NanochipMy main question here is: what generation of PCIe will USB4 v2 enable? Is it PCIe4.0? 5.0? And how many lanes of PCIe bandwidth? Finally, we've heard that with PAM3 signaling, existing TB4 (and USB4 40 Gbps?) cables will be able to take advantage of the new data speeds, no problem. But what about existing TB4 controllers...will a firmware update enable them to pump data at faster than 40 Gbps? I doubt this: my thought is TB4 controllers are connected to the PCIe root complex at PCIe 3.0... foreclosing any opportunity for a speed bump via a firmware update.

Below is a drawing of current TB4 solution and expected TB5 solution with PAM 3. It could give an answer about anticipated PCIe standard used both for USB4 2.0 and TB5. It seems that PCIe switch will be Gen5 x4. It will link to four Gen5 lanes on Intel's CPU or AMD's CPU for USB4 2.0 solution. A few years later such PCIe switch could wire to chipset once those get PCIe 5.0 capability. Or sooner, if the PCIe switch wires to eight Gen 4 lanes on the chipset.

PCIe switch chip needs to be Gen5 x4. There are several reasons for that. It gives an access to 128 Gbps of data speed. If symmetrical solution is implemented (easier one), this total bandwidth could be distributed into two ports running at Gen4 x4 each, so that each port gets up to 64 Gbps of PCIe data to any peripherals, which would largely be Gen4 for years. As Thunderbolt 5/USB4 2.0 port will have 80 Gbps bandwidth in total in one direction, this makes sense.

TB5 port (80 Gbps) expected capabilities:
1. one DP 2.0 80 Gbps display or two DP 2.0 40 Gbps displays (4K/120 10-bit RGB panels) - if two such displays are daisy-chained, bandwidth of one port is saturated with DP data only
2. PCIe Gen4 x4 up to 64 Gbps, if no other traffic consumes more than 16 Gbps, such as DP monitor and/or USB device. PCIe data available bandwidth will be reduced if high-end monitor is connected, as display data get priority
3. USB 20+ Gbps (xHCI 1.2 or newer controller)
4. PD 3.1 up to 240W (TI most likely)
5. Networking - more than 10 GbE?




In this light, USB4 2.0 chip/port might have slightly less stringent obligatory features than TB5, such as one mandatory display instead of two, etc. But yes, USB4 2.0 should use PCIe 5.0 x4 on CPU link and minimum PCIe 4.0 x4 on port PHY.

Existing cables will not be able to use 80 Gbps speeds, as new chips are needed in cables to recognize and transmit all those standards that will be implemented. The same applies for HDMI and DP cables. Cables designed for older standards cannot miraculously use most of new features. For example, Intel designed TB redriver JHL5040D to work with 40 Gbps cables. This chip implements DP 2.0 UHBR10 traffic for 40 Gbps connections and USB 3.2 Gen 2x2 for 20 Gbps traffic. Such redriver cannot operate with 80 Gbps traffic, as it is not capable of supporting DP UHBR20 traffic. Nothing is known about tunnelling capabilities of this redriver.

You are right. Firmware update cannot bake in hardware capabilities. New IP solutions are needed.

LycanwolfenUSB 4 great more USB that cannot do USB speeds. USB 2.0 was 60 mb/s and today we do have some USB drives that can finally get that speed. USB3 speed is 600 mb/s a sec well some drives now can get that but most drives maybe 30 mb/s Now USB 4 comming out. Give me a break. How long will it take before devices can actually do 2400 mb/s or 4800 mb/s or even 120 GB a sec. Nothing maybe in 40 years something will get that. I have some samsung bar USB 3.1 drives and the fastest transfer on those is maybe 250 mb/s not even USB 3.0 speeds.

Not sure what you're talking about, USB 3.0/3.1 Gen1 (5Gbps) can easily handle 450MB/sec with an external enclosure with a SATA3 SSD in it as long as UASP is enabled. USB 3.2 Gen2 (10Gbps) enclosures are capable of 950MB/sec....there's always a little overhead which doesn't allow for full throughput, but it's been really close, is the overhead loss what you're referring to? Because that's a fact of life for any I/O protocol...either way USB has been pretty close to the claimed bandwidth as long as you have the proper hardware to fully utilize it.

AssimilatorI think Thunderbolt is officially dead now. I don't see any advantage it has, or can have, over USB4 now. And that's a good thing for the industry because Thunderbolt is a royal PITA.

It's not dead at all. Almost every Intel and Apple laptop has one or more TB ports. I use TB3 Samsung monitor daily. It's connected to Asus ProArt B550 motherboard that has two TB4 ports and DP-in ports for routing video data from GPU. It works fine. I also use TB networking at 10 GbE with direct link between two PCs at home. Asrock and Asus enjoy installing TB chips on several motherboards.

AquinusThat is quite literally the difference between passive and active cables. Passive Thunderbolt 3 cables are super cheap, but they're also only a foot or so long. USB4 has the same issues when it comes to cable length. At some point it needs to become an active cable to handle the distance. When push comes to shove, active cables for TB3 or USB4 are essentially doing the same thing. I would not expect production costs when ramped up to the same capacity to be all that different. USB4 and TB3 really aren't all that different from an implementation perspective. It has more to do with licensing, not the underlying technology.

There's no Intel tax, which makes USB4 cheaper.
You can go up to almost 2 meters with a passive USB4 cable, but it has to be a quality cable.

Tek-CheckIt's not dead at all. Almost every Intel and Apple laptop has one or more TB ports. I use TB3 Samsung monitor daily. It's connected to Asus ProArt B550 motherboard that has two TB4 ports and DP-in ports for routing video data from GPU. It works fine. I also use TB networking at 10 GbE with direct link between two PCs at home. Asrock and Asus enjoy installing TB chips on several motherboards.

Intel won't let it die for that that matter.

MusselsI do like the concept of the assymetrical connections, if only they could negotiate in realtime - allowing external hard drives for example to swap around in real time depending if you're reading or writing

External hard drives? Do you mean HDD? Faster HDDs on the market are ~260 MB/s, which is much lower than one PCIe 3.0 lane uses when converting SATA signal from HDD. 40 Gbps cable can already serve many hard drives in external enclosures, and at top speed. Am I getting this right?

AquinusThat is quite literally the difference between passive and active cables. Passive Thunderbolt 3 cables are super cheap, but they're also only a foot or so long. USB4 has the same issues when it comes to cable length. At some point it needs to become an active cable to handle the distance. When push comes to shove, active cables for TB3 or USB4 are essentially doing the same thing. I would not expect production costs when ramped up to the same capacity to be all that different. USB4 and TB3 really aren't all that different from an implementation perspective. It has more to do with licensing, not the underlying technology.

Both with copper and FO cables can be completely passive but for long distances the hosts themselves need to boost the signal. This in turn puts the burden of higher power draw, heat and increased BOM on the end devices. But such is life, you can't have it all at the same time and at the same place.

LycanwolfenUSB 4 great more USB that cannot do USB speeds. USB 2.0 was 60 mb/s and today we do have some USB drives that can finally get that speed. USB3 speed is 600 mb/s a sec well some drives now can get that but most drives maybe 30 mb/s Now USB 4 comming out. Give me a break. How long will it take before devices can actually do 2400 mb/s or 4800 mb/s or even 120 GB a sec. Nothing maybe in 40 years something will get that. I have some samsung bar USB 3.1 drives and the fastest transfer on those is maybe 250 mb/s not even USB 3.0 speeds.

External NVMe Gen4 drive can do ~5 GB/s, which already saturates 40 Gbps connections and cannot operate at its full speed over PCIe 3.0 in Thunderbolt and USB4 solutions.

zlobbyI wonder what specs we'd get for the USB4.1 Gen 6 Rev.2 PD Alt-mode Cat9 Class 8 High Speed???
This has gone beyond anything sane... :banghead:

It's about more capable monitors and faster data transfer from external NVMe drives, USB4 devices and eGPUs. A leap towards 80 Gbps connection is necessary to accommodate new devices and tech.

TheLostSwedeThere's no Intel tax, which makes USB4 cheaper.
You can go up to almost 2 meters with a passive USB4 cable, but it has to be a quality cable.

Intel won't let it die for that that matter.

And it's easier to drive several protocols over single USB-C connector than install all other ports, apart from HDMI, but here adapters could be used too, such as new HDMI-USB-C 40 Gbps cable from Club.

TheLostSwedeYou can go up to almost 2 meters with a passive USB4 cable, but it has to be a quality cable.

USB4 and TB3 are almost the exact same thing if you exclude the whole royalty thing. I have never seen a passive TB3 cable longer than a few feet and I find it hard to believe that a USB4 cable capable of PCIe passthrough at TB3 speeds wouldn't have the same kind of limitation. The standards are simply too similar for there to be that much of a difference. Not to mention that Intel still exercises its right to require certification against the standard, so while there is no "tax" per say, it still costs businesses to go through that certification process which is undoubtedly not free.

Tek-CheckExternal NVMe Gen4 drive can do ~5 GB/s, which already saturates 40 Gbps connections and cannot operate at its full speed over PCIe 3.0 in Thunderbolt and USB4 solutions.


It's about more capable monitors and faster data transfer from external NVMe drives, USB4 devices and eGPUs. A leap towards 80 Gbps connection is necessary to accommodate new devices and tech.


And it's easier to drive several protocols over single USB-C connector than install all other ports, apart from HDMI, but here adapters could be used too, such as new HDMI-USB-C 40 Gbps cable from Club.

How are HDMI, TB, DP inadequate for monitors, so that USB will bring something on top of that?

eGPU? I'm yet to see someone making use of these.

AquinusUSB4 and TB3 are almost the exact same thing if you exclude the whole royalty thing. I have never seen a passive TB3 cable longer than a few feet and I find it hard to believe that a USB4 cable capable of PCIe passthrough at TB3 speeds wouldn't have the same kind of limitation. The standards are simply too similar for there to be that much of a difference. Not to mention that Intel still exercises its right to require certification against the standard, so while there is no "tax" per say, it still costs businesses to go through that certification process which is undoubtedly not free.

No, that only applies if Intel gets to tell you the information.
I suggest you read my article about USB4 if you haven't. The cable thing is explained in it among many other things. It's related to the loss budget, which was changed for USB4.
It might not apply for PCIe tunnelling though, so you could be correct there.
www.techpowerup.com/review/usb4-guide-info-technology-details/

TheLostSwedeNo, that only applies if Intel gets to tell you the information.
I suggest you read my article about USB4 if you haven't. The cable thing is explained in it among many other things. It's related to the loss budget, which was changed for USB4.
It might not apply for PCIe tunnelling though, so you could be correct there.
www.techpowerup.com/review/usb4-guide-info-technology-details/

I have a while back when you posted it. It's a good light read. For what it's worth, I am talking about the things that make USB4 and TB3 the same, which is support for the 40Gbps connection. That definitely has much more demanding cable requirements than 5, 10 and 20Gbps USB revisions, but if we talk about TB3 and its analogous support in the USB4 spec, then I think you'll find that you're almost always going to have super short cables or an active cable since really nothing changed with the signaling as TB3 (as I understand it,) is backwards compatible with USB4 and is how USB4 hits the 40Gbps mark, because it's basically TB3, just without the Intel tax and a handful of other small additions.

So yes, I agree in the sense that below 40Gbps that it has some flexibility, but if we're talking tit for tat between TB3 and USB4, I think you'll find very little difference in the implementation to achieve 40Gbps. That's really all I was getting at.

Tek-CheckIt's not dead at all. Almost every Intel and Apple laptop has one or more TB ports. I use TB3 Samsung monitor daily. It's connected to Asus ProArt B550 motherboard that has two TB4 ports and DP-in ports for routing video data from GPU. It works fine. I also use TB networking at 10 GbE with direct link between two PCs at home. Asrock and Asus enjoy installing TB chips on several motherboards.

The only reason they have Thunderbolt is because USB 3.x was slower than TB3 and didn't have DisplayPort alt mode. USB4 is now faster than TB4 and guarantees DP alt mode, so there is literally zero reason for OEMs to care about Thunderbolt any more; all it does is add cost and complexity to designs.

AquinusI have a while back when you posted it. It's a good light read. For what it's worth, I am talking about the things that make USB4 and TB3 the same, which is support for the 40Gbps connection. That definitely has much more demanding cable requirements than 5, 10 and 20Gbps USB revisions, but if we talk about TB3 and its analogous support in the USB4 spec, then I think you'll find that you're almost always going to have super short cables or an active cable since really nothing changed with the signaling as TB3 (as I understand it,) is backwards compatible with USB4 and is how USB4 hits the 40Gbps mark, because it's basically TB3, just without the Intel tax and a handful of other small additions.

So yes, I agree in the sense that below 40Gbps that it has some flexibility, but if we're talking tit for tat between TB3 and USB4, I think you'll find very little difference in the implementation to achieve 40Gbps. That's really all I was getting at.

Maxwell's and Shannon's (among countless others') works tell us that we can only pass so much information without losses over a wire with given characteristics. Unless somebody revolutionizes physics soon, nobody is running PCIe 'riser' cables for their GPU on the balcony.

Why do you think there are no 100Gbps ETH cables, let alone long ones?

TheLostSwedeIt's USB4, not USB 4.

Actually more like USB4 Version 2.1 ...
/facepalm

Prima.VeraActually more like USB4 Version 2.1 ...
/facepalm

I thought they were going to go for USB4 Version 2.0 3x1...

The real 4G mobile internet speed is 24 year old USB 1.1 speed!

AssimilatorUSB4 is now faster than TB4

USB4, TB3, and TB4 all have the same limits IIRC because they're all basically doing the same thing at the higher link speeds. TB4 is only different in so far as it has a higher minimum bandwidth requirement for the PCIe tunneling bit. Now, with this change, you could have more bandwidth in a given direction at the cost of bandwidth in the opposite direction. At the end of the day, you still have the same bandwidth, you're just allocating it differently. I wouldn't call it "faster" though. It could be faster for particular use cases.

AquinusUSB4 and TB3 are almost the exact same thing if you exclude the whole royalty thing. I have never seen a passive TB3 cable longer than a few feet and I find it hard to believe that a USB4 cable capable of PCIe passthrough at TB3 speeds wouldn't have the same kind of limitation. The standards are simply too similar for there to be that much of a difference. Not to mention that Intel still exercises its right to require certification against the standard, so while there is no "tax" per say, it still costs businesses to go through that certification process which is undoubtedly not free.

USB4 standard is more relaxed than TB4, and there are more optional features, so that OEMs could pick and choose what they want to support.

PCIe tunnelling is optional (although, and interestingly, Windows mandates all OEMs to include PCIe tunnelling on all USB4 ports. Well done Windows!), cables do not have to be 40 Gbps, but half of it, cables are not standardised as yet, one display less, lower charging requirement, lower minimum power for accessories, no mandatory networking, no mandatory certification for the ecosystem.

AssimilatorThe only reason they have Thunderbolt is because USB 3.x was slower than TB3 and didn't have DisplayPort alt mode. USB4 is now faster than TB4 and guarantees DP alt mode, so there is literally zero reason for OEMs to care about Thunderbolt any more; all it does is add cost and complexity to designs.

Not that fast. USB4 introduces confusion among consumers because we do not know exactly what we are getting with USB-C port. That's a problem. The reason for that is too many optional features and more vague package of features, which opens advertisement to deception, misinformation and abuse.

USB4 is not currently faster than TB4. Where did you get that information from? DP Alt Mode is currently limited to DP 1.4 until first AMD Rembrandt laptops come out with UHBR10 40 Gbps speed certification. OEMs that implement TB4, which is literally almost all of them with Intel CPUs in laptops, know that once TB4 port is installed, a defined package of features would reach consumer. With USB4, I'd imagine a lot of complaints from disgruntled consumers.

ReadlightThe real 4G mobile internet speed is 24 year old USB 1.1 speed!

With LTE 3CC you can get 650Mbps in DL. Granted, that's LTE-A but then what is 'real' LTE? Not quite sure what are you talking about?

zlobbyHow are HDMI, TB, DP inadequate for monitors, so that USB will bring something on top of that?

eGPU? I'm yet to see someone making use of these.

Those are not inadequate. It's just that USB-C port simplifies connectivity for audio-video, PCIe and USB data, and there is less reason to install all different ports.
Traditional DP ports have already disappeared from laptops. DP protocol goes now over USB-C port. Increasing number of GPUs also have USB-C port for DP 1.4 video data. USB4 type-C port should also work with TB devices.

The only one left is HDMI, as it has its own spec. Therefore, I would expect in most laptops USB-C, USB-A legacy, HDMI, LAN and perhaps SD card reader, but some OEMs are even reducing some of those ports.

AquinusSo yes, I agree in the sense that below 40Gbps that it has some flexibility, but if we're talking tit for tat between TB3 and USB4, I think you'll find very little difference in the implementation to achieve 40Gbps. That's really all I was getting at.

This remains to be seen and tested. Rembrandt laptops are the first ever devices to have USB4 support on die and Raphael and Raptor Lake CPUs do not have this controller, as far as I know. For desktops, separate USB4 chip is needed, such as the one from AsMedia 4242.

PC World tried to test USB4 on Rembrandt laptops. Multiple BIOS and chipset updates are needed for this to work properly. We still need a comprehensive review of USB4 in devices.

Tek-CheckThis remains to be seen and tested.

On paper, it shouldn't be all that different because how they operate at 40Gbps is essentially the same which would imply similar cable requirements. I don't think we need to test something like this out to make an educated guess on how it will behave. A non-Intel controller only adds variability in that equation, particularly if they don't go through the Intel certification process. When push comes to shove, signaling between TB3, TB4, and USB4 at 40Gbps is practically the same, so one can imply that signal integrity requirements would also be similar, if not the same.

zlobbyWhy do you think there are no 100Gbps ETH cables, let alone long ones?

There are FO cables of even higher speeds for specific environments. I have a few Cat 8 cables at home, rated for 40 Gbps, but currently no devices to use it. I use it for 10 GbE traffic. Consumer space for LAN cables moves very slowly... Even 10 GbE ports are rare on high-end routers and non-existent on standard routers and motherboards.

Tek-CheckThere are FO cables of even higher speeds for specific environments. I have a few Cat 8 cables at home, rated for 40 Gbps, but currently no devices to use it. I use it for 10 GbE traffic. Consumer space for LAN cables moves very slowly... Even 10 GbE ports are rare on high-end routers and non-existent on standard routers and motherboards.

Emphasis on 100GbE. The specs for copper end at 40GbE. And for a good reason. Max lenght for 40GbE over copper is 25m (or was it 27m?)

AquinusOn paper, it shouldn't be all that different because how they operate at 40Gbps is essentially the same which would imply similar cable requirements. I don't think we need to test something like this out to make an educated guess on how it will behave. A non-Intel controller only adds variability in that equation, particularly if they don't go through the Intel certification process. When push comes to shove, signaling between TB3, TB4, and USB4 at 40Gbps is practically the same, so one can imply that signal integrity requirements would also be similar, if not the same.

True. I am more concerned about features offered over USB4, as those are more relaxed than TB4. Microsoft was so annoyed by the USB4 spec that they ordered all OEMs to support PCIe tunnelling on every laptop that uses Windows.

Tek-CheckTrue. I am more concerned about features offered over USB4, as those are more relaxed than TB4. Microsoft was so annoyed by the USB4 spec that they ordered all OEMs to support PCIe tunnelling on every laptop that uses Windows.

Ehhh, you can say that about TB in general. TB4 and TB3 are practically the same thing at the end of the day. The only difference between them and USB4 is that you're going straight to the support for the 40Gbit link. All in all, I don't really care a whole lot about the features in USB4 as I use two TB3 displays. The only thing I really care about is that if I get a newer device with USB4, that the same displays will continue to work. I already get a lot of the lower USB revision speeds via the same ports and if I really were to need something faster than 5Gbps, I can always use one of the two additional TB3 ports I have for a TB device. To me, USB4 is just the glue to ties Thunderbolt and USB together, but TB already supported a lot of this, so it's not like a whole lot is new for current users of TB3.