[Computex] AsMedia Demos USB 3.2 Gen2x2 (Up To 20Gbit/s) PHY with Controller due in 2019

[Dabombinable]

*still wishes firewire hadn't been dropped but improved further*

[bcredeur97]

at this rate motherboards are going to come with internal USB ports instead of SATA connectors.

Are we just giving up on the SATA drive form factor? I mean arguably with hard drives it doesn't need to be any faster than 6gbit/s but idk...

[Dabombinable]

2 minutes ago, bcredeur97 said:

at this rate motherboards are going to come with internal USB ports instead of SATA connectors.

Are we just giving up on the SATA drive form factor? I mean arguably with hard drives it doesn't need to be any faster than 6gbit/s but idk...

The problem with USB is that it has more overhead than SATA.

[Drak3]

9 minutes ago, bcredeur97 said:

at this rate motherboards are going to come with internal USB ports instead of SATA connectors.

Are we just giving up on the SATA drive form factor? I mean arguably with hard drives it doesn't need to be any faster than 6gbit/s but idk...

SATA III is a dedicated 6Gbps per port. USB Is 5/10/20 per controller, and each port divides that.

 

So a system like mine, having 6 SATA can do a theoretical combined 36Gbps if writing or reading from all drives at once.

[AlTech]

50 minutes ago, Sauron said:

I understand that, the problem is that it doesn't make a difference - I can't reliably tell what a port will do for me just by looking at it or its name, and that is a problem.

As I said, they should have waited, released Type C along with 3.2 and called it a  day. This way we'd know exactly what a Type C port can do for us and what a Type A can't.

In Linus' last Apple MBP video, he commented on how tons of people on Apple's support website were so confused and left bad ratings cos they didn't understand the difference and were totally confused. 

[Sauron]

50 minutes ago, AluminiumTech said:

In Linus' last Apple MBP video, he commented on how tons of people on Apple's support website were so confused and left bad ratings cos they didn't understand the difference and were totally confused. 

Of course, if Apple hadn't put 4 physically identical but electrically different ports on the same machine it might have helped...

[AlTech]

17 minutes ago, Sauron said:

Of course, if Apple hadn't put 4 physically identical but electrically different ports on the same machine it might have helped...

Yup. Are the left ones 4x lanes each and the right 2x lanes each? cos allegedly the left are more powerful than the right .

[Shadestones]

Why can't they just call it USB 4.0 and be done with it? Who is in charge of their marketing team?

[AkatsukiKun]

TBH I blame anyone that ever marketed USB 3.1 without noting the gen or actual speed because it's all terrible marketing from them. Although still disappointed of the lack of USB 3.1 (10Gbps) devices, yes  I know they are there but just the cost of it is somewhat hard to eat up. And so my Crosshair VI Hero's 3.1 ports are lonely.

[Guest]

13 hours ago, themctipers said:

Thunderbolt 3 take over for fucks sake.

If I remember Thunderbolt 2 was 20GB/s with 2 lanes. 

[AlTech]

32 minutes ago, RorzNZ said:

If I remember Thunderbolt 2 was 20GB/s with 2 lanes. 

Gbit/s but yes.

[Doobeedoo]

They're making even more confusion and fragmentation like this with these namings. Also how Type-C was introduced. They should've just called it 4.0 and have speeds that are faster than A can achieve. Not like how it's now. 

Anyway, everyone needs to hurry incorporating Type-C everywhere already. It's a slow transition. Still more expensive cases without it. Barely any peripherals with it. Motherboards only having one. Lame. 

[Trixanity]

I wonder what the thought process behind USB naming conventions are. You'd think they could find a common ground between engineering and marketing because if you give the job to either you get stupid names like this. 

Really makes you think how a guy could come up with this and some other guys patted him on the shoulder and said good job.

[Drak3]

14 hours ago, AluminiumTech said:

Yup. Are the left ones 4x lanes each and the right 2x lanes each? cos allegedly the left are more powerful than the right .

If all 4 are used, they're x2. Only 8 lanes are used for TB.

I believe there is one controller per side, use two on the left, x2. Same for the right. Use one from each side, x4.

[AlTech]

12 minutes ago, Drak3 said:

If all 4 are used, they're x2. Only 8 lanes are used for TB.

I believe there is one controller per side, use two on the left, x2. Same for the right. Use one from each side, x4.

Then why the discrepancy regarding regarding performance on the left vs right?

[Drak3]

2 minutes ago, AluminiumTech said:

Then why the discrepancy regarding regarding performance on the left vs right?

People probably have a tendency to plug stuff in on the same side as their dominate hand.

[Guest]

12 hours ago, AluminiumTech said:

Gbit/s but yes.

Ty.

 

i think also this opens up new avenues for eGPU solutions 

[Unixsystem]

USB-C 5, supports X protocols

USB-C 10, supports X+Y protocols

USB-C 20, supports X+Y+Z protocols

 

Was that so hard?

 

[schwellmo92]

Is this true 20Gb? TB2/3 have always annoyed me in that they round up their speeds to the nearest 10.

[Sniperfox47]

On 2018-06-08 at 1:31 PM, AluminiumTech said:

Here's some context:

USB 3.2 is the most recent USB Standard that was released a while back. 

 

USB 3.2 has 4 compatibility modes:

• Gen1x1 - Basically the same as USB 3.1 Gen 1 and USB 3.0 (5Gbit/s using 1 lane)
• Gen1x2 - 10Gbit/s using 2 lanes
• Gen2x1 - Basically the same as USB 3.1 Gen 2 (10Gbit/s using 1 lane)
• Gen2x2 - 20Gbit/s using 2 lanes

 

Quote

This also causes more issues as currently USB 3.1 support is entirely hit or miss regarding 5Gbit/s vs 10Gbit/s . USB 3.2 supports 5GBit/s 1 lane, 10Gbit/s 2 lane, 10Gbit/s 1 lane, and 20Gbit/s 2 lane which is likely to cause more fragmentation over which standard is actually used by Device OEMs and Controller support.

 

So yeah, I'm liking the improved bandwidth but do we really need so many compatibility modes and configurations to work with everything from USB 3.0 to now?

 

Sources:

https://en.wikipedia.org/wiki/USB_3.0#USB_3.2

https://www.anandtech.com/show/12922/asmedia-demos-usb-32-gen-2x2-phy-usb-32-controller-due-in-2019

Yeah we do for power savings. All of these modes are used for different things.

 

Gen1x1 and Gen 2x1 are used for compatibility with USB 3.1gen1 and gen2 devices to avoid compatibility issues with the new PHY layer. It's one of the big benefits to XHCI over previous host interfaces, the ability for a USB controller to fall back to older speeds to ensure 100% compatibility without needing a whole different host layer. They're for legacy peripherals only.

 

Gen1x2 isn't expected to be widely used, but could be useful for USB 3.1gen1 hubs. Rather than taking 1 USB 3.1gen1 lane and splitting it 4 ways, you could have 2 USB 3.1gen1 lanes and then only split each two ways. This doubles up the effective bandwidth for devices on your hub which is useful for high speed storage. It's only for peripherals with legacy support and *must* be able to fall back to Gen1x1 for compatibility.

 

Gen2x2 is the new mode and supports new USB 3.2 devices. If your peripheral is USB 3.2 it's using this mode.

 

A downstream facing (i.e. on the host computer/phone) Superspeed (3.1gen1) or Superspeed+ (3.1gen2)  port has to support gen1x1 or gen2x1 respectively.

A downstream facing Enhanced Superspeed+ (3.2) port *must* support all four modes.

The spec also technically allows Enhanced Superspeed without the plus which would just be Gen1x1 and Gen1x2 but that's kinda silly so I doubt there will be any controllers for it.

 

An upstream facing (i.e. on your peripheral) port must also support all lower modes than what it's designed for. So a USB 3.2 Enhanced Superspeed+ USB drive will still fall back to Gen2x1 and Gen1x1 for USB 3.1 gen2 and gen1 ports.

 

There's no increase in fragmentation here because all of the devices are inter-compatible. It has so many compatibility modes specifically to be sure that nothing breaks. That's one of the reasons it's taken so long for ASMedia to actually make a controller in the first place.

 

On 2018-06-08 at 2:20 PM, Sauron said:

If Type C had been a single, well defined standard, sure - but it wasn't, and only now are we starting to get all the features the cable theoretically allows. Nobody was in a rush to get usb 3 on smartphones (even with quick charge we're nowhere near the peak power delivery Type C supports) and waiting a year or two wouldn't have hurt.

It would have hurt by delaying the start of adoption by 2 years. USB-C also has *NOTHING* to do with the speed of the port or cable. Just like USB-A/B had *NOTHING* to do with the speed of the port or cable. You can have USB-A at USB 1.0, USB 1.1, USB 2.0, USB 3.0, USB 3.1gen1 or USB 3.1gen2.

 

USB 3.X wasn't a rush but USB-C definitely was for smartphone devs. And USB 3.X was a byproduct of them wanting Displayport Alt-mode. Before you had to have your Micro-USB port inside which was bigger internally than Type-C. Plus you had to have a separate controller chip for Slimport or MHL. Then you had to have a janky USB switch between the two modes and couldn't run them at the same time because MHL/Slimport were not USB compliant. And then you had to have all kinds of proprietary standards with their own charging chips.

 

USB-C allows you to cut all that down to 1 maybe 2 chips. You have your USB controller which may be built into your SOC. And then you have your USB-C Matrix Switch which takes in Power, USB, and Displayport and handles them all over the USB-C port at the same time, in addition to the port itself taking up less space.

 

And I don't know where you're getting your info. USB-C *is* a single well defined standard. The fact that a bunch of companies weren't half-assed bothered to implement it in a compliant fashion comes more down to the USB-IF not holding anyone accountable, and the huge rush of cheap aftermarket brands bringing cheaply made stuff out to take advantage of the initial rush.

 

13 hours ago, Waffles13 said:

USB-C 5, supports X protocols

USB-C 10, supports X+Y protocols

USB-C 20, supports X+Y+Z protocols

 

Was that so hard?

That's basically what they've done.

USB 3.1gen1 (Superspeed) supports 5G

USB 3.1gen2 (Superspeed+) Supports 5G and 10G

USB 3.2 (Enhanced Superspeed+) supports 5G, 10G, 2x5G, and 2x10G

Those are the consumer facing names.

 

7 hours ago, VegetableStu said:

kinda curious if there's any performance difference in a x2 USB 3.2 10G connection and a x1 USB 3.1 10G connection o_o

(or is it a 1 kg sack of cotton and 1 kg slab of iron thing)

Performance difference there shouldn't be. Compatibility difference there definitely is, particularly with hubs.

 

A gen2x1 connection is faster for a single device but will have difficulty getting full speed when split by a hub due to added switching overhead. Won't work great for lots of 3.1gen1 devices but will be great for a few 3.1gen2 devices. A gen1x2 connection with hub won't get you the same performance for a single device (USB 3.0 speeds only) but will handle more of those 3.0 devices better. Your hub should automatically handle switching modes for you, since that's also part of the spec, with the hub monitoring the fastest downstream connection and using that speed with as many lanes as possible upstream.

 

For devices If your device supports 2 lanes it should support all four modes anyways, so it'll just use 3.1gen2 or 3.2 speeds, whichever is better. Unless you have some weird controller with gen1x2 but no gen2 support... which is technically allowed but nobody's currently making controllers for it because it's dumb.

 

 

[Unixsystem]

6 hours ago, Sniperfox47 said:

That's basically what they've done.

USB 3.1gen1 (Superspeed) supports 5G

USB 3.1gen2 (Superspeed+) Supports 5G and 10G

USB 3.2 (Enhanced Superspeed+) supports 5G, 10G, 2x5G, and 2x10G

Those are the consumer facing names.

By "protocols" I mean things like HDMI/DP/Ethernet over USB. Correct me if I'm wrong, but I believe that the ability of a specific port to support one or all of those protocols is completely independent of which USB spec the port is. That's needlessly complicated and results in unnecessary confusion in what needs to be plugged in where to get full functionality. If the different speed standards were much more clearly delineated it would be a lot easier for a laptop manufacturer to just go "this system has 2 USB-5 ports, 2 USB-10 ports and one USB-20 port", and you'd know immediately what that means in terms of connectivity with displays, internet adapters, etc. No fussing with how many lanes the port is, how much power it can carry, or what alternate modes it supports.

[Sniperfox47]

5 hours ago, Waffles13 said:

By "protocols" I mean things like HDMI/DP/Ethernet over USB. Correct me if I'm wrong, but I believe that the ability of a specific port to support one or all of those protocols is completely independent of which USB spec the port is. That's needlessly complicated and results in unnecessary confusion in what needs to be plugged in where to get full functionality. If the different speed standards were much more clearly delineated it would be a lot easier for a laptop manufacturer to just go "this system has 2 USB-5 ports, 2 USB-10 ports and one USB-20 port", and you'd know immediately what that means in terms of connectivity with displays, internet adapters, etc. No fussing with how many lanes the port is, how much power it can carry, or what alternate modes it supports.

You don't have to worry about how many lanes the port is. That's all on the implimentation side.

 

You also don't need to know specifically how much power the plug can carry, you just need to know if it can or can't carry power, which is basically "Did it come with a USB-C power cable? Then yeah it supports USB-C power. Did it come with monoplug? Then no it probably doesn't support USB-C power, use a monoplug adapter." I mean yeah you need to get a power adapter at least as powerful as the plug draws, but the same is true of *any* power adapter, even Apple's magsafe adapters where they charge the same price for every wattage anyways.

 

And alt modes, they're optional for a reason. There's plenty of implimentations and form factors where they're not needed or not feasible to impliment. Every laptop maker other than Apple clearly marks every port with the Alt-modes it supports. They're also not USB.

 

Consider a laptop. How many display outputs does it have? If you want to include 4 USB ports you probably don't have enough graphics lanes to give them all displays.

 

Consider a budget phone. You only have one display lane on your GPU. You *can't* impliment DP Alt-modes.

 

And none of this is any worse than the previous option for display over USB (DisplayLink) which could bug out on certain USB controllers, added a ton of latency due to framebuffer copying, would drop frames and lose quality while the controller was in use by, worked better or worse with certain ports on the same device. And required proprietary drivers that barely worked and frequently broke even with the operating systems they were officially supported on.

 

And with regards to Ethernet adapters? Literally nothing has changed with USB-C. Nobody is using the GBE alt mode except with Thunderbolt 3. Nobody else is. You're still just using the old "pass Ethernet frames over USB protocol" method. Which is why you have so many issues still with any given adapter maybe working on a specific device or maybe not since there's like 3 different networking driver protocol stacks that the USB adapters can support with some devices supporting one or the other or both or all 3. The GBE protocol would have at least clarified things in a decade when everything finally moved to using that.

[UpTrance]

excelente

[Delicieuxz]

Will USB 3.2 require specific drivers, or will it function with basic USB 3.0 drivers installed?