Hacker News Reader

Comment by Neywiny | Hacker News Reader

I actually like the 3.2 naming. Gen is speed, "by" is width. It puts it very roughly on par with PCIe's naming which nobody complains about. I just don't like that USB 3, USB 3.1, and USB 3.2 are the same things. And that sales people don't seem to understand that saying a chip supports 3.1 or 3.2 tells me it's anywhere from 5-20gbps which isn't ideal.

View on HN

PCI-E has had the same standard since its inception: 1.0, 2.0, 3.0, etc. USB has changed multiple times and has remained confusing for the vast majority of people. What was 3.0 is now not 3.0. Even 3.1 has changed. There is no reason to use this naming convention they currently have but for some reason they stick with it..

The problem with USB, is the revisions are all encompasing. So USB 2.0 includes the 1.5 Mbps and 12.0 Mbps modes from usb 1.x as well as the new 480 Mbps mode, so you could have a USB 2.0 device that only did 12 Mbps (high speed!). It might actually be your old usb 1.1 device with a new label.

PCI-e requires hosts and devices to be backwards compatible, but the interface speed is a required part of the spec. Nobody makes a PCI-E 2.0 device that only works with PCI-E 1.0 encodings/speeds, or anyway, it wouldn't be acceptable.

PCIe also had things like "1.1", "2.1" and "3.1" - that fixed issues and added functionality - but there wasn't the same crossover between "feature sets and spec revisions" and "speeds" we see in USB today.

Manufacturers of mainstream consumer motherboards never used 1.1, 2.1, etc. for PCI-E though. What is 4.0 on the spec sheet will be 4.0 to the buyer. My old 2016 motherboard has a slew of 3.0 labelled USB ports that are now not 3.0, hence the conundrum. It just doesn't make sense why they changed established naming conventions. Is this something that causes me sleepless nights? Not in the least. But it's still an annoyance for consumers and even advanced users as detailed in that latest Geerling video et al.

1.1 was very much commonly used in consumer marketing, to the level where there's many instances today of people referring to pcie1.x speeds as "1.1". And I'm pretty sure I've seen 2.1 in consumer marketing contexts. But you're right I didn't know 3.1 existed until I looked it up :p

But USB 3.0 is pretty much the only "speed" that hasn't changed - it always required the extra connectors for 5Gbps from the start - but no more. What about those ports is now not "3.0"?

Possibly they stick with it because it's usable (ish) and it was driving everyone up the wall when they'd change it?

And not only the sales people. Windows doesn't report anywhere what your motherboard is capable of, and even if you connect with a device it will not tell you the speed it agreed on.

Out of curiosity, what do you use the higher 20gbps transfer speeds for? Video production?

I use USB-C displays, but they run in DP Alt mode. I don't have many (any?) storage devices that can max out a 20gbps connection, and usually don't exceed 5gbps

This goes back to another point I've historically made which is that except for storage devices, pretty much nothing supports those speeds. I think there are some USB adapters that don't use alt mode and that can have some advantages on some hosts but usually that's a disadvantage.

USB interface chips are, as far as I've seen, a Cypress/Infineon FX3 or a bit more rare FTDI FT600/FT601. I even talked with the FTDI guys at s conference and they said nobody's asking for higher than 5gbps. Infineon just recently, after I think 10+ years, came out with 10 and 20gbps chips. But only for receive. Seems to be for cameras mainly. So surprisingly yes, video production.

But I want it for other reasons professionally. For example, if you look at the signalhound (which uses the fx3) series of products, they often cap out at 40 Msamples/sec for USB. This is a classic 5gbps limit. To compete with the big boys they need 250 MHz if not more. That's 8 gbps before protocol overhead. It doesn't help that USB is extremely dependent on host compute capability to keep throughput up but assuming your PC is up to the task, 20 gbps could interface some serious data to the real world.

Besides storage devices, i.e. external SSDs, which are very frequently used and they need a USB port as fast as possible, the other frequent application that needs the fastest USB ports is the use of USB Ethernet interfaces.

Also eGPU. I have a tiny NUC-size system with decent internal GPU and a (physically much larger) game system with a slower CPU that idles at only a bit under twice the maximum power of the NUC. It would be handy to be able to just plug in an eGPU when needed. The power and cooling requirements of fancy GPUs are so much higher than that of CPUs that large cases designed around the CPU don't make much sense. Even the pysical stability of a large GPU in an ATX style case is not ideal.

> Out of curiosity, what do you use the higher 20gbps transfer speeds for?

Images, videos, movies, file transfer/backup. 50 megapixel RAW images from a DSLR that can capture up to 20 images a second get big. My daughter is a much better volleyball player than I am sports photographer, so I have to spray-and-pray to capture those high-speed hits at the net.

Transferring a few hundred such photos via a card reader was so glacially slow it was worth adding a 20Gbps USB3.2 2x2 port to my home server (Ryzen 5600x) via a dedicated PCI-E card. The USB3 ports on a good enthusiast-class mobo for that generation (only 4 yrs ago) max out at 5Gbps (theoretical). I would have added a 40Gbps Thunderbolt port instead but then I'd have to take a hit on the top speed of my second NVMe drive due to sharing PCI-E lanes.

While the increasing deployment of true USB4 ports is wonderful, it's not quite a panacea. Just because a port is labeled USB4 doesn't mean you necessarily get 40Gbps performance. USB bandwidth is shared across multiple ports via internal hubs and then the PCI-E lanes the hub is connected to might be shared with other peripherals (GFX, NVME, I/O cards). And different USB ports have different trade-offs depending on how they're internally connected, which isn't always documented well by mobo makers.

Sadly, in consumer systems the lack of PCI-E bandwidth can still be an issue if you want your expensive GPU to run maximally fast and have multiple fast NVMe drives. You have to spec your system carefully, get the latest generation hardware or pay 3-4x for HEDT/Enterprise chipset motherboards. Getting even 10s or 100s of gigs of data in or out of a PC reasonably quickly and conveniently has always been a bottleneck that's only getting somewhat better quite recently.

Any external NVMe SSD from the last 7-8 years easily saturates a 20 Gbps connection, because already from that time the NVMe SSDs were able to saturate a 32 Gb/s PCIe 3.0 4-lane connection.

For at least 7-8 years I have been using USB external enclosures for M.2 Key-M NVMe SSDs, which always saturated whatever kind of USB port they were connected to, i.e. 5/10/20 Gb/s.

I do not remember when I have last used a SATA SSD, which is slower than 10 and 20 Gb/s USB, but I think that this was about a decade ago.

Reply on news.ycombinator.com