"Given that the Tiger Lake platform doesn't support USB 3.2 Gen 2x2, it is going to be interesting to watch how this plays out in the near future."
Glad to see that you caught that. I was pretty surprised when I saw the lack of USB3 dual-lane operation while perusing the Tiger Lake UP3 datasheet. I had really hoped that Thunderbolt 4 would support all of the headline features of both the USB 3.2 and USB4 specifications.
USB 3.x / SuperSpeed USB 5Gbps is Gen 1 signaling. USB 3.x / SuperSpeed USB 10Gbps is Gen 2 signaling. USB 3.2 / SuperSpeed USB 20Gbps is Gen 2 signaling plus dual-lane operation, a.k.a. Gen 2x2.
USB naming standards are such a mess, made shopping for a motherboard this past year an absolute pain with virtually every manufacturer inconsistently naming their I/Os.
If would be very simple if only they had made something like this:
USB 3: 5Gbit/s - Type A connector (known as USB 3.2 Gen 1×1) USB 4: 10Gbit/s - Type C connector (known as USB 3.2 Gen 2×1) USB 5: 20Gbit/s - Type C connector (known as USB 3.2 Gen 2×2) USB 6: 40Gbit/s - Type C connector (known as USB 4 Gen 3×2)
I'm sure that with hindsight the USB-IF might have elected to do things differently. However, from their standpoint, it probably seemed easier to add a couple new signaling modes as optional features in point releases and not have to roll new licensing or maintain additional specifications and associated testing / compliance programs.
Also, USB 3.1, which introduced the Gen 2 (10Gbps) PHY, was developed for Standard/Micro-A/B connectors and was finalized over a year prior to the Type-C Cable and Connector Specification. And USB4 hosts/devices only need to support Gen 2x2 (20Gbps) signaling, or up to Gen 2x1 (10Gbps) in USB3 mode.
Optional features can be beneficial to manufacturers and consumers alike, but they make the specification version number an unreliable indicator of device capabilities. And to a certain degree, the differences between the USB 3.x variants really only matter in edge cases. For years now, nobody has complained that USB 2.0 encompasses devices that might be capable of low-speed (1.5 Mbit/s), full-speed (12 Mbit/s), or high-speed (480 Mbit/s) signaling, despite the fact that they are separated by orders of magnitude. As long as the signaling rate is appropriate for the device, nobody cares. PC host ports are pretty much the only place where you might want to pay attention to the signaling capabilities, but a 10Gbps device only operating at 5Gbps is unlikely to impact most real-world workflows.
I am curious why you say 'a 10Gbps device only operating at 5Gbps is unlikely to impact most real-world workflows' ; For content creators on the go, for example, backing up work files from their desktop PC before venturing out to the field is a legitimate real-world workflow. A device operating at half the advertised speed is bound to double the waiting time of the user.
I do agree that all the version numbers have been designed from the viewpoint of what manufacturers need to do / support to obtain certification - However, based on prior experience, USB-IF must have recognized that the version numbers are bound to be marketed directly to consumers. Reading more into the USB4 standard and its implementation in Tiger Lake, I am strongly convinced that it is shaping out to be a disaster in terms of consumers understanding its capabilities (akin to what has been happening with Type-C over the last few years).
The only time you benefit from Gen 2 signaling is when your attached device is capable of throughput greater than 435 MB/s. That means multiple SATA drives in a hardware based striped RAID configuration or PCIe SSDs with a Gen 2 bridge. Although the theoretical max throughput for Gen 2 (~1050 MB/s) is more than double that of Gen 1, the real-world throughput difference is dependent on the capabilities of the storage media and the access pattern. Your benchmarks prove this out. Take any of the write benchmarks from the AnandTech DAS Suite and compare them to the same device plugged into a Gen 1 port.
The time is money set switched to Thunderbolt eight years ago and haven't looked back. USB 3.2 Gen 2x2 is essentially nonexistent and it appears that Intel wants to keep it that way by moving everyone to USB4 / Thunderbolt 4 instead. There are very few device categories for which USB3 Gen 2 silicon exists, and even basic items such as hubs are still exceedingly rare in the marketplace.
In many cases, USB3 is preferable to Thunderbolt because devices are considerably cheaper and will work just fine if you plug them into any USB port. But I think some users would trade the speed bump that comes with Gen 2 signaling simply to be able to use passive cables longer than 1m. The differences between USB 1.1, 2.0, and 3.0 were so great that they enabled entirely new use cases. What's the killer application for USB3 Gen 2 aside from "sometimes it's a bit quicker"?
Top tip: when it comes to USB 3.1 and 3.2 gear, ignore the number to the right of the decimal point, because the specification version number does not determine device signaling capabilities. Look instead at the part where it says 5/10/20Gbps, which all licensees are supposed to include in any marketing materials or packaging where the USB 3.1 or 3.2 version numbers are mentioned.
Also, 20Gbps is only for USB 3.2 Type-C ports that support dual-lane operation, which currently requires a discrete controller and are as rare as hen’s teeth.
I bought a 2TB Samsung X5 Thunderbolt 3 SSD (at great expense), but when I switched from an Intel NUC to an AMD Threadripper system, I came to realize just how proprietary Thunderbolt is to Intel, and with no Thunderbolt add-in cards existing for my AMD system, my drive is now a paperweight. My Zenith II Extreme board does have USB 2x2, so I might give one of these a try. I'm also hoping we see USB 4 add-in cards for AMD systems. I won't be buying any more proprietary Thunderbolt devices.
Hopefully there will be a full-featured USB4 AiC that does PCIe tunnelling, and your X5 can start working with your Threadripper system once you put that in place.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
14 Comments
Back to Article
repoman27 - Wednesday, September 30, 2020 - link
"Given that the Tiger Lake platform doesn't support USB 3.2 Gen 2x2, it is going to be interesting to watch how this plays out in the near future."Glad to see that you caught that. I was pretty surprised when I saw the lack of USB3 dual-lane operation while perusing the Tiger Lake UP3 datasheet. I had really hoped that Thunderbolt 4 would support all of the headline features of both the USB 3.2 and USB4 specifications.
JfromImaginstuff - Wednesday, September 30, 2020 - link
I didn't even know that there was such a standard as USB 3.2 Gen 2X2. Well there's always more to learn.repoman27 - Wednesday, September 30, 2020 - link
USB 3.x / SuperSpeed USB 5Gbps is Gen 1 signaling.USB 3.x / SuperSpeed USB 10Gbps is Gen 2 signaling.
USB 3.2 / SuperSpeed USB 20Gbps is Gen 2 signaling plus dual-lane operation, a.k.a. Gen 2x2.
brutedawg - Wednesday, September 30, 2020 - link
USB naming standards are such a mess, made shopping for a motherboard this past year an absolute pain with virtually every manufacturer inconsistently naming their I/Os.Arsenica - Wednesday, September 30, 2020 - link
If would be very simple if only they had made something like this:USB 3: 5Gbit/s - Type A connector (known as USB 3.2 Gen 1×1)
USB 4: 10Gbit/s - Type C connector (known as USB 3.2 Gen 2×1)
USB 5: 20Gbit/s - Type C connector (known as USB 3.2 Gen 2×2)
USB 6: 40Gbit/s - Type C connector (known as USB 4 Gen 3×2)
repoman27 - Wednesday, September 30, 2020 - link
I'm sure that with hindsight the USB-IF might have elected to do things differently. However, from their standpoint, it probably seemed easier to add a couple new signaling modes as optional features in point releases and not have to roll new licensing or maintain additional specifications and associated testing / compliance programs.Also, USB 3.1, which introduced the Gen 2 (10Gbps) PHY, was developed for Standard/Micro-A/B connectors and was finalized over a year prior to the Type-C Cable and Connector Specification. And USB4 hosts/devices only need to support Gen 2x2 (20Gbps) signaling, or up to Gen 2x1 (10Gbps) in USB3 mode.
Optional features can be beneficial to manufacturers and consumers alike, but they make the specification version number an unreliable indicator of device capabilities. And to a certain degree, the differences between the USB 3.x variants really only matter in edge cases. For years now, nobody has complained that USB 2.0 encompasses devices that might be capable of low-speed (1.5 Mbit/s), full-speed (12 Mbit/s), or high-speed (480 Mbit/s) signaling, despite the fact that they are separated by orders of magnitude. As long as the signaling rate is appropriate for the device, nobody cares. PC host ports are pretty much the only place where you might want to pay attention to the signaling capabilities, but a 10Gbps device only operating at 5Gbps is unlikely to impact most real-world workflows.
ganeshts - Thursday, October 1, 2020 - link
I am curious why you say 'a 10Gbps device only operating at 5Gbps is unlikely to impact most real-world workflows' ; For content creators on the go, for example, backing up work files from their desktop PC before venturing out to the field is a legitimate real-world workflow. A device operating at half the advertised speed is bound to double the waiting time of the user.I do agree that all the version numbers have been designed from the viewpoint of what manufacturers need to do / support to obtain certification - However, based on prior experience, USB-IF must have recognized that the version numbers are bound to be marketed directly to consumers. Reading more into the USB4 standard and its implementation in Tiger Lake, I am strongly convinced that it is shaping out to be a disaster in terms of consumers understanding its capabilities (akin to what has been happening with Type-C over the last few years).
repoman27 - Thursday, October 1, 2020 - link
The only time you benefit from Gen 2 signaling is when your attached device is capable of throughput greater than 435 MB/s. That means multiple SATA drives in a hardware based striped RAID configuration or PCIe SSDs with a Gen 2 bridge. Although the theoretical max throughput for Gen 2 (~1050 MB/s) is more than double that of Gen 1, the real-world throughput difference is dependent on the capabilities of the storage media and the access pattern. Your benchmarks prove this out. Take any of the write benchmarks from the AnandTech DAS Suite and compare them to the same device plugged into a Gen 1 port.The time is money set switched to Thunderbolt eight years ago and haven't looked back. USB 3.2 Gen 2x2 is essentially nonexistent and it appears that Intel wants to keep it that way by moving everyone to USB4 / Thunderbolt 4 instead. There are very few device categories for which USB3 Gen 2 silicon exists, and even basic items such as hubs are still exceedingly rare in the marketplace.
In many cases, USB3 is preferable to Thunderbolt because devices are considerably cheaper and will work just fine if you plug them into any USB port. But I think some users would trade the speed bump that comes with Gen 2 signaling simply to be able to use passive cables longer than 1m. The differences between USB 1.1, 2.0, and 3.0 were so great that they enabled entirely new use cases. What's the killer application for USB3 Gen 2 aside from "sometimes it's a bit quicker"?
repoman27 - Wednesday, September 30, 2020 - link
Top tip: when it comes to USB 3.1 and 3.2 gear, ignore the number to the right of the decimal point, because the specification version number does not determine device signaling capabilities. Look instead at the part where it says 5/10/20Gbps, which all licensees are supposed to include in any marketing materials or packaging where the USB 3.1 or 3.2 version numbers are mentioned.Also, 20Gbps is only for USB 3.2 Type-C ports that support dual-lane operation, which currently requires a discrete controller and are as rare as hen’s teeth.
Jorgp2 - Wednesday, September 30, 2020 - link
Fuck, USB is confusing.Intel states TG supports USB 4 Gen 2 x 2, which is the same speed as USB 3 Gen 2 x 2 but electrically incompatible.
boeush - Wednesday, September 30, 2020 - link
"USB 3.2 Gen 2x2 SanDisk Extreme PRO Portable SSD v2"weak.
Not nearly enough 2's in that name to keep me interested. Needs more 2's.
hubick - Thursday, October 1, 2020 - link
I bought a 2TB Samsung X5 Thunderbolt 3 SSD (at great expense), but when I switched from an Intel NUC to an AMD Threadripper system, I came to realize just how proprietary Thunderbolt is to Intel, and with no Thunderbolt add-in cards existing for my AMD system, my drive is now a paperweight. My Zenith II Extreme board does have USB 2x2, so I might give one of these a try. I'm also hoping we see USB 4 add-in cards for AMD systems. I won't be buying any more proprietary Thunderbolt devices.ganeshts - Friday, October 2, 2020 - link
Hopefully there will be a full-featured USB4 AiC that does PCIe tunnelling, and your X5 can start working with your Threadripper system once you put that in place.soulkeeper4 - Tuesday, November 24, 2020 - link
What about apple's M1 usb 4?are they compatible with these speeds?