even if they are low load devices. You also may need to wait for everything to handshake and initialize after you connect them for a WHILE, then see better stuff working.

[@realbarries]

I don’t know what’s more impressive, the fact it (almost) handled 127 USB devices, or the fact they were able to find 127 devices around the office that hadn’t already been stolen (as we’ve seen is a common theme from $5000 tech upgrade videos.)

[@dataterminal]

It was a bring your keyboard and mouse to work day. With the amount of staff LTT has now, they had more than enough.

[@phaloxy]

If anyone is looking for the software Linus is using to list the devices plugged in, its called "USB Device Tree Viewer". Made by some german person called Uwe Sieber :)

[@mtt59]

thank you

[@metalale5529]

Savior

[@mrfoodarama]

This should be the most upvoted comment! 🤣

[@Mike90317]

Thank you!

[@ryancummings8273]

Thank you

[@RealGrouchy]

14:15 I'd appreciate a short video explaining why Windows thinks some thumb drives always have a "problem", even when scanning it reveals no issues.

[@cometcom3660]

such a useful video, im constantly in need of usb ports. i have a 10-port powered hub n a shit ton of 1 to 4 splitters just to keep everything running. it still amazes me how many ports it can handle until sth is malfunctioning.

[@TheHarryChanne1]

In the unix/linux world, handling interrupts for low speed i/o is a huge PITA. A bunch of USBs sounds like a great stress test.

[@arnabbiswasalsodeep]

I guess that kinda answers my question if linux would far better or not given server & supercomputer usage

[@TheHarryChanne1]

@@arnabbiswasalsodeep today's small systems have more oomph than the big systems from when Unix was created. The big difference was I/o channels. You don't need those now thanks to things like iscsi, but back in the day, you had to attach it to the system, e.g. DASD on block mux connectors for disks and tape drives.

[@konstakuosmanen]

​@@arnabbiswasalsodeepit is

[@U1TR4F0RCE]

@@arnabbiswasalsodeepit probably would be better for some stuff storage comes to mind where the lettering of drives in my experience isn’t a thing.

[@ToasterWithFur]

​@@U1TR4F0RCEBlock devices under Linux are still lettered (sda, sdb, ...) but when more then 26 a present another letter gets added (sdx, sdy, sdz, sdaa)

[@JazzCabbage00]

i stopped working in traditional IT departments way before it was "cool" this channel is a gem for old timey IT guys who miss the crazy IT department shenanigans we would do in the server room on off days. thanks for these!

[@tonyupham354]

A fantastic hub! Have had one for the last couple years and the speed is great and not one issue at all.... A great video!

[@keltoi30]

Just wow! This has got to be one of your best videos I have seen yet. I learned so much about usb.

[@VorpalGun]

Would love to have seen how Linux would have handled it. Would have been a good test to determine if it was a driver issue or not.

[@cdvideodump]

Linux gang gang gang

[@thisfeatureisbad]

I doubt that Linux would have any issues with that amount of devices.

[@kodi0223]

I would be interested too. At the very least it would probably need less CPU

[@autohmae]

I can just imagine a: watch 'lsusb | grep -v hub | wc -l' command running...

[@autohmae]

@@kodi0223 well, I think what they might have had issues with in Windows is every time you plugged something in it would start to load extra drivers, which itself causes a big spike in CPU usage, etc. Linux doesn't do this.

[@wisteela]

Very informative. Love that with Linus not knowing it was a heated mouse. 😁

[@countbowl]

Just did bring up on PCIe on a new platform and it works in a similar way when enumerating the bus. It seems like black magic when you first start, but it’s really just lots of simple operations repeated over the whole device tree. Was fun learning about it

[@matthewlozy1140]

If I'm not mistaken the CPU polls each device many times a second as well, so that's a good amount of CPU overhead. I wouldn't be surprised if there's significant delay between pollings and some devices not liking that.

[@ptamog]

All those gaming things demanding polling at unreasonable speeds

[@GeekProdigyGuy]

​​@@ptamogyeah >1kHz peripherals don't make any sense unless you're actually a pro and you've got a 360Hz+ monitor, consistent fps higher than the refresh rate, and optimized all the low hanging software settings. even going from 1kHz to 8kHz that's shaving only 0.875ms off input lag...

[@BCProgramming]

I believe the actual polling is done by the USB Host Controller, not the System CPU. When there is a result from the poll, when say a key is pressed or mouse movement occurs and so on, there could be CPU overhead servicing that, though.

[@Samuel.55]

​@GeekProdigyGuy they make plenty of sense. The higher the peripheral refresh rate, the more consistency between each individual mouse movement. They matter mostly for quick precise actions such as gaming.

[@tschuuuls486]

And if the polling can't happen in time the device probably times out and resets the connection. Which will make the situation worse lol

[@shei69]

Linus is in his side-missions era… Should’ve left the CEO position sooner 😂

[@TheMonDon1721]

A lot of these crazy videos are coming from tanner which is amazing

[@DaimyoD0]

So true. I love this kind of content. Like it might seem just fun and trivial but I actually feel like I learned useful stuff from this.

[@jswclips8459]

Wait what he's not CEO anymore?

[@robinrumpfkeil9546]

​@@jswclips8459yeah, he stepped back and let his former boss take his place, so he can completely focus on content creation again, without the whole headache that comes with being the CEO of a multi million dollar company

[@DaimyoD0]

@@jswclips8459 Nah he hired someone else as CEO so he can do the stuff he really likes in the company. He's like, Chief Vision Officer or something like that now?

[@Grundch]

Love this video (and type of video) so much, really pulls you with it and the education to entertainment ratio is, as the full product, Gold!

[@interlace84]

Another challenge connecting 100+ input devices is it'll only take one with one stuck button or key to cause issues, let alone all the razer-like devices auto-installing software next to their drivers it's still mindblowing you managed to get so far!

[@j.m.8895]

22:34 The delivery just cracked me up.

[@ethanblanch6517]

I love how it just turned into a Techquickie episode half way through, even thought they had gone onto a green screen until Linus grabbed a hub from the table lol

[@DizzyBusy]

Yes, why wasn't this a Techquickie episode?

[@harrasika]

​@@DizzyBusybecause it's not very quick

[@notenoughmonkeys]

So tempted to see how many USB Killers that janky setup could trigger at once. Maybe more something the next time ElectroBOOM is in the area!

[@olwynskye417]

This is going to become handy. I'm so used to using USB hubs with my laptop to connect external hard drives, controller, mouse/keyboard and other things. Good to know how they work and how far I can go with it until I get a proper desktop PC.

[@williamhall2386]

Finally… I need more info on this. Flight sim IO is acting weird.. I have researched USB tiers and PCIE lanes vs onboard chipsets/controllers and it’s just nuts

[@FlameSoulis]

I can confirm this absolutely with VR equipment, but also better stability with flight sim gear.

[@debbiebernhardt5406]

Mythbuster moment for ltt

[@pokepokepoke64]

Valve index dongles are weird too

[@lemonoxygen8846]

@theflightwrightsgaming906 I assume they mean IO == input output so all the controls like switches and throttles

[@williamhall2386]

@@lemonoxygen8846 lemon is correct. I was referencing the amount of available input/output devices

[@devrolo]

Please do more videos like these in the future. I love the idea of exploring the theoretical limits of different PC components.

[@michaelfitzgerald8727]

I've always wanted this video since I knew what a usb hub was, I never had the resources though. Thank you Linus, my mind can rest satisfied :)

[@alexyoung6418]

A lot of the 127 limitations with such devices have stuff to do with the protocol makers who thought it would be sufficient to only allocate the 7 least significant bits of a single byte in a fixed-length frame to represent the amount of a certain type of objects. The most significant bit is usually used as a flag to indicate some status, such as the validity of the value. Since the USB protocol has plenty of fixed-length data structures such as requests and descriptors, it can be a huge pain in the ass to work around these limits. Take the totaly amount of modifier keys simultaneously pressed on a keyboard for one. Plus, I'd be happy to see new USB HID protocols and DirectX versions surpass the 127-button limitation with a single USB game controller device. This again has something to do with a byte field in the report descriptor, however in this case the descriptor format is flexible. It doesn't seem to be crazy difficult to expand the protocol to accept longer frames while keeping downward compatibility. Older devices become legacy over time and support for them may eventually be dropped. Kudos to AMD with the expanded endpoint support, they could have bragged about it. I never bothered looking into how endpoints are handled on the controller side, information like this is probably off limits to me anyway. An endpoints is, as the name suggests, either the ultimate origin where data is collected before sending over USB, or the ultimate destination where that incoming data is received into for a devices to consume. It's just a USB-relevant way of calling a section of the RAM of a tiny computer chip. An endpoint comes in various sizes and it's often much easier to allocate a portion of RAM for a USB device - just write the code and hit compile - than managing the RAM space of a host controller chip for a whole bunch of different devices. This RAM has to be large enough to begin with, then this RAM would also have to be dynamically allocated and released as USB devices get plugged or unplugged. It's actually pretty amazing to see how robust the USB interface is, in far less crazy settings.

[@JordanProtsko]

Linus, question: Is it possible that some of the keyboards that you used could also act as a hub and have a USB port on it increasing the amount of devices connected? Just a thought experiment.

[@Hbk1998G]

Most keyboards I know have only one usb port which acts like an extensions for another port on your pc.

[@D3nn1s]

​@@Hbk1998Gno there are many that actually are usb hubs. Take the old apple keyboard as an example

[@MasterGeekMX]

there is one for the Raspberry Pi with two of them.

[@benhook1013]

@@Hbk1998GIf the keyboard is USB and it offers a USB port on it, it has to have an internal hub, else its own use as the keyboard would be all it could do.

[@Simon-tr9hv]

​@@benhook1013nope, most if not all of them only offer passthrough, you have 2 usb connector, one for keyboard, one pass through port on the keyboard

[@NoLongo]

I’m not sure if it’s part of Labs but it would be cool to see how much power each peripheral draws. In this video it would be interesting to see how much power was flowing to the hub from the motherboard.

[@DanKaschel]

When they were trying to figure out why everything was still working at the beginning I really wanted them to check the actual power draw.

[@tonnentonie2767]

There are 5$ USB passthrough devices that can do this, I use them to test cables as cables don't tell you what protocol they can run

[@donchaput8278]

I have a suspicion the one hub with the USB's may have been overheating and the chips crashing

[@DanKaschel]

@@tonnentonie2767 LTT has those as well as fancier ones, they just didn't pull them out

[@balaran4349]

Hey! Im a hub!

[@OtherK]

What type

[@styrelsefksu4463]

A good follow up would be to see what happens to windows 11 if you have more then the drive letter C - to Z and where the limit is and also if a Linux based system could beat that number of drives/ partitions since it is meant to be a server system and not a home PC test the same thing with windows server and also test the same thing on a mac to see if there is any limitations in what each of the OS is capable to handle or if it is in the USB specs itself that sets the limit

[@troyquigg4411]

Funny enough, I ran this experiment a number of years ago (2005 ish) running under Windows XP. Ultimately, I was able to get 212 devices (plus 20+ hubs) connected to a single Dell laptop. For the "127" device limit, I found that that was 127 devices per host controller. Most PCs have a couple of host controllers - this Dell with a docking station had 3 or 4. The other thing I found was that each device plugged in seemed to take about .5 percent of the CPU ... I suppose to maintain communication with it or some type of overhead with the driver. After I had around 200 devices online this .5 percent had finally added up to nearly 100% CPU utilization just sitting there. As I turned on the last 12 devices, the friendly "ding dong" sound Windows makes when a new USB device comes online slowed waaaay down. Device 212 literally had a 15 second long diiiiiiiing-doooooong sound. 213 never registered being powered up. I think the hardware could have theoretically handled more devices, but the CPU was maxed out. This was all for an actual legitimate purpose. I wasn't using thumb drives and it took some creative mix of powered and unpowered hubs to be able to power all the devices themselves... but, way past the "127" mark.

[@KainniaK]

I want to know the software limit for 95, 98, 2000, XP, 7, 8 and10. There must be a software limit. usb emulator software should be able to easily find it.

[@spike5499]

how and why did you get 212 devices 😭

[@troyquigg4411]

@@spike5499 haha - i was working for a science company and some researchers wanted to build a cheap lysimeter network (in this case, a lot of scales plugged into a PC.) 212 was the max number of scales I got (sorta) working on the laptop. It was stable and very usable with about 60-80 devices connected. Learned lots of tricks and hacks of USB to get it all working. 1 well powered (4+Ah) 7 port hub that could power a 4 port unpowered hub on each of its 7 and then a scale in the 28 available ports. A "branch" like that X4 will get you a little over 100 devices. For the 212 push, I had to use every port on the laptop plus docking station; including cutting a hole in the dock so I could still use the rear USBs that were normally replicated (but blocked) by the dock itself. I found that to keep it all stable, it worked best forcing all the USB to v1.0/1.1 ... disable 2.0 capabilities. It was much more forgiving to the number of hubs and long cable lengths this way.

[@toastylee1234]

Would've been interesting to see if behaviour was much different in linux, since it seemed to become a driver limitation at the end.

[@mrlithium69]

They would have needed wendell to know whats going on in the linux kernel

[@sannjogsos3009]

Does mac also have that limitations? Since they use thunderbolt, will speed or any other component be affected when that many devices are connected?

[@joee7452]

@@sannjogsos3009 The 127 endpoint limitation is currently built into USB because of the addressing. The USB addressing scheme is limited to 7 bits. No matter what system the controller is running on, each root controller can only address 127 endpoints to communicate to.

[@fromfin90]

the hub discussion at 5 minutes answered so many quaestions i had, thank you

[@Bamfhammer]

Linus stacking hubs in his had has been the best footage you have shot of the year. Absolutely perfect radial symmetry.

[@DarkRider2k3]

This is actually relevant to me since my flight sim setup uses multiple Arduino boards as well as other USB devices. I currently have 6x Arduino Mega, 2x Nano, Logitech rudder pedals, Honeycomb Yoke/Throttles, Wireless headset, Keyboard, Mouse. Sure I have 2 powered USB hubs, but considering I still have more planned upgrades knowing the USB limits is useful.

[@SupremeLordGeek]

Buying a PCIE USB card will probably allow more devices if you need it

[@normified]

​@@SupremeLordGeekthis

[@MacroAggressor]

Yeah, before I got into flight sim I would've just shrugged this off and said "but who cares", lol

[@DarkRider2k3]

@@SupremeLordGeek So far so good but I'm planning on at least 2 more Mega boards. Running a 5800x 3d so I'm hoping Zen 4 does me well with USB support lol. I have 2x Sabrent powered USB hubs similar to what they used.

[@jeremydeath]

I work on a computer that actually uses 128 (plus 16 hubs) USB 2.0 devices plugged into a single Linux machine. We really did use 128 USB thumb drives to test it out before assembling the full mess. It got quite a bit harder when USB 2.0 root hubs stopped becoming easily available. USB makes a great sideband protocol for bootloading and management.

[@kingmaxlol64]

For what scenario do you need that many devices?

[@drgncabe]

@@kingmaxlol64 I'm going to guess automation. I've dealt with pharmacy automation systems that use USB for communications have ran over 80 devices on a single system but never that many. I notice crappy controllers get flakey at around 50 or devices but solid controllers handled the 80 no problem (HPe workstations).

[@IsaqueFontinele]

@@kingmaxlol64 I will guess Phone Farming? Unfortunately they are common in certain places.

[@MikeWilliamsTheComputerGuy]

@linustechtips You can increase the number of USB thumbdrives by mounting the drives in empty NTFS folders instead of using drive letters. It would be REALLY cool to see this test done again with only thumbdrives!

[@wilh3lmmusic]

🐧 or just use GNU/Linux and don’t deal with this silly drive letter stuff

[@BBROPHOTO]

Those Startech hubs are pretty much standard for astronomy / telescope setups now. They're the only ones truly reliable as multiple cameras pushing a decent amount of data (plus many many USB accessories) are needed for telescope setups.

[@samtherat6]

“Something is depressed” “Sorry” Had me WHEEZING

[@SparkTubes]

FYI: A USB device will have a different ID when moved between ports - say like a serial adaptor moved from USB 1 to USB 2 would go from COM1 to COM2. Keep this in mind when using a USB device - Windows will remember each device plugged in, and which port it was on. Prepping USB devices for shipment can fill the registry with crap that's old, etc. Great video BTW. You are "The EXTREME!"

[@ArtisChronicles]

It's annoying, but helpful that it does this. One time I had a device malfunction on one port, yet it was fine on another. It stayed broken on that one port too.

[@Meshamu]

@@ArtisChronicles It was helpful that it stayed broken?

[@Work_Data_PV]

this was one of your most fun technical videos with actual application in normal people's daily life. thanks.

[@raymondpavlov7624]

There needs to be a sequel! How many thunderbolt devices can you plug in at once, or how many usb c devices as opposed to usb a

[@Infinitrium]

I remember reading some real old computer industry magazine in which they were talking about the release of Windows 95 OSR2 and the new USB 1.0 standard so a few industry people wanted to test USB 1.0's 127-device connection limit. The article says they had a bit of trouble finding 127 unique devices but eventually they did and they all worked when connected

[@jihadalsweed]

- "something really depressed" - "oh sorry" -" not that kind of depressed" I laughed so hard. Poor linus employees :(

[@STMUN]

this actually explained a common issue I run into, doesn't change how I fix it but it explains why my fix works

[@isaacmills1200]

This was such an interesting video, and I know a lot more about USB than I did before, but is there a reason that you guys couldn't conduct this test on a Linux machine? That would solve the drive letter problem, *and* possibly the driver limit problem you guys encountered. I'd love to see an appendix to this video about if you could reach 128 USB devices on a Linux machine.

[@me2olive]

It's also worth noting that some "active" USB extension cables (typically cables over 5 metres) actually use a hub chip, they're effectively a single-port hub according to the system. I think other kinds of usb "boosters" may exist, but presumably as hub chips are cheap that seems to be how active extension cables usually come.

[@bosstowndynamics5488]

From what I understand an active USB cable can be made with a very simple logic buffer circuit that adds pretty much no latency to the system. I suspect cables using hub chips is either reusing defective hub chips (eg only one downstream port works) or companies that already have tons of hub chips on hand for making hubs to the point that it's cheaper to use those than it is to also stock buffer chips.

[@last0titan]

As someone with questionably too many peripherals, I experienced problems with connections and disconnections constantly, first I thought it was a PSU problem, but further diving found it was more of the bus/controller. StarTech makes a PCIE card that comes with 4 usb slots each with their own individual controllers and I've not dealt with any issues since buying. Love seeing it being validated like this.

[@platinumjimjam]

This is easily one of the most fascinating videos from the entire channel.

[@djjay7777]

This is interesting and very educational. I always wondered what would happen with lots of usb devices plugged together, hence why I bought the Crosshair Hero VIII motherboard it has more than enough usb port for my devices.

[@jordilopezgomez1160]

The limit is not the number of devices but the number of endpoints per hub controller. With newer USB 3 you could connect up to 4096 endpoints. Intel xhce controllers are famous for not complying with that limit, furthermore different versions of the chip have different limits ranging from 64 to 192. AMD supports a lot more of endpoints.

[@KainniaK]

If intel figures out a way of saving some money so one of their theoretical limits is lower but they don't think anybody in the real world will complain about it, they will do it.

[@fireskydiver7]

Intel Sucks

[@FlickTakFlakAttack]

"nobody will use more than 64 devices at once" intel laughed while not seeing Linus Sebastian quickly approaching behind them

[@potatoes5829]

@@KainniaK I think its fine so long as they include it in their documentation (I don't think they do tho)

[@ABaumstumpf]

"Intel xhce controllers are famous for not complying with that limit" There is no non-compliance there - and it is xHC*I*.

[@aLotofLlamas]

For what it's worth, the videos where things go off the rails are always my favorite.

[@WASH3D]

I run a 3950x / 3070 system on an x570 MSI chipset and can confirm, the 2 4port StarTech hubs work like champs in my rig. I was able to connect 11 webcams for a live stream test, for anyone curious about bandwidth limits and ways to expand their live stream capabilities.

[@kuhrd]

Most modern computer motherboards will have more than one host controller. My laptop has 2 host controllers sharing the same chip package on board with the same bandwidth but my main desktop has 4 physically separate host controllers separately connected to the system 2 via the PCIe lanes on the motherboard from the CPU and the second two connected to a PCIe multiplexing switch chip on the motherboard. Having 2 or more USB host controllers is quite common on any standard motherboard these days. Windows however does have some limits if you add too many USB devices. Linux is much better at handling large numbers of devices since the OS handles them a little differently. It is very understandable that the system is using a ton of CPU resources to poll all of those ports, especially the devices like keyboards and mice that often try to maintain sub 1ms polling rates. This is also why you don't see hubs used in ethernet installations any more since the overhead of managing all of the ports can begin to overwhelm the devices and controllers all using the same string since each hub in a string doesn't have any memory to store and forward the data so the main host controller must poll each device one at a time and the more devices you add to the host controller the slower overall that network of devices gets. If multiple devices need to send data on a hub they have to send their data packets one at a time and the controller will switch between devices. In the world of USB if you have any devices that stream data they will need to be connected to or very close to the root host controller with very few if any other devices on the same host device. This is also why USB sound cards and video capture devices can have unacceptable delays in their input or output if they are not combined in the same device like delays between a webcam and microphone during a live stream that have to be adjusted for in software.

[@hkhan1989]

One of the best videos you've made. Theoretically understood all of the limitations, but to actually physically test this just looks like amazing fun! And the AMD tests are just damn bizarre and awesome!

[@benruss4130]

I love AMD's yolo approach, it has saved me time and effort in the past.

[@Offsecwrkr]

Me over here using 3,094 USB devices on my desktop with a custom intel chipset designed for my pc and my pc only.

[@kieranlee9610]

i don't think its an extremely powerful USB host i think its just more bugged than tanking because none of those actually worked AMD is still very buggy for no reason

[@benruss4130]

@@kieranlee9610 its more that the y will let you pile it on until you crash. It gives you a little bit beyond the official standard, whilst simultaneously becoming exponentially more buggy the further beyond the standard you go.

[@TheHammerGuy94]

basically, mad scientists

[@AwesomeGames56]

This is absolutely insane, and I love it! I don't have enough USB ports on my system so I've always been interested in getting a powered hub and that 16 port looks awfully interesting now.

[@D3nn1s]

What the hell do you need 16 more usb ports for lmao

[@MasonMouse]

@@D3nn1s I see someone has never made a USB RAID device Just Because™

[@Carlosdesignergrafico]

@@D3nn1s music production

[@jimstanley_49]

@@D3nn1s a flight simulator cockpit.

[@evilofalfa]

Ikr man

[@nasroml12]

This video is way beyond the other videos, congrats Linus this is a great video

[@raftmatrix]

I love how Linus was comfortable with everything but that one heated mouse

[@GrumpyCoffeeBean]

I'm so glad there's somebody like Linus around. This is exactly the same sort of stuff I thought would have been doing when I was younger but couldn't afford. Now that I'm older and still curious and could afford I don't need to because of your shenanigans!

[@ferybrand1990]

He really is like a Bob Ross or Mr. Rodgers for our little corner of the internet.

[@GrumpyCoffeeBean]

@@ferybrand1990 with a little but of Bill Nye mixed in.

[@morbus5726]

@@ferybrand1990 to bad he screws over tiny companies like billet labs, auctions their parts without consent, and refuses to admit his mistakes.

[@ferybrand1990]

@@morbus5726 brother if you think capitalism woes sway me youre barking up the wrong tree. We love in a society that allows exploitation, wanna cry about it? Fix it

[@morbus5726]

@@ferybrand1990 what? I'm just pointing out that he does some horrible things, countering your point that hes a "Bob Ross" or "Mr. Rodgers" type of guy. I'm not sure how you want me to fix it??

[@Reanimatedyt]

The answer: yes.

[@YNBD]

One thing to mention. You can have more "drives" than 26. Drive letters are not required. You can also have them unmounted... wich would defeat the point. OR you could assign drives to Mount Points instead of Drive Letters. as in, mount them as if they are folders.

[@guardianfkc]

Ok, what I've learned, hub on hub not ideal. Another very useful experiment, love this. Thank you guys!

[@techteam431]

This video is very relevant to my work. I got 20 port powered hubs that take 2 usb B 3.0 inputs. (basically two 10 ports hubs in one unit) I flash 40 galaxy phones at a time and it takes forever because of the bandwidth, pushing 7.5 GB to 40 devices. I have 3 usb PCI cards so basically 10 ports per controller. I would love to flash more at a time but wasn't sure where my bottleneck is

[@brad8122]

Why the f are you flashing that many phones? Spam bot?

[@GameTimeWhy]

Are you the reason so many bots exist?

[@dedtrox6424]

i never heard of something like this. why do you have to flash so many phones???

[@Apollo-Computers]

@@dedtrox6424 the guy that flashes the stolen phones...

[@benp439]

@@dedtrox6424 Probably the work phones for the company they work at. A lot of companies use them.

[@ClearLampOil]

2:21 *POV No nut November is over*

[@angryscottishidiot]

Thank you - On my music station occasionally my mouse would stop responding. Powered up my USB hub and its fixed :D Had a music keyboard with lots of led plugged into the hub too :D

[@shakthinandan7124]

I've never that windows running out of drive letters to assign was a thing until now truly a great video❤

[@DiscipleOfMan]

Great video and really cool to find out more about how controllers distribute bandwidth and power. The amount of USB ports on modern motherboards is insane, back in the day I remember having plenty of add in cards.

[@myaiua]

I loved the video for being so educational as many pointed out, and I also love Tanner's energy. Props to this man for the joke at 22:48. Although depression is not a joke, but an illness, being able to chuckle about it is what makes your life a bit brighter. Thank you, Tanner, this one made my evening.

[@Bimboms]

The only thing that depresses me are the number of USB ports on my motherboard. Never felt more inadequate

[@dillonh321]

Same I hope to see more of him in the future.

[@geevee9728]

He also has a mouse infestation 🐁

[@Meidri9292]

5:00 I was having this issue, so it occurred a question to me how many USB devices can you plug into your computer and this video pops up, a fairly new linustechtips video for a problem I just had what a coincidence.

[@jakehiggins5118]

"Something is depressed" "oh, sorry" that made me giggle more than it should have

[@dangingerich2559]

I remember back in the days before USB, we used serial ports, and a PC would likely have one or two, and the most I saw on one system, with a controller that cost more than my gaming PC at the time, had 8. The resources used by each port pretty much limited it to 8 ports per system. Having so many devices today is such a luxury.

[@Henrik_Holst]

I worked at a place in the late 90:ies where we had loads of more serial ports than that on a single pc, the trick is having the kind of cards that offloaded some of the processing to the card itself. It was used to run a modem pool, but I think that I remember it being the same number as you, aka 8 per card.

[@WackoMcGoose]

Yeah, I remember reading one of dad's old "how computers work" books, and it talked about SCSI daisy-chaining (the predecessor to doing the exact same with USB)... Apparently, you could make some real crazy stuff happen back in the day if you had enough devices, enough cables and ports, and zero fscks given.

[@skeennah1927]

serial and rs232, those were the days, not forgetting the parallel printer port.................

[@dangingerich2559]

@@skeennah1927 Ever have to deal with RS485?

[@skeennah1927]

@@dangingerich2559 nope, haven’t heard of that

[@archoniteX]

Along the same lines of this video, I would love to see the team investigate the USB 3.0 interference with 2.4GHZ devices (such as wireless USB dongle mice, & bluetooth even).

[@EvilDaveCanada]

A while ago, when I was playing around with using powered USB 3.0 Hubs. At the time I found that some USB Ports will allow the excess power from the hub BACK into the USB port. It wasn't noticed until I powered down the system. At that time, the CPU fan keep running and was still lit. It seemed that the power from the USB port was going back into the PCIe bus, which is where CPU fans & lights get their power from. To solve the issue, I made little pigtail cables with the DC +/- both using diodes and added a barrel jack to allow external power as well. For a power source, I use a device designed for allowing mass charging of cell phones with 32 USB-A 5v2a per port with no data lines connected. I use it to supply individual power to each of my HUBs& anything else that uses a USB power brick. So my KVM switch, LED lamp and my additional input devices all have power independent of the PC's PS. These type of pigtails were used back in the day when devices first came out and needed more than the max power that a USB port could supply. They added an additional USB cable that would just take the power from an additional USB port. I'm not sure if they had any diodes back than as I used anything that used this trick.

[@rob7xiv]

We reach on one Project a USB3 32(96) Limit, so we changed the USB Cable Extensions to USB2. A USB Cable 5m have a Hub/Device. And Some Hub devices with more then 7 ports are full chained, so hub1 last port go to hub2, hub2 last port go to hub3, etc., so you get fast to tree limit.

[@muadeeb]

Would love to see how linux tries to deal with this. You're not limited to 26 drives there.

[@realryleu]

lsblk would run for a solid minute

[@supergoofy123]

For USB storage devices : In windows you also have the drive letter limit, so you cannot add more than upto drive Z (I think there was a hack to be able to also use drive letters A and B which are for floppy disk drives). In linux though there is a possibility that it might work.

[@edwardecl]

There is no limit in Linux, you can have as many drives as your IO can handle.

[@regrettspaghettt]

Wouldn't linux also be limited in the same way because each drive is sda, sdb, sdc, and so on?

[@ZeroX252]

This is true only for drive letter assignments, windows allows attaching disks to empty folders just like Linux does, and can address a nearly infinite number of volumes this way.

[@ZeroX252]

@@regrettspaghettt No - Linux wraps around at sdz to sdaa, sdab etc, and then will move on to sdba after that until you reach sdzz at which point it will go to sdaaa and so on.

[@cameronwebster6866]

In windows you can mount drives to folders Unix style, using disk manager. In disk manager, right click on the drive, select "change drive letter and paths" remove the drive letter(optional) and in "add" you will see an option for "mount in the following empty NTFS folder"

[@Rerbun]

7:20 When using Raspberry Pis as a NAS with a USB HDD you also need to be careful with power draw. I had to buy a powered hub because my drive was not getting enough power to spin the disk and would keep shutting down every couple of seconds. I'm lucky the drive survived in the first place.

[@jonnivanzelatti4073]

From what I understand from the usb specification, the limit of 127 devices is per usb port, not per controller, but we have a limit of 5 hubs per port

[@CyberbrainPC]

I found this out a few years ago! Depends on the system! Also, some Unraid systems wont boot if you have so many USB drives in that it pushes the bootable USB off the boot device list.

[@Idiomatick]

Lol, that sounds like a fun headache. I recently changed motherboards and there are physical slots for 8drives but the silicone only supports 6, so when you turn it on it would randomly select 2 drives and not detect them. I figured it out quickly enough, but didn't notice this at time of purchase and needed to buy a raid card.

[@ErulianADRaghath]

...you could say...It got booted 😉

[@CyberbrainPC]

@@ErulianADRaghath lmao get out :-p

[@CyberbrainPC]

@@Idiomatick Yikes, that also sounds like a headache and bad design!

[@Idiomatick]

@@CyberbrainPC It was clearly written in the spec so i have only myself to blame..... though a better eufi diagnostic message would certainly be a welcome improvement. I tried switching cables before checking spec, so i guess it'd have saved me 10mins.

[@requiemriot8619]

Linus semi panicking about the hot mouse was pretty amusing

[@thetomoffel]

Now power them up all together and so how the computer handles this! Again awesome video!

[@ThePixelDreams]

What software is being used to investigate the USBs? Not sure if I missed that info, it looks super helpful!

[@dominikgrassler6034]

USB Device Tree Viewer by Uwe Sieber - It's mentioned a bit lower in the comments

[@Ntce]

15:00 this makes me feel better about my cable management

[@dedr4m]

Worth noting, since around 2010-ish (maybe before?) many motherboards started to up the amps to 2A and above, the reason the typical 5Vsb rail is above 2A (I've seen up to 5A) is in part for "USB charging support", so I'd not be surprised if the ports would handle 3A just fine and 3.75A before problems start. Then again, many boards still only allow 2A max using something like a "FETfuse USB protection IC" (if I recall the name correctly). I know the Dell E6400 had "USB charging support" for charging things like iPODs and early phone/smartphone types and defo could handle 2.5A each port.

[@JMBuLLeTCL]

"Billet Labs asked for the prototype back on June 28th. This would allow them to send it out to other media or use it for further development. LMG agreed at least twice to return the product. At LTX, LMG put the one-of-a-kind water block up for auction at its extra life auction event WITHOUT the permission of Billet Labs." "Billet told us that it is now stalled it no longer has its best prototype for continued development. It's also missing one of its GPUs." - How scummy can you get, LTT?

[@rythie]

I’d like to see how Linux and Mac handle this, since it mostly seems like operating system limits that are being hit.

[@gnomecs2]

Linus - Genuine idea for a video. Do a USB hub rundown of cheap or unbranded hubs compared to good ones. I recently bought a hub marked as USB 3.0 and when a USB 3.0 device was plugged into a USB 3.1 port via this hub it slowed the data transfer down to USB 1.1 .... It was clearly a con, but there are thousands of sellers trying to flog them with dodgy misleading descriptions on eBay and Amazon (who don't care). Call the scam products out and do a wonderful consumer awareness video all in one 👍

[@TheAtqthe30th]

To be honest powered cheap USB hubs kind of scare me. I've seen some actually send voltage down from the hub into the motherboard. You got to even watch out for some USB devices having malicious firmware which may have keyloggers on them. I just buy directly from trusted manufacturers for things like USB devices, SD cards, etc. Found that out after getting scammed with a fake Samsung micro SD card which looked legitimate until I started running some tests.