Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I expect them to switch from pure digital (0/1) to multi-voltage (0/1/2/3) signaling – this requires tighter (adaptive) calibration but does not raise the signal frequencies while still doubling data throughput.

This also fits with existing passive cables still working (because multiple levels are not a problem there, but higher frequencies would be) but active cables needing to be replaced (because the existing active cables have digital signal amplifiers in each end which can't cope with multilevel signaling, so new active cables with multilevel-enabled amplifiers will be needed).

I have no actual inside information, but that would be the most feasible path to take and it would be consistent with the cable compatibility they have announced.
That's the same approach Intel seems to be considering for TB5, though more specifically they've been testing PAM-3, which is (0/1/2) (presented here, equivalently, as –1/0/+1):

 
I expect them to switch from pure digital (0/1) to multi-voltage (0/1/2/3) signaling – this requires tighter (adaptive) calibration but does not raise the signal frequencies while still doubling data throughput.

This also fits with existing passive cables still working (because multiple levels are not a problem there, but higher frequencies would be) but active cables needing to be replaced (because the existing active cables have digital signal amplifiers in each end which can't cope with multilevel signaling, so new active cables with multilevel-enabled amplifiers will be needed).

I have no actual inside information, but that would be the most feasible path to take and it would be consistent with the cable compatibility they have announced.

Could be PAM-3

" ... The second line confirms that this is ‘USB 80G is targeted to support the existing USB-C ecosystem’, which follows along that Intel is aiming to maintain the USB-C connector but double the effective bandwidth. ..."
https://www.anandtech.com/show/1685...bolt-5-photo-80-gbps-and-pam3-then-deletes-it

Whether it will be Thunderbolt 4.2 or TB v5 or something else might be up in the air. Perhaps the USB-IF is taking a slighlty different route ( that would be surprising). Throws even more 'wrinkles' in tracking all the names if TB and USB are on different major number versions for the same thing (TB with less optional stuff).

So a combination of frequency increase and 1.5 bits ( without going all the way to PAM-4 complexity in singnalling.)
The cost of these active control circuits in the cables is going to be a factor.

The current TBv4/USBv4 active cables don't have to be replaced if the speed is neogiated through a standard handshake protocol. If the cable effectively tells the port "I can only go to 20Gb/s" then there should not be any problem. The ports on both sides should dial back to at least 20Gb/s . ( if the one of the ports says even lower ... then everything goes backward in speed/bandwidth).


*IF* they were going to weave PAM-3/PAM-4 back to 20Gb/s zone then that would be messier. But because 80 is only in the new range only really have to deal with the new transceivers also being clean transmitters/receivers when not doing PAM-3/4 . The encoding should be done at the ports. The transceivers are just to make sure the signal doesn't get lost in the 'noise' of the cable. They should faithful reproduce their inputs.
 
That's the same approach Intel seems to be considering for TB5, though more specifically they've been testing PAM-3, which is (0/1/2) (presented here, equivalently, as –1/0/+1):


Thunderbolt 5 is extremely likely not decoupled from USB-IF at all. That's , in part , where USB-IF is getting it from.
Intel did a large section of the work to create the Type C connector for USB-IF also.

Thunderbolt 4 was largely USB 4 with less 'optional features'. The USB-IF standards usually allows implementors to cherry pick feature they would like to do or not. Which gets a bit of a hodge podge of compatiblity, but many with lower prices. Thunderbolt sets a higher floor and usually means system prices are higher, but more consistent across vendors.

Same thing is probably happening at "USB 4 version 2 " and "Thunderbolt 5". One is riddled with loopholes so can avoid higher component costs and the other has better interoperability , consistency at higher costs.
 
  • Like
Reactions: dabotsonline
would be great if my iphone had that transfer speed then my a7iv wouldn’t be difficult to import videos using sd card reader
 
Just making sure. This is USB 4 version 2.0 to support USB 3.2 ports? Ok got.

It’s getting to a point that I don’t even want to try to explain tech to my non-tech family members. It’s just, get this cable..
 
  • Like
Reactions: dabotsonline
Could just call it USB 5 and move away from this daft situation we are in with multiple things being called the same thing.

True. If it's called usb4 2.0 then they are shooting themselves in the foot, again, like before with usbc 3.0. It's time to streamline, clean up. Let it be called usb5 or thunderbolt5 or whatever but not usb4 2.0
 
  • Like
Reactions: dabotsonline
Also, I don't understand, 40gbps could work over passive cables. what? I always thought it required active electronics and timers to clean the signal.
 
Thunderbolt 5 is extremely likely not decoupled from USB-IF at all. That's , in part , where USB-IF is getting it from.
Intel did a large section of the work to create the Type C connector for USB-IF also.

Thunderbolt 4 was largely USB 4 with less 'optional features'. The USB-IF standards usually allows implementors to cherry pick feature they would like to do or not. Which gets a bit of a hodge podge of compatiblity, but many with lower prices. Thunderbolt sets a higher floor and usually means system prices are higher, but more consistent across vendors.

Same thing is probably happening at "USB 4 version 2 " and "Thunderbolt 5". One is riddled with loopholes so can avoid higher component costs and the other has better interoperability , consistency at higher costs.
Naively, it doesn't seem they can get 2 x the bandwidth out of a ternary system like PAM-3. It seems that, if you didn't want to increase the frequency, you'd need a quarternary system, like what you proposed (0/1/2/3). I'm not familiar with these higher-multiplicity systems, so I tried this simple calculation:

If the word length of a binary system is x bits, then that encodes for 2^x values. To determine the number of bits, y, that would give you the same no. of values in a ternary system, you'd just solve 3^y = 2^x, giving y = x * ln(2)/ln(3) = 0.63 x. I.e., you'd only need 63% as many bits, giving 1/0.63 = 1.6 times the bandwidth for the same frequency.

To get double the bandwidth, without increasing the frequency, it seems you'd need ternary, which would give you y = x/2.

So, at least according to this back-of-the-envelope calculation, it sounds like they are going to have to increase the frequency by a factor of 2/1.6 = 1.25.

I guess they decided PAM-3 plus a 25% frequency increase would be easier to implement than "PAM-4".
 
Last edited:
  • Like
Reactions: dabotsonline
Also, I don't understand, 40gbps could work over passive cables. what? I always thought it required active electronics and timers to clean the signal.
Thunderbolt and USB 4 have supported passive cables which pass through the signal as is. They also offer active cables that repeat the signal with active electronics. You will find there are cheaper passive cables that are under 1 m, and more expensive active cables about 2 m (and there’s now Apple’s new 3 m cable).
 
  • Like
Reactions: akbarali.ch
That is fast :)

For non-technical people this will be hell - as they will always be using the wrong type of USB-C cable
(As today - it will just be worse)
 
Sounds incredible but once again they are screwing up the USB naming 🤦‍♂️
According to Benson Leung: “This announcement is about a spec version bump, which is a technical document, not a branding document.

Here's what my guess is on what the branding will be for gear with the new speed:

USB4 80Gbps.”
 
Does this mean once we have a 5K2K monitor with 100Hz refresh rate, driving that bit rate may actually be possible?

I had assumed the bottle neck was always the data throughput of USB 4 / Thunderbolt 4 cables and ports?
 
I'm assuming here that your post was a response to mine. So I agree with your point when the identically looking cables (many of which have ICs or other electronics/intelligence at each end) have inherently different capabilities. However, if the universal cable is just "dumb" with effectively infinite bandwidth and a fixed DC power transmission capability then you can use that to connect systems together e.g. a computer to your TV or external hard drive. The electronics driving the cable at each end could conceivably handshake and figure out lowest common denominator protocols etc., but people wouldn't be loading up landfills with obsolete cables anymore. People generally know the capabilities of the hardware they're trying to interconnect and a generic cable would always appear to work well in such cases. A generic looking cable with different capabilities is another (frustrating) story altogether.
MY post wasn't specifically a response to yours. There's no shortage of complaints of one sort or other in this thread.

The notion of a universal, dumb cable???? Ain't no such thing. "Infinite bandwidth" is expensive, and the definition of "adequate bandwidth" for computing has increased with every passing decade. Do you feel competent to define how much bandwidth will be required 10 or 20 years from now? And just how much DC power transmission capability do you need? Should a smartphone or wireless earbud charger cable come equipped to carry as much power as a desktop computer or display requires? For all those concerned about wasting resources... Those over-built "universal" cables would use a whole lot more copper and plastic, too.

I have a long history in pro audio, video, and computing, with a bit of household electrical work on my days off. Fabricating cables of one sort or other has been a big part of it all. For the most part, higher data speeds demand more expensive cable fabrication techniques, such as twisted pairs with a high number of twists per foot of running length. The difference in construction and cost between a traditional Cat 5 Ethernet cable (10/100 mbps) and a Cat 8 (25/40 gbps) is substantial. For higher data rates the individual twisted pairs are themselves twisted together in carefully-engineered manners. Will pairs be individually shielded to reduce cross-talk? The higher the data rate the more likely that will be necessary, but shielding not only adds bulk but capacitance, which tends to reduce the maximum practical length of the cable. For power you need large current-carrying capacity, which is bulky (typically done with larger-guage conductors, but in the USB-C power cables it's a multiplicity of light-guage wires). Fabrication of these many-paired-and-conductored cables is not cheap, so manufacturers prefer to use simpler, cheaper cables for less-demanding applications (lower data rates, non-power delivery, etc.)... and consumers prefer to pay lower prices, too. Building to the highest common denominator generally fails all economics tests.

You want a dumb cable? Classic example is basic, Thomas Edison-style 110 VAC mains power delivery. Even there, manufacturers are much happier when they can use two-wire cables (hot and neutral) rather than three-wire (hot, neutral, ground). They're also not going to use the 12-guage conductors used in a 20-amp circuit to connect a lamp with a single, 125-watt-rated incandescent lamp socket (18-guage).

One size fits all is a pipe dream.
 
Dare I say, it makes me miss SCSI...
Humor aside, yes and no. I miss the clarity of “this device requires a higher transfer rate so use X.”

If you have the right cable, USB-C enabled devices are way easier than flipping dip switches to make sure a device has the right SCSI ID.

The difficulty is that “if.” A single connector is easier for users but had the unintended impact of making the cable market hell. Even some host devices and peripherals can be hell as you try to figure out what you are buying.
 
Should multiply by 10 like Ethernet: 10 Mb > 100 Mb > 1 Gb (1000 Mb) > 10 Gb
And then the messed it up by retrospectively adding 2.Gbps, 5 Gbps and 25 Gbps iirc. I think 2.5 was a cheap stopgap to support ax WAPs, and 25Gnps was for some DC stuff (oh yea a cheaper top of rack to server connection when fiber was to expansive and 10Gbps was not enough, I seam to remember tha packet pushers saying something about it being used mostly to converge FCoE and other traffic on one port so servers didn't need 4 ports (1 for FC and one for ethernet on redundant top of rack switches. But I'm straying well away from consumer here so I'll stop
 
  • Like
Reactions: dabotsonline
So will the 80Gb/s speed be labeled USB UltraSpeed, UltraSpeed+ or UltraSpeed++? :p

Seriously, happy to see even faster speeds. Will we now see a corresponding ThunderBolt 5 with 80Gb/s?
whu not just lable it with the actual bitrate an power delivery it supports ie 100W/20G or similar that would cut down on confusion
 
  • Like
Reactions: dabotsonline
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.