The USB-IF Announces New Logos, Kills Off SuperSpeed Branding

[LabRat 891]

Not gonna quote all of 'ya, but I wanted to say:
Sometimes, I adore being wrong.
Your corrections to my presumptions on USB4/PCIe tunneling brought great news.
Thanks folks

 

[Ruru]

Totally a mess. I lost count how many times USB3.0 has been rebranded.

 

[TheLostSwede]

I meant as founding or current members of the group.
Anyway, since Intel has released GPUs with DisplayPort 2.0 at 40 Gbps, do you have any hint about DP 2.0 monitors? Even TFT is not able to gather any meaningful information about it.

It sounds bizarre that nobody publishes anything substantial about it. Even since VESA's rep had an interview with Tom's Hardware in Janaury 2021!, he said "end of 2021". It's end of 2022 and zero news. I am confused.

The only news I have heard about it was when AMD officially certified DP 2.0 port on Rembrandt laptop in San Jose during display symposium. There was a mention of "reference monitor" with MediaTek DP 2.0 chip inside.

It all sounds secretive, hush, hush, unclear, muddy. Are there any announcements in the pipeline that you are aware of? Do you guys plan an article about it? Perhaps a follow up interview with the same VESA rep Craig W.?

Also, there are no leaks whatsoever. Are any DP 2.0 monitors going to be announced alongside RDNA3 launch? It's usually then and during CES that we get some info.

Sorry, can't help there, not my expertise and I don't have any contacts at companies that makes those kind of products.

 

[Chrispy_]

Totally a mess. I lost count how many times USB3.0 has been rebranded.

After the first screwup (where they added decimal points and then generations instead of just calling it USB4) I stopped counting because I didn't care.

I guess if you needed bleeding-edge data transfer rates you probably would have cared but the USB standard has always been at least an order of magnitude faster than the fastest devices you'd typically expect to connect over USB.

When 10Gbps USB was released in 2013, the fasted drives you could possibly get were SATA-based and it was rare outside of synthetic sequential testing that any of those drives could saturate the old 5Gbps bandwidth.

Now with 80Gbps USB you are unlikely to find any device on the market that can reliably use that bandwidth. PCIe Gen5 SSDs in an enclosure could theoretically do it, but the USB-to-PCIe bridge chip doesn't yet exist, and first-generations of it are likely to suck, like they always do. It's genuinely hard to find an external drive that can read at more than about 25Gbps and that's if you spend silly money on overpriced external drives.

Basically, as long as you're using the fastest port on your device with the latest device, you're generally going to be fine, whatever the hell they label it as; Real-world hardware always trails behind the current-gen platforms by enough that it's almost never an issue - and when it is an issue, you have to truly ask whether the limitation is the USB interface bandwidth, or if it's something more primitive like running out of pSLC cache, performance loss in the PCIe-to-USB controller, lack of HMB since the USB controller can't transparently pass through PCIe/NVMe etc. There are countless reasons why a USB device is slower than something connected directly to the CPU via PCIe and until USB becomes a physical hot-swappable PCIe connector, it will always be subject to extra steps that slow it down.

 

[Ruru]

After the first screwup (where they added decimal points and then generations instead of just calling it USB4) I stopped counting because I didn't care.

I guess if you needed bleeding-edge data transfer rates you probably would have cared but the USB standard has always been at least an order of magnitude faster than the fastest devices you'd typically expect to connect over USB.

When 10Gbps USB was released in 2013, the fasted drives you could possibly get were SATA-based and it was rare outside of synthetic sequential testing that any of those drives could saturate the old 5Gbps bandwidth.

Now with 80Gbps USB you are unlikely to find any device on the market that can reliably use that bandwidth. PCIe Gen5 SSDs in an enclosure could theoretically do it, but the USB-to-PCIe bridge chip doesn't yet exist, and first-generations of it are likely to suck, like they always do. It's genuinely hard to find an external drive that can read at more than about 25Gbps and that's if you spend silly money on overpriced external drives.

Basically, as long as you're using the fastest port on your device with the latest device, you're generally going to be fine, whatever the hell they label it as. real-world hardware trails behind the current-gen platforms by enough that it's almost never an issue - and when it is an issue, you have to truly ask whether the limitation is the truly the USB interface bandwidth, or something more primitive like running out of pSLC cache, performance loss in the PCIe-to-USB controller, lack of HMB since the USB controller can't transparently pass through PCIe/NVMe etc.

Good points there and I agree more or less. Personally the 5Gbps has been enough for me (and I don't even have type-C devices), so when it's just USB3, it's fine for me.

 

[trsttte]

After the first screwup (where they added decimal points and then generations instead of just calling it USB4) I stopped counting because I didn't care.

I guess if you needed bleeding-edge data transfer rates you probably would have cared but the USB standard has always been at least an order of magnitude faster than the fastest devices you'd typically expect to connect over USB.

When 10Gbps USB was released in 2013, the fasted drives you could possibly get were SATA-based and it was rare outside of synthetic sequential testing that any of those drives could saturate the old 5Gbps bandwidth.

Now with 80Gbps USB you are unlikely to find any device on the market that can reliably use that bandwidth. PCIe Gen5 SSDs in an enclosure could theoretically do it, but the USB-to-PCIe bridge chip doesn't yet exist, and first-generations of it are likely to suck, like they always do. It's genuinely hard to find an external drive that can read at more than about 25Gbps and that's if you spend silly money on overpriced external drives.

Basically, as long as you're using the fastest port on your device with the latest device, you're generally going to be fine, whatever the hell they label it as; Real-world hardware always trails behind the current-gen platforms by enough that it's almost never an issue - and when it is an issue, you have to truly ask whether the limitation is the USB interface bandwidth, or if it's something more primitive like running out of pSLC cache, performance loss in the PCIe-to-USB controller, lack of HMB since the USB controller can't transparently pass through PCIe/NVMe etc. There are countless reasons why a USB device is slower than something connected directly to the CPU via PCIe and until USB becomes a physical hot-swappable PCIe connector, it will always be subject to extra steps that slow it down.

The exception to that is docking, where you need the correct features and bandwith starts to matter really quickly. The way I see it Thunderbolt 4 will remain the preferable solution simply because you know what you get with the important bits all mandatory to meet certification.

 

[Solaris17]

What if your colourblind?

Matt green = 5 Gbps, light green = 10 Gbps, dark green = 20 Gbps, army green = 40 Gbps

 

[Mussels]

After the first screwup (where they added decimal points and then generations instead of just calling it USB4) I stopped counting because I didn't care.

I guess if you needed bleeding-edge data transfer rates you probably would have cared but the USB standard has always been at least an order of magnitude faster than the fastest devices you'd typically expect to connect over USB.

When 10Gbps USB was released in 2013, the fasted drives you could possibly get were SATA-based and it was rare outside of synthetic sequential testing that any of those drives could saturate the old 5Gbps bandwidth.

Now with 80Gbps USB you are unlikely to find any device on the market that can reliably use that bandwidth. PCIe Gen5 SSDs in an enclosure could theoretically do it, but the USB-to-PCIe bridge chip doesn't yet exist, and first-generations of it are likely to suck, like they always do. It's genuinely hard to find an external drive that can read at more than about 25Gbps and that's if you spend silly money on overpriced external drives.

Basically, as long as you're using the fastest port on your device with the latest device, you're generally going to be fine, whatever the hell they label it as; Real-world hardware always trails behind the current-gen platforms by enough that it's almost never an issue - and when it is an issue, you have to truly ask whether the limitation is the USB interface bandwidth, or if it's something more primitive like running out of pSLC cache, performance loss in the PCIe-to-USB controller, lack of HMB since the USB controller can't transparently pass through PCIe/NVMe etc. There are countless reasons why a USB device is slower than something connected directly to the CPU via PCIe and until USB becomes a physical hot-swappable PCIe connector, it will always be subject to extra steps that slow it down.

Ever heard of hubs?

the 10GB/s of 80Gb, is going to split nicely with a USB-C hub that works for a 10Gb ethernet adaptor, an NVME drive and some regualr 5Gb ports

 

[Chrispy_]

Ever heard of hubs?

the 10GB/s of 80Gb, is going to split nicely with a USB-C hub that works for a 10Gb ethernet adaptor, an NVME drive and some regualr 5Gb ports

Good point, but the IO panel on the back of the motherboard is all the hub I need. I'm a luddite and distrustful of the quality of most hubs; Based on empirical data a high proportion of the hubs I've used have been total garbage. In saying that, I've never sought out nice hubs, either.

The exception to that is docking, where you need the correct features and bandwith starts to matter really quickly. The way I see it Thunderbolt 4 will remain the preferable solution simply because you know what you get with the important bits all mandatory to meet certification.

Yeah, though it only matters if you're docking an external high-res display over USB-C.
For the sake of one additional cable I'd rather just have a display out via HDMI (and the MUX chip) rather than going over Thunderbolt/USB-C
Single-cable convenience will be a deciding factor for some people, I'm sure....

 

[Lianna]

Good point, but the IO panel on the back of the motherboard is all the hub I need. I'm a luddite and distrustful of the quality of most hubs; Based on empirical data a high proportion of the hubs I've used have been total garbage. In saying that, I've never sought out nice hubs, either.


Yeah, though it only matters if you're docking an external high-res display over USB-C.
For the sake of one additional cable I'd rather just have a display out via HDMI (and the MUX chip) rather than going over Thunderbolt/USB-C
Single-cable convenience will be a deciding factor for some people, I'm sure....

I guess you use desktop only. Direct connections are the best if you have enough of them.
Hubs or docks (docking stations) may be necessary if your primary system is a laptop. I agree, most hubs are garbage and have misleading marketing, finding one with 10Gbps USB, HDMI 2.0 and/or even GbE takes time.
Direct HDMI connection - yes, unless OEM saved a few pennies on LSPCON and your card supports, say, HDMI 2.0 but your (laptop) port is 1.4 only; that's a common case. It makes USB-C to HDMI 2.0 adapter really useful.

 

[Mussels]

Good point, but the IO panel on the back of the motherboard is all the hub I need. I'm a luddite and distrustful of the quality of most hubs; Based on empirical data a high proportion of the hubs I've used have been total garbage. In saying that, I've never sought out nice hubs, either.


Yeah, though it only matters if you're docking an external high-res display over USB-C.
For the sake of one additional cable I'd rather just have a display out via HDMI (and the MUX chip) rather than going over Thunderbolt/USB-C
Single-cable convenience will be a deciding factor for some people, I'm sure....

My setup is ~3 meters to the side of the tower itself now. I run a USB extension to a 7 port hub.

5Gb is enough for a bajillion USB 2.0 peripherals, 80Gb would let me run my monitor off that connection too, which is just insane.

 

[Prima.Vera]

So what is the difference between USB3 40Gbps and USB4 40Gbps ?? cable wise also??

 

[trsttte]

So what is the difference between USB3 40Gbps and USB4 40Gbps ?? cable wise also??

Hmm... it's that one of those doesn't exist

If you meant USB3 20gbps and USB4 20gbps, I was going to go into the rabbit hole but looked at this table from wikipedia and decided it was best to just give up (it seems to be about the same with a slightly different encoding - packages with half the size - but I don't know).

 

[Nanochip]

Did you read my USB4 primer article?

USB4 is Coming - Everything You Need to Know

USB4 is the next evolution of the USB standard, but it's quite different from its predecessors as it brings with it support for DisplayPort connectivity, as well as optional PCI-Express and Thunderbolt 3 support. It tries to fix some technical issues, but adds a few of its own, but at least the...

www.techpowerup.com

I guess not, based on your comments.
If it says 20 Gbps, that's all you need to know. Tunneling will work, regardless of what you're sending in the tunnel, that's why tunnelling is so great.


See article above. I visited ASMedia and talked to them about this. One port can do 40 Gbps while the other can do 20 Gbps simultaneously.

Reading is fundamental. I said the label 20 Gbps is ambiguous. The protocol used is still relevant. Users still need to know if it’s usb3.2 port or a usb4 port.

case in point I have a usb 3.2 gen 2x2 20port on my motherboard. A 20 Gbps port. But alpine ridge devices connected to this port do not enumerate or connect. Why? Because the port doesn’t support pcie tunneling or tb3 signaling. And since alpine ridge doesn’t have a usb3 fallback mode it must be connected to a tb3 or usb4 port to connect. The ambiguous label “20 gbps” does nothing to tell me if my alpine ridge device will work with the port, or not.

 

[TheLostSwede]

Reading is fundamental. I said the label 20 Gbps is ambiguous. The protocol used is still relevant. Users still need to know if it’s usb3.2 port or a usb4 port.

case in point I have a usb 3.2 gen 2x2 20port on my motherboard. A 20 Gbps port. But alpine ridge devices connected to this port do not enumerate or connect. Why? Because the port doesn’t support pcie tunneling or tb3 signaling. And since alpine ridge doesn’t have a usb3 fallback mode it must be connected to a tb3 or usb4 port to connect. The ambiguous label “20 gbps” does nothing to tell me if my alpine ridge device will work with the port, or not.

Well, Thunderbolt isn't USB, so there's your problem right there. That's not the fault of USB, but Intel. It's also not the fault of the cable.
The fact that Intel is trying to sell Thunderbolt as USB is their marketing problem, not that of the USB standard.

 

[Mussels]

Reading is fundamental. I said the label 20 Gbps is ambiguous. The protocol used is still relevant. Users still need to know if it’s usb3.2 port or a usb4 port.

case in point I have a usb 3.2 gen 2x2 20port on my motherboard. A 20 Gbps port. But alpine ridge devices connected to this port do not enumerate or connect. Why? Because the port doesn’t support pcie tunneling or tb3 signaling. And since alpine ridge doesn’t have a usb3 fallback mode it must be connected to a tb3 or usb4 port to connect. The ambiguous label “20 gbps” does nothing to tell me if my alpine ridge device will work with the port, or not.

You raise a valid point, I cant find answers yet.
Wiki pages are stating that due to the tunneling some overheads are reduced so say an NVME SSD on 20Gb/s 4.0 should be still faster than 20Gb/s 3.x


This likely falls on the motherboard manufacturers and their shitty backplates with poor labelling, than anything else
(Although with how few USB-C ports we have, it'll likely be fairly obvious)

 

[Nanochip]

You raise a valid point, I cant find answers yet.
Wiki pages are stating that due to the tunneling some overheads are reduced so say an NVME SSD on 20Gb/s 4.0 should be still faster than 20Gb/s 3.x


This likely falls on the motherboard manufacturers and their shitty backplates with poor labelling, than anything else
(Although with how few USB-C ports we have, it'll likely be fairly obvious)

This is intel’s fault. they implemented alpine ridge in a weird way where if you connect your usb-c device to a usb-c port, the device may or may not work, depending on whether the port is a tb3/usb4 port or not. Confusing for some users!

Intel fixed the problem with titan ridge and later, but still…