News USB 4's 80 Gbps Spec Released Alongside New Logos

[Admin]

USB 4 with 80 Gbps is now official, along with new logos for the standard.

USB 4's 80 Gbps Spec Released Alongside New Logos : Read more

 

[spongiemaster]

Are there any USB 4 devices yet? And we're already on USB 4 V2.0. Is there any standard with a more confusing naming scheme than USB?

 

[RichardtST]

So, like 4 different logos for the new generation of USB 4. That's nice. Between all these versions, the HDMI versions, and the DP versions, I've got enough cables to run an undersea connection to French Polynesia, where you will find me trying to memorize all the various combinations. Seriously though, this is getting ridiculous. It has to stop. Someone please find a permanent solution!

 

[Colif]

Are there any USB 4 devices yet? And we're already on USB 4 V2.0. Is there any standard with a more confusing naming scheme than USB?

So far all I have found is a few cables and an nvme enclosure https://www.orico.cc/us/product/detail/7328.html

Expecting lots of fakes though.

 

[JamesJones44]

So, like 4 different logos for the new generation of USB 4. That's nice. Between all these versions, the HDMI versions, and the DP versions, I've got enough cables to run an undersea connection to French Polynesia, where you will find me trying to memorize all the various combinations. Seriously though, this is getting ridiculous. It has to stop. Someone please find a permanent solution!

The issue is power and speed vs cost. If they force everyone to make one cable capable of 80 Gbps and 240 watts with active management then everyone is going to get upset that a 6 inch cable costs $15. The only ways to make it stop is break the laws of physics, come up with a room temperature super conductor on the cheap or remove features from USB.

 

[dehjomz]

So where is the first USB 80 Gbps controller chip? Maple Ridge as we know maxes out at 40 Gbps, and on the latest NVMs, it doesn't properly support hotplugging of certain older Alpine Ridge devices (i.e., JHL6240).

I'm looking forward to hopefully a new Intel thunderbolt 5 controller or an USB4 80 Gbps controller from AsMedia MediaTek or some other entity. With full hotplugging support of older devices Full speed ahead! Seems that Intel z790 /Raptor Lake are still using the 'older' 40 Gbps technology, and we'll have to wait until at least MTL next fall for 80 Gbps on motherboards.

 

[InvalidError]

The only ways to make it stop is break the laws of physics, come up with a room temperature super conductor on the cheap or remove features from USB.

The USB spec is becoming far too bloated. I'm very much in favor of a whole new connector dedicated to 20+Gbps high-speed/60+W high-power devices.

 

[Kamen Rider Blade]

The USB spec is becoming far too bloated. I'm very much in favor of a whole new connector dedicated to 20+Gbps high-speed/60+W high-power devices.

Or we can just Update USB 4.X to support Type-A & Type-B plugs and update the lonely single Full Duplex SuperSpeed Channel to support the latest protocols & speeds.

No need to reinvent the wheel, we can just use a existing plug that just needs a micro controller chip updates on both ends to the new specification.

Then you can have a SuperSpeed Type A port capable of 20 Gbps & 40 Gbps @ Full Duplex speeds.

If you want Assymetric Speeds, you can have 40 Gbps & 80 Gbps one direction with USB 2.0 speeds covering the opposite direction for Temporary SpeedBoosts.

And USB Type-A can handle up to 60 watts according to official USB PD specs.

So charging wise, you're good.

Also SuperSpeed branding should be used as the label for all <Integer ###> Gbps bandwidths.

As long as your bandwidth is > 1.0 Gbps, it should be considered "SuperSpeed" class.

This way you don't conflict with legacy branding.

The issue is power and speed vs cost. If they force everyone to make one cable capable of 80 Gbps and 240 watts with active management then everyone is going to get upset that a 6 inch cable costs $15. The only ways to make it stop is break the laws of physics, come up with a room temperature super conductor on the cheap or remove features from USB.

That's the problem with USB Type-C.
Everybody that thinks "One Cable to Rule them All" is the solution.
Well you run into the issue where it's one cable that needs to do everything and that gets VERY expensive.
Does it do what you want, yes. Is it cheap, HELL NO!

That's why I think updating SuperSpeed USB Type-A/B ports to support the latest bandwidths / protocols is paramount.

A SuperSpeed Type-A/B port with the latest specs can support up to 40 Gbps @ Full Duplex communication speeds and 80 Gbps in a Assymetric configuration with USB 2.0 for the other direction.

It's a "Good Enough" solution.

And we need a cheaper new USB Nano-A port that offers Full Duplex HiSpeed USB 2.1 that fixes any of the flaws with the original USB 2.0 spec.
The Nano-A port would be square shaped and Fully reversible, and smaller tha a 3.5 mm Headphone jack in Diameter.

Imagine the new wired interconnects that can be made with such a slim & tiny cord based on USB Type-C's micro pins.

We also need a USB Mini-C connector that replaces the old SuperSpeed Micro-B, but is a Single Lane at Full Duplex.

Given that regular Type-C is Multi-Lane and with non-standard configurations, can have more than 2x SuperSpeed lanes, you should be able to see where I'm going with this.

Type-C will be the Large Bandwdith connector to the USB Root Port.

SuperSpeed Type-A/B ports or Mini-C will be the Medium Bandwidth for most devices.

HiSpeed Nano-A will offer smaller devices that need Low Bandwidth and tiny form factor along with a fully reversible setup.

It solves the problem of the "Ultra Expensive" port trying to rule everything. We don't need everybody buying "Ultra Big Gulps" sized Bandwidth connectors to solve everything.

We need finesse, the right size solution for every job.

 

[thisisaname]

Can this new cable/port be used in Europe or does the standard have to change to allow it?

 

[InvalidError]

And USB Type-A can handle up to 60 watts according to official USB PD specs.

So charging wise, you're good.

I wrote 60+W, so 60W would barely hit the minimum hypothetically guaranteed by USB4v2.0-gen4x2b. In a "one plug for everything" ecosystem, you need to think about things like bus-powered monitors with built-in high-powered hubs, in which case you need 20-70W to run the monitor itself + however much extra power any devices powered off of it may require which can be another 100+W when that includes a laptop, daisy-chained bus-powered monitors, external HDDs for bulk off/near-line storage and whatever else people may throw on there. You will also be losing a few watts in the cables themselves.

It may sound far-fetched today but the USB-IF did pitch USB4v2 at people using 4k240 or 10k60 monitors, so entirely USB-powered monitors may actually happen and once it does, 60W won't be anywhere near enough.

 

[Kamen Rider Blade]

I wrote 60+W, so 60W would barely hit the minimum hypothetically guaranteed by USB4v2.0-gen4x2b. In a "one plug for everything" ecosystem, you need to think about things like bus-powered monitors with built-in high-powered hubs, in which case you need 20-70W to run the monitor itself + however much extra power any devices powered off of it may require which can be another 100+W when that includes a laptop, daisy-chained bus-powered monitors, external HDDs for bulk off/near-line storage and whatever else people may throw on there. You will also be losing a few watts in the cables themselves.

It may sound far-fetched today but the USB-IF did pitch USB4v2 at people using 4k240 or 10k60 monitors, so entirely USB-powered monitors may actually happen and once it does, 60W won't be anywhere near enough.

So you're talking about those small portable monitors the size of LapTop monitors that lots of people like to use as a secondary monitor?

 

[Kamen Rider Blade]

Anyways, I still think it's a fundamental mistake to get rid of the "SuperSpeed" moniker.

That should be the indicator to the average user that your transmission rate is in the Gbps era if they're not reading the details too clearly.

Spoiler: My version of the SuperSpeed USB Transfer Speed Packaging Logo

 

[InvalidError]

So you're talking about those small portable monitors the size of LapTop monitors that lots of people like to use as a secondary monitor?

No, those mobile displays only use 5-10W and most don't provide downstream hub functionality. I'm talking 24+" monitors eventually getting USB-powered if USB4v2 become common enough that monitor and GPU manufacturers can be bothered to fully support it.

Putting type-C ports on monitors and GPUs has already been tried before: there was a short time where some AMD and Nvidia GPUs had type-C outputs and now that feature has gone mostly extinct. We'll see if the USB-IF's new attempt at "one port to rule them all" will be any more successful. It'll probably fail due to adding too much cost.

 

[Kamen Rider Blade]

No, those mobile displays only use 5-10W and most don't provide downstream hub functionality. I'm talking 24+" monitors eventually getting USB-powered if USB4v2 become common enough that monitor and GPU manufacturers can be bothered to fully support it.

Putting type-C ports on monitors and GPUs has already been tried before: there was a short time where some AMD and Nvidia GPUs had type-C outputs and now that feature has gone mostly extinct. We'll see if the USB-IF's new attempt at "one port to rule them all" will be any more successful. It'll probably fail due to adding too much cost.

I don't see Type-C ports catching on in DeskTop Discrete GPU's. Those add too much cost / complexity.

Mini DisplayPort would sooner return and populate those empty bracket space than use "Type-C".

It seems like "Type-C" is a mobile first solution and maybe a DeskTop MoBo Rear IO solution for iGPU's.

Most normal MoBo's would have at least 1x Full-Size HDMI & DP port.

But they can route the two extra display slots optionally on the DeskTop MoBo iGPU through the Type-C port if the end user wants them.

 

[kkratr]

USB4v2.0 ....can someone explain to me why they didn't call it usb5? Why do they keep doing the same mistakes over and over again?

 

[InvalidError]

USB4v2.0 ....can someone explain to me why they didn't call it usb5? Why do they keep doing the same mistakes over and over again?

Because too many execs are morons not worthy of their grossly excessive paychecks and marketing has far too much leadership influence in Big Tech.

 

[InvalidError]

I don't see Type-C ports catching on in DeskTop Discrete GPU's. Those add too much cost / complexity.

And where does that complexity come from? Its bloated spec with a dozen alt-modes, power delivery ranging from 500mA 3V through 48V 5A, four separate cable specs depending on power delivery current and voltage (old cables for legacy USB that only go up to 500-700ma, cables with the "trick resistor" for up to 2A, 20V5A chipped cables and 48V5A chipped cables), six cable specs depending on speed (USB2, 3.x-gen1, 3.x-gen2, 3.x-gen2x2, USB4 and active USB4v2), etc. not counting alt-mode cable variants.

Imagine how much simpler a high-power, high-speed spec that does 48V up to 5A as baseline and runs a lean encapsulation layer for hot-plug PCI for data with no alt-modes whatsoever would be. Only one power spec, only one cable spec (fully wired for high-speed, high-power), only one mode, only one protocol, no legacy luggage apart from what would come from using PCI as the foundation.

 

[Kamen Rider Blade]

Because too many execs are morons not worthy of their grossly excessive paychecks and marketing has far too much leadership influence in Big Tech.

QFT!

And where does that complexity come from? Its bloated spec with a dozen alt-modes, power delivery ranging from 500mA 3V through 48V 5A, four separate cable specs depending on power delivery current and voltage (old cables for legacy USB that only go up to 500-700ma, cables with the "trick resistor" for up to 2A, 20V5A chipped cables and 48V5A chipped cables), six cable specs depending on speed (USB2, 3.x-gen1, 3.x-gen2, 3.x-gen2x2, USB4 and active USB4v2), etc. not counting alt-mode cable variants.

You answered the question for me, it's overly complex, doesn't add enough to a Discrete GPU / Video Card to justify the time / expense / effort.
It's not like implementing the Type-C port isn't without it's troubles.

I'd rather see more Mini DisplayPort 2.1 eat up the port space on the back of the PCIe slot bracket than a Type-C port.

Realistically, Display Out via Type-C port (Doesn't matter if it's HDMI or DP protocol) is a Mobile Device FIRST solution, anybody else that uses it is a bonus.

Imagine how much simpler a high-power, high-speed spec that does 48V up to 5A as baseline and runs a lean encapsulation layer for hot-plug PCI for data with no alt-modes whatsoever would be. Only one power spec, only one cable spec (fully wired for high-speed, high-power), only one mode, only one protocol, no legacy luggage apart from what would come from using PCI as the foundation.

I can see a external PCIe connector as a dedicated solution.

Spoiler: Proposed Reversible External OCuLink connector

The proposed external OCuLink connector was going to be slimmer thatn USB Type-A ports and a bit wider by a few mm.
I don't know if PCI-SIG & Amphenol ever got done with the spec and finished mass producing it.

So far, I haven't seen anything, but this proposal happened right before COVID-19 Pandemic hit.

So alot of things went crazy around the world.

As for 48v, 5A; good luck getting PCI-SIG to approve.
At best, I can see them using the existing internal PCIe bus power standards/specifications, but moved to external ports as the power spec.

It's the least amount of work / friction for them.

BTW, that proposed connector handles up to PCIe 4.0/5.0 x4 lane width.

All the internal wiring tech is based on internal OCuLink for Enterprise, so it's a proven connector.
Just ruggidized for external use.

 

[InvalidError]

As for 48v, 5A; good luck getting PCI-SIG to approve.

The PCI-SIG takes its marching orders from the server and datacenter market and from what I read, the PCI 5.0 CEM spec includes provisions for up to 600W of 48V HPWR. So the PCI-SIG is apparently already rocking the 48V boat.

 

[Kamen Rider Blade]

The PCI-SIG takes its marching orders from the server and datacenter market and from what I read, the PCI 5.0 CEM spec includes provisions for up to 600W of 48V HPWR. So the PCI-SIG is apparently already rocking the 48V boat.

Cool Beans that 48V HPWR exists, but would they ever consider merging that with OCuLink?

That's the kicker.

OCuLink is based off of existing PCIe spec, and that doesn't have provision for 48V.

 

[InvalidError]

OCuLink is based off of existing PCIe spec, and that doesn't have provision for 48V.

At a glance, OCuLink doesn't even have 12V, only two 3.3V and two 5V tiny pins, which basically means barely enough power to run active cables if necessary. OCuLink appears to be a data-only standard. If you want integrated power and data, you need something else.

 

[Kamen Rider Blade]

At a glance, OCuLink doesn't even have 12V, only two 3.3V and two 5V tiny pins, which basically means barely enough power to run active cables if necessary. OCuLink appears to be a data-only standard. If you want integrated power and data, you need something else.

Exactly, & I hope it gets standardized.

We could use a true External PCIe connector for the masses.

 

[InvalidError]

Exactly, & I hope it gets standardized.

We could use a true External PCIe connector for the masses.

One that is actually capable of powering stuff.

OCuLink provides so little bus power that even M.2 adapters require external power. I can't imagine very many people being thrilled about needing two cables and an extra power adapter or using two ports on their laptop/PC to access their external NVMe SSD, which would likely be the most common consumer use-case than external GPUs if this went mainstream at a price commensurate with it being little more than dumb PCIe pass-through.