Not All USB 3.0 Implementations Are Created Equal

[qubit]

Despite the fact that today's USB 3.0 products center on the same NEC controller, we compared a handful of different USB 3.0 drives and found performance to range from 113 to 173 MB/s, depending on the implementation used. Should you be worried?

You might think that the performance of one USB 3.0-equipped product would be (or at least should be) the same as another. If so, you'd be wrong. We found that five different components, all based on the NEC controller, deliver different performance.

Doesn't seem like it's ready for prime time, does it? This reminds me of the significant performance differences at stock speeds of mobos in the 90s which used the same chipsets.

Tom's Hardware

 

[FordGT90Concept]

USB 3.0 requires a lot of bandwidth and some chips simply don't give it enough or there's too many other perhiperals eating away at the available bandwidth. I'm sure you'd see the same issue with SATA 6 Gb/s

 

[PVTCaboose1337]

This is why I am slow to adapt to new standards. Maybe wait till USB 3.1? I just stick to ESATA for now, although it does have a TON of flaws, it works fairly well.

 

[newtekie1]

113MB/s? Still about 4 times as fast as USB 2.0, so sounds good to me.

 

[qubit]

USB 3.0 requires a lot of bandwidth and some chips simply don't give it enough or there's too many other perhiperals eating away at the available bandwidth. I'm sure you'd see the same issue with SATA 6 Gb/s

I'm sure you're right and I have to confess, that I've only read the first few paragraphs of this article, so they might mention this point in it somewhere.

113MB/s? Still about 4 times as fast as USB 2.0, so sounds good to me.

+1 I'd take it, ready for prime time or not, as long as it works without errors. Performance would only improve over time.

 

[mlee49]

Dont some USB2.0 devices benefit just from being connected to a USB3.0 slot? I want to say Mussels got better transfer rates(not 2-4x better) just by using the 3.0 slot.

I'm not sure on this, but would be nice to see.

 

[PVTCaboose1337]

Dont some USB2.0 devices benefit just from being connected to a USB3.0 slot? I want to say Mussels got better transfer rates(not 2-4x better) just by using the 3.0 slot.

I'm not sure on this, but would be nice to see.

I have no idea why you think connecting a 2.0 device to a 3.0 would magically increase performance. I think you fail to see the concept of a bottleneck. I could be wrong, but I am almost 100% sure putting a 2.0 device into a 3.0 port would do nothing (to increase performance).

 

[mlee49]

I think you fail to see my last sentence:

I'm not sure on this, but would be nice to see.

Dont worry when my UD5 comes on Saturday I'll fire it up and see for ya.

 

[qubit]

Dont worry when my UD5 comes on Saturday I'll fire it up and see for ya.

I'd be keen to see if it improves performance.

There can be unexpected results sometimes. Say, the throughput rate isn't steady on a USB 2.0/2.0 connection, something one sees all the time on a network (try transferring a big file between two networked PCs and watching the throughput with NetMeter to see how it often drops massively). Making a USB 2.0/3.0 connection might resolve this and wouldn't surprise me at all.

 

[PVTCaboose1337]

Ok everyone, I made a diagram of how USB works, from 2 perspectives. Stand back:

To mlee, not trying to pick on you, just think this is hilarious.

 

[mlee49]

Massive win!!! Thanks PVT, caught a laugh while doing homework.

 

[qubit]

Ok everyone, I made a diagram of how USB works, from 2 perspectives. Stand back:

http://forums.techpowerup.com/attachment.php?attachmentid=37906&stc=1&d=1283831434

To mlee, not trying to pick on you, just think this is hilarious.

Dunno if you saw my answer in post 9. In theory, of course you should see no difference at all, but it's possible that making a USB 2.0/3.0 connection might make the data transfer rate more consistent and allow it to reach full USB 2.0 speed, assuming it wasn't already for some reason.

I haven't tried this out myself, but it sounds like a plausible scenario to me.

 

[PVTCaboose1337]

Dunno if you saw my answer in post 9. In theory, of course you should see no difference at all, but it's possible that making a USB 2.0/3.0 connection might make the data transfer rate more consistent and allow it to reach full USB 2.0 speed, assuming it wasn't already for some reason.

I haven't tried this out myself, but it sounds like a plausible scenario to me.

I understand what you are saying, talking about how 3.0 would make 2.0 work to its full potential. This is not the case however. Let's say the 2.0 device transfers max of 1 mb/s, and the 3.0 a max of 9000mb/s. Regardless, the max you can have the 2.0 go is 1 mb/s. Now if in a 2.0 port the 2.0 device was going .9mb/s then yes maybe it can go 1mb/s in the 3.0 port, but in reality, the bottleneck is IN the 2.0 device, meaning that there should be no change. Theory says no change. Reality will make no change. I can do some tests for you on a 3.0 card with a 2.0 device and a 2.0 device in a 2.0 port on the same machine. I promise you the results are so close together that the difference in negligible.

 

[qubit]

I understand what you are saying, talking about how 3.0 would make 2.0 work to its full potential. This is not the case however. Let's say the 2.0 device transfers max of 1 mb/s, and the 3.0 a max of 9000mb/s. Regardless, the max you can have the 2.0 go is 1 mb/s. Now if in a 2.0 port the 2.0 device was going .9mb/s then yes maybe it can go 1mb/s in the 3.0 port, but in reality, the bottleneck is IN the 2.0 device, meaning that there should be no change. Theory says no change. Reality will make no change. I can do some tests for you on a 3.0 card with a 2.0 device and a 2.0 device in a 2.0 port on the same machine. I promise you the results are so close together that the difference in negligible.

Thanks for that. It's just that I've seen how crappy data transfer performance can be over a network - and I'm talking about a very simple 100Mb (and later 1Gb) connection between two Windows PCs at home connected through a hub - so I figured if the same thing happens with USB it might improve it. Perhaps this problem doesn't exist with a USB connection? I really don't know and I'd be very grateful if you could do that experiment and post the result here.

 

[PVTCaboose1337]

Thanks for that. It's just that I've seen how crappy data transfer performance can be over a network - and I'm talking about a very simple 100Mb (and later 1Gb) connection between two Windows PCs at home connected through a hub - so I figured if the same thing happens with USB it might improve it. Perhaps this problem doesn't exist with a USB connection? I really don't know and I'd be very grateful if you could do that experiment and post the result here.

I was speaking hypothetically, but you have now given me a reason to get a 3.0 card for my laptop. I might as well...

 

[mudkip]

I have no idea why you think connecting a 2.0 device to a 3.0 would magically increase performance. I think you fail to see the concept of a bottleneck. I could be wrong, but I am almost 100% sure putting a 2.0 device into a 3.0 port would do nothing (to increase performance).

When I use my External HDD with USB 2.0 on my USB3 the transfer rates are about 40-50-60 MB/s instead of the usual 20-25 MB/s with USB 2.0

 

[qubit]

When I use my External HDD with USB 2.0 on my USB3 the transfer rates are about 40-50-60 MB/s instead of the usual 20-25 MB/s with USB 2.0

Yeah, I'm not surprised. It shouldn't make a difference, yet it does. Perhaps it's to do with drivers? Who knows.