Booting from SSD/SATA

You lack the ability to properly get your point across and have other actually understand you. Please don’t interject. The adults are talking.

1 Like

I am sorry, that I exceeded your ability. I will try to formulate it easier, next time. So you can understand it too.

as i alway to say , banana pi is not a perfect product, and open source software need all user to support us and help us .

but we try to do our best to support it. all banana pi product ,all is long time support .

when we do a product ,cost is very high , we must do all test . and Ensure the quality of the hardware, we need get all CE ,FCC,RoHS certificate, and do all wireless test ,and do all Validation test for all banana pi board . it not just for maker , you also can use banana pi do any cool product.

we also know ,Tkaiser not like us , We don’t know how to make him feel happy at here. but for BPI-M3, have about 50K at user hand .we just can try to support software , we will keep our "bad“ support , this is base work for open source community, we hope every one can help us grow up.

armbian have do a great work for arm board , we also learn from armbian community, we have finish our software framework, let us easy to support all banana pi board. so we will support all hardware and easy to maintain software.

This sums it up perfectly: You don’t care at all about your user’s needs but only your own ‘development’ needs. Not a single user has ever demanded an OS image that’s able to run on more than one of your boards. The only people interested in this are you and maybe a few colleagues of you.

Your users want to have one or two OS images per board that do work and fully support the hardware. And not countless crappy ones lacking features but containing some bootloader voodoo easing vendor’s testing efforts (if you would only really test the stuff you release). Why don’t you read the threads here and start to try to understand your users: Banana Pi images? + FEEDBACK

when are you going to listen to what they are saying?

BPI M3 is not a bad product but it’s hard to develop it. BPI fans can refer its SOC spec by, it features GPU by PowerVR which is not open source like Mali can be opened and released for users developing easily.It is tablet SOC without SATA function that makes sense that’s why we add extra IC to support SATA for serving BPI fans.


A83T has 1 USB host port and 1 USB HSIC port. The USB-to-SATA bridge you use on BPi M3 is the slowest in the world. Why did you choose this? Is this part of Allwinner’s reference design? Since CubieTech and Xunlong use the same ultra slow GL830 on their boards? At least they were smart enough to attach GL830 to an own USB port. On BPi M3 the 2 USB receptacles and the SATA bridge share the bandwidth of one single SoC USB port and the other USB port is not used at all.

Add the lousy decision to use Micro USB for DC-IN, the thermal problems, your broken THS settings (you just took the defaults from Allwinner which are inappropriate for an SBC), not a single fully working OS image for BPi M3, the state of ‘support’ where not even the most basic questions are answered here in the forum and you might get the idea what’s wrong (of course you won’t since thinking from a customer’s perspective is nothing worth a thought).

M3 onboard SATA and microUSB for DC-In on 2nd version due to BPI fans required. Of course, When they got samples that they found above feature and power change are not good and gave us feedback, we replaced microUSB with DC-IN for MP version.

Our SW is improving step by step and fans are pleased our upgrade and adopt various OS images to fulfill different applications.

We have shipped hundred K quantities, we must release more OS images by global users demand by applications. I checked CubieTech and Xunlong shipping quantities which can’t compete BPI’s volume, we will face more users feedback and demand on BPI models.

Your advice will be considered for upcoming project!


Ok, and you’re Banana Pi’s product manager, true?

There exist quite a few different USB-to-SATA bridges. The one you used (Genesys Logic GL830) is the worst known (only 15 MB/s write throughput and 30 MB/s read throughput – any other bridge exceeds 35 MB/s easily in both directions. Are you able to spot the difference? Are you also able to spot the difference between this horribly low performance and real SATA performance BPi M1/M1+ show: 45MB/s write performance, ~200MB/s read performance and way better random IO. Are you able to get the difference? Do you get the idea that it is just fooling potential customers when you tell them BPi M3 ‘has a SATA port’?).

Even Genesis Logic has better chipsets available. So why did you chose the worst chip that makes this onboard SATA port so useless and performance so painfully low? That was a simple question. Paired with an assumption that you simply copied Allwinner’s reference design and they use GL830 there. And that was the reason I mentioned Xunlong and Cubietech since they use the same shitty chip on OPi Plus, Plus 2 and Cubietruck Plus (a choice nobody understands since why choosing an ultra slow chip when faster are available).

I didn’t ask for shipping quantities, I just named two other Allwinner board vendors that also use this unbelievable crappy USB-to-SATA bridge. To ask you the simple question why you chose exactly this chip. Since using a good USB-to-SATA bridge onboard is somewhat ok but why choosing the slowest in the world? That was the simple question you either fail to answer or to understand. Or why don’t you answer this simple question?

Apart from that I really fear you’re believing the stuff you wrote before…

all user maybe wise than you, maybe are you want deceive someone. all is know what our hardware spec and tell use sata port is USB bridge .and choose by user. someone maybe just need a sata port , not need quick speed . if user need a real sata ,they know to choose M1,M1+,R1, all is for long time support.

Not one development board can meet the needs of all people, we only do our work and choose by user. why you so angry. if you not like us ,just as i say ,go away, if not ,please keep support .:joy:

you are not any experience on hardware,

No matter A83T or H8 or R58 without SATA support from SOC, Allwinner only has A20 with real SATA, what’s your advice for 8-core project to support SATA if customers insist this function?


Don’t care about what he said, maybe He have other purposes,But he took a wrong way.

I collected the results of countless hours of testing where it belongs to: in the community wiki:

As a side note: I would suspect A83T’s USB implementation to be the same as H3’s so if you would care about your users and keep in mind that linux-sunxi community might get mainline kernel support for A83T working in the future choosing an UASP capable USB-to-SATA bridge would be the best choice since both sequential transfer speeds improve and especially random IO will get faster with mainline kernel.

Again: Why did you choose the slowest USB-to-SATA bridge in the world? There must be a reason why you chose exactly the worst option possible? Do you develop the hardware you sell? Or why can’t you tell why you chose inferior components?

for some user need this function on BPI-M3, it can support they function , if you not need ,just not use it. so easy.

These SoCs have 3 USB ports. One OTG, one host port and one HSIC. So they clearly lack bandwidth compared to all your other boards. So why did you chose to limit IO bandwidth even more? You do not use the one USB host port at all but connect the HSIC port to an internal USB hub and behind this hub is both the slow SATA bridge and the 2 externally available USB ports. This way performance will further decrease when bandwidth intensive USB peripherals are used.

Why not doing it like Cubietech on their H8 design (using both host and HSIC port and choosing a more intelligent USB hub)? Can you spot the difference between your and their USB solution?

Unfortunately they use the same slow GL830 SATA bridge but at least it has not to share bandwidth with other USB ports.

BTW: Neither you nor the famous @sinovoip support muppet had a look at the question this thread started with. Don’t you think providing documentation containing some advices like ‘install to eMMC if possible since…’ would be a good idea since you chose to never answer support questions here in this channel?

we are try to write documents on gitbook. and know how to install image to SD card or eMMC flash

not good ,but user can reference

maybe we need waitting for a H8 hardware, when it achieve, you can understand why @tkaiser always not like us.:slight_smile:

I didn’t choose H8 last year due to only support Android 4.4 but Android 5.x~6.0. For M3 SW will be improved, I don’t suggest H8 onboard, we must focus on M3 SW development. Org developers got M3 samples, they will help us too, don’t worry!


i just mean an other H8 board, not do by us. this is why…

One final word from me. If you would care about your customers and given the poorly designed USB situation on BPi M3 (you waste one USB host port completely for no reason – see how Cubietech does it better above) I would strongly recommend to document for your customers that the Micro USB OTG port can be used almost identical as an USB host port now that you fixed the fex configuration for this port after customers complained over half a year while you weren’t listening to them).

Performance of the OTG port in host mode is almost as good as with a native USB host port. See my test results here (H3 and A83T perform identical on the OTG port, on the type A receptacles and on the SATA port it’s a different story since you f*cked up your board design here so badly choosing to waste one USB port and this slow USB-to-SATA bridge)

It should also be possible to use BPi M3 as an USB gadget device through the USB OTG port but I would assume the same small patch is necessary as outlined here for H3 BSP kernel:

Apart from that @sinovoip is really doing a great job publicly outlining what to expect from your company. And it should be noted that at least here in Europe BPi M3 customers all believe M3 would be a powerful upgrade compared to M1, that it has a real SATA port since every reseller lists exactly that and that over 95 percent of devices are just collecting dust in the meantime and all those people will never buy from you again since they can’t do anything useful with this board and feel they got duped (they paid the extra price for a reason since they thought they get a beefy board capable of doing a great job as server or desktop replacement. No one is dumb enough to buy M3 to be used for GPIO stuff since it’s way too expensive for that)

But from your perspective as product manager it’s better to ignore all of this of course :slight_smile: