News USB 4 Debuts With Twice the Throughput and Thunderbolt 3 Support

[abryant]

The USB Promoter Group announced it is stepping forward to the USB 4 interface that enables up to 40 Gbps of throughput and supports Thunderbolt 3. Read more here.

PAUL ALCORN
@PaulAlcorn
Paul Alcorn is a Senior Editor for Tom's Hardware US. He writes news and reviews on CPUs, storage and enterprise hardware.

 

[TripleHeinz]

USB 4.0 Gen 4x4

 

[InvalidError]

USB 4.0 Gen 4x4

On USB Type-D connectors and cables since Type-C only has two high-speed lanes in each direction.

If the USB-IF followed its own naming convention with 3.1 where each generation doubles the per-lane rate, 40Gbps on two lanes would be 3.x gen3x2.

 

[hannibal]

My brain hurts... good to get more speed.... but make the naming more simple and more usefull!

 

[InvalidError]

My brain hurts... good to get more speed.... but make the naming more simple and more usefull!

With USB being so stupidly fragmented, I'm starting to wish we could just go back to specific-purpose ports and cables that work for their more limited intended purposes instead of having multiple protocol variants (USB, PCIe, ThunderBolt, DP, etc.) and multiple cable variants (USB2, USB3, USB3.1, USB3.2, USB4, etc.) sharing the same connectors but need to be correctly matched to work as intended.

 

[Evil Marshmallow]

So will Ryzen 3000 series CPU's have USB 4 Baked In also?

 

[InvalidError]

So will Ryzen 3000 series CPU's have USB 4 Baked In also?

Way too late for that. It takes at least a year for a newly finalized standard to get in production hardware. USB 4.0 isn't a final spec yet, may not make it into Ryzen 4000 and 600-series AMD chipsets next year.

 

[alextheblue]

USB 4.0 Gen 4x4

Don't forget the marketing name. UltraSpeed USB 40 Gbps.

 

[alextheblue]

With USB being so stupidly fragmented, I'm starting to wish we could just go back to specific-purpose ports and cables that work for their more limited intended purposes instead of having multiple protocol variants (USB, PCIe, ThunderBolt, DP, etc.) and multiple cable variants (USB2, USB3, USB3.1, USB3.2, USB4, etc.) sharing the same connectors but need to be correctly matched to work as intended.

I've been saying this for a while now, they have taken port consolidation a little too far. It would be fine if every USB-C port supported the same protocols... but they don't. That doesn't even touch Power Delivery, or adapters. Hurray.

 

[truerock]

So HDMI 2.1, DisplayPort 1.5, USB 4.0 and Thunderbolt 3 run at about 40Gb/s over USB-C.
So, can we just settle on a single protocol?
Having 4 protocols doing almost identical things is asinine.
And get rid of Ethernet and Bluetooth and WiFi while we are at it.

 

[Zaporro]

" certified USB Type-C cable... "

There we go, the "certified" cable bullshit all over again.

 

[SteelCity1981]

So HDMI 2.1, DisplayPort 1.5, USB 4.0 and Thunderbolt 3 run at about 40Gb/s over USB-C.
So, can we just settle on a single protocol?
Having 4 protocols doing almost identical things is asinine.
And get rid of Ethernet and Bluetooth and WiFi while we are at it.

I agree this whole thing is getting ridiculous. If it's hard enough for the enthusiast to follow, all these protocols, just imagine how hard it is for the avg consumer to follow all these protocols now.... There really should be a universal standard in place for stuff like this...

 

[InvalidError]

There really should be a universal standard in place for stuff like this...

Or just quit trying to cram 50 different things down a single connector. A single standard that does everything means a whole lot of added complexity to things that don't need it. Most plain wired keyboards today still use a 8051/80186-style micro-controller from ~30 years ago updated ~20 years ago to add USB1.x support. Puting an OmniBus micro-controller in those keyboards and similar devices would increase their power draw by 100+ times, while increasing the cost of the controllers themselves by at least 10X before you add bus certification fees, standard association membership fees, whatever applicable license and royalty fees, etc.

One bus to rule them out wouldn't be practical.

 

[joevt1]

It will probably be USB 4.0 gen 3 x2:

• USB 4.0 is the spec.
• gen 3 is the speed per line (20 Gbps)
• x2 is the line count for each direction (two lines for receive and two lines for transmit - 4 lines total which requires a USB Type C cable, as Type A only has one line for each direction - not including the USB 2.0 lines).

USB 4.0 will probably include the previous specs like USB 3.2, so it will have:

• USB 4.0 gen 1 x1 (5 Gbps 8b/10b)
• USB 4.0 gen 1 x2 (5 Gbps x2 8b/10b)
• USB 4.0 gen 2 x1 (10 Gbps 128b/132b)
• USB 4.0 gen 2 x2 (10 Gbps x2 128b/132b)
• USB 4.0 gen 3 x1 (20 Gbps? ??) - I am guessing this one.
• USB 4.0 gen 3 x2 (20 Gbps? x2 ??)

USB-C uses separate lines for DisplayPort and USB signals. The x2 modes don't allow simultaneous DisplayPort alt mode. The x1 modes only allow 2 lane DisplayPort alt mode (instead of 4 lane which allows only USB 2.0).

Thunderbolt transmits bits at 10.3125 Gbps or 20.625 Gbps (64b/66b) using 2 lines in each direction. PCIe and DisplayPort traffic are transmitted as packets on the same lines. While the Thunderbolt 3 controller takes PCIe 3.0 x4 and DisplayPort 1.2 x4x2 as inputs, only approximately 22 Gbps of the 40 Gbps total can be PCIe traffic (for some unknown reason). That is reduced by DisplayPort traffic that exceeds 18 Gbps. Thunderbolt reduces DisplayPort traffic by removing horizontal and vertical blanking stuffing symbols and recreating them in the peripheral device's Thunderbolt controller. I am not sure, but I think Thunderbolt can use x1 in the case where there is a problem negotiating x2 over a bad cable?

Since Thunderbolt 3 only allows 22 Gbps of PCIe traffic, USB 4.0 gen 3 x2 has the potential of being significantly faster than Thunderbolt 3.

 

[littleleo]

Or just quit trying to cram 50 different things down a single connector. A single standard that does everything means a whole lot of added complexity to things that don't need it. Most plain wired keyboards today still use a 8051/80186-style micro-controller from ~30 years ago updated ~20 years ago to add USB1.x support. Puting an OmniBus micro-controller in those keyboards and similar devices would increase their power draw by 100+ times, while increasing the cost of the controllers themselves by at least 10X before you add bus certification fees, standard association membership fees, whatever applicable license and royalty fees, etc.

One bus to rule them out wouldn't be practical.

"One BUS to rule them all and in the darkness bind them!" Sounds like a familar movie refrain. I don't think it's that bad but then I remember when USB 1st came out and we called it the Useless Serial Bus because nothing used it, my how times have changed.

 

[InvalidError]

I don't think it's that bad

Wait a few more years when you'll have to sacrifice a USB4 port to plug in a keyboard... all of the potential of a fully fledged USB4/TB3/DP2/etc. port to run a peripheral that only uses the USB1.0 (1.5Mbps) part of the spec!