Discussion

read/write speed of a HDD through Guitar USB3.0 Micro-B Receptacle

29 2992
hange  
I found that the cubieboard has the same question, just look at the pictures:
cubieboard.png
USB3.0pic.png
U48.png

tkaiser  
Edited by tkaiser at Tue Nov 3, 2015 03:24

Ok, so now you did 'an SSD test'. And what does the results mean? The Guitar is horribly slow, right? Just above 100 MB/s with an ultra fast most recent SSD is nothing you should publish. Better hide the results.

But what have you tested in reality? Nothing, since you used a wrong SSD for this purpose. SSDs with small capacity are not that fast compared to their bigger siblings. When you use an SSD with advanced cache technologies like Samsung's TurboWrite the size of the cache also matters. And you chose a SSD with 3 GB cache and a test size of 4 GB:

http://www.anandtech.com/show/8747/samsung-ssd-850-evo-review/2 (you would have to choose an EVO with at least 500 GB to get results with a meaning)

And the whole approach is totally useless unless you use a 'reference setup' where you did the very same test to get a clue whether your SSD is the bottleneck (true in your case), your test setup might be wrong (also true since you seem to be clueless regarding storage technologies like SLC/TLC) or the USB3 implementation is slow (this was the question but your test setup failed to provide answers).

You would need to use an SSD known to be able to get beyond USB3 limits (+400 MB/s when using UASP) connected to your reference setup. And when you then use the exactly same test methodology on the Guitar, THEN you will get results that have any meaning. That's due to the simple fact the the USB-to-SATA bridge used in the USB enclosure can also have a massive impact on performance (which chipset did you use? Without knowing that the results are worthless!). That's why you need a reference setup using the same relevant hardware components (same USB enclosure and same USB cable -- haha, yours is custom made and doesn't work outside the LeMaker world) and the same testing methodology.

If you're able to get performance values like this on a MacBook and use the same USB enclosure together with the same cable (haha, not possible) afterwards to measure the Guitar's USB3 performance (consisting of both hardware and software!) then you get results with a meaning. Otherwise not.



But again: Time to stop. It's moronic to discuss this stuff with the manufacturer if even such basics aren't considered and you as the manufacturer of the device seem to do performance tests (wrongly) now after the sales already started. And given the current situation (USB3 NOT useable due to lack of commercially available cables) the whole discussion is useless and the Guitar's storage performance sucks anyway due to USB 2.0.

tkaiser  
hange replied at Mon Nov 2, 2015 20:59
I found that the cubieboard has the same question, just look at the pictures:

On the Cubieboard's schematics there's written Micro-AB and not Micro-B. And according to what I understood from reading through USB specs (unbelievable that a vendor forces a customer or reviewer in this case to do that) this would've been the correct receptacle: https://en.wikipedia.org/wiki/USB_On-The-Go#OTG_micro_plugs

But I stop here since the Guitar doesn't make any sense to me. There are tons of SBC available featuring the RPi's GPIO header. Why choose an expensive Guitar?

Rate

1

View Rating Log

cxy  
This performance of the USB3.0 has been very good to me, and i can use it replace with the Bpro .
@ tkaiser Could you recommend better board to me ? i hope it have higher read/write speed in the storage and higher cost performance.

bprpcb  
tkaiser replied at Tue Nov 3, 2015 01:34
On the Cubieboard's schematics there's written Micro-AB and not Micro-B. And according to what I u ...

You are joking ,  i use the cubieboard for a long time, and it use the Mirco-B not Mirco-AB, please don't mislead me and view the following picture.
cubieboard.png

You can through the picture below to see the difference between Mirco-B and Mirco-AB .
mirco-b.png

I am planning  to apply for a Guitar, and build my wireless router.

tkaiser  
cxy replied at Tue Nov 3, 2015 07:37
This performance of the USB3.0 has been very good to me, and i can use it replace with the Bpro . :l ...

Sorry? USB3.0 performance 'very good'?

Since noone outside LeMaker is able to test USB3.0 since they chose the super superiour Micro-B connector with non-standard Pin mappings we have to rely on the test results posted here.

When you get 118 MB/s sequential write and 106 MB/s sequential read from a device that's able to reach 150 and 540 MB/s than this is a clear indication that the performance of the test setup really sucks. First of all they still choose only wrong devices (which are too slow to draw conclusions from when performance is bad -- this applies to the SSD they used as well since it features only a 3 GB TurboWrite buffer) and then the whole test setup is crap since it's not clear what might be responsible for this bad performance (the USB-to-SATA bridge, the 'custom cable' being responsible for retransmits, the S500's USB3 hardware implementation, the driver status, silly settings, and so on).

Given this is the maximum limit one can reach with S500's single USB3 connection this would mean if I want to use the Guitar as NAS an GBit-Ethernet adapter and an USB-to-SATA bridge have to share bandwidth and you end up with a network performance of maximum 50 MB/s. Sorry but with the old Banana Pi you can get close to 44/72 MB/s if you setup all parameters correctly. If the Guitar's USB3 performance is really limited to approx. 110 MB/s then where's the point?

But as already said: the whole test is pretty useless. And you can not test storage performance when you ignore EVERYTHING that's important in this area. This applies especially to slow ARM devices where you have to take care of cpufreq settings and IRQ affinity and the like. And without a reference setup known to be able to reach the USB3 limits every test result is completely meaningless.

If they would've chosen a suitable SSD (able to exceed at least 300 MB/s) connected to a JMS567 USB-to-SATA bridge and get these 300 MB/s when connected to a known performant USB host implementation (eg. any Intel chipset with included USB3 controller manufactured after 2012) and if they afterwards use exactly the same test setup with the Guitar and results are in the range of 110 MB/s then we know that there's something wrong with the Guitar's USB3 implementation (be it software -- for example wrong IRQ settings or the still missing UASP support since Actions Semi has no interest to get their SoCs supported in mainline kernel -- or hardware).

A recommendation for a 'better' board? What does 'better' mean? It always depends on the use case and what's important to you (price, features, performance, whatever).

Maybe someone over there is able to do real USB3/storage testing: http://forum.linux-xapple.org/t/ ... rage/37/2?u=tkaiser

tkaiser  
bprpcb replied at Tue Nov 3, 2015 08:29
You are joking ,  i use the cubieboard for a long time, and it use the Mirco-B not Mirco-AB, ple ...

Woohoo, that's great, so the picture above (from their schematics?) is wrong. There is written Micro-AB and that's all I know because I don't own a Cubieboard 4 since Linux support for the board really sux.
But that's interesting. I just had a look for it in a local reseller's shop. Cubietech bundles these two cables (maybe for a reason LeMaker still doesn't get?):


bprpcb  
Edited by bprpcb at Tue Nov 3, 2015 09:20
tkaiser replied at Tue Nov 3, 2015 08:42
Woohoo, that's great, so the picture above (from their schematics?) is wrong. There is written Mic ...

May be, But i think the USB3.0 MIRCO-B on the Guitar and Cubieboard 4 meet USB3.0 standards, just for compatibility with the USB3.0 Device and USB 3.0 Host on the Mirco-B, have to customize the USB3.0 OTG wire.

tkaiser  
bprpcb replied at Tue Nov 3, 2015 09:18
customize the USB3.0 OTG wire

Why? Why USB3 OTG?

It's still not about OTG. It's the host mode that isn't working. USB OTG works but this isn't USB3 but USB2 instead. But anyway, if the manufacturer of the device knows for sure that you can't buy the cable you would need to use USB3 peripherals then the only reasonable decision is to bundle such a cable with the device like Cubietech did. Since otherwise USB3 isn't useable. And that describes pretty much the situation with the Guitar: No cable bundled with baseboard B --> no USB3 available.

bprpcb  
tkaiser replied at Tue Nov 3, 2015 09:29
Why? Why USB3 OTG?

It's still not about OTG. It's the host mode that isn't working. USB OTG works  ...

Test speed data from this post,USB3.0 Host mode on the Guitar is Ok. Although the read/write speed can't reach the level of the computer, and  they have used the USB3.0 OTG cable, as follows
usb3.0 OTG.png
I think your conclusion is too one-side. On the contrary, I believe the test data. I have already applied for the LeMaker Guitar, If i can get one, and I will show the USB3.0 speed to you again.

Comments

Well, go and read http://forum.armbian.com/index.php/topic/311-quick-review-of-lemakers-guitar/ when you understood, return.  Post time Nov 03, 2015 15:00

You have to log in before you can reply Login | Sign Up

Points Rules