Rock64 micro-sd controller defective?

[soichiro]

HHi3rqe4


Hello everyone,

I am not sure if this is software related problem or is just my Rock64's micro-sd controller that is defective. I tried different micro-sd cards and they all perform really badly.


The following results were made with DietPi_v156_Rock64-arm64-(Stretch).img

1-Samsung 32GB EVO Plus Class 10 Micro SDHC with Adapter (MB-MC32GA/AM) --- 95MB/s Read, 20MB/s Write

I got 18.1MB/s Read, 22.9MB/s Write
https://drive.google.com/open?id=18kWYaM...cH0KqzkJ9y


2- Samsung Pro Plus 64GB MicroSDXC Memory Card --- 95MB/s Read, 90MB/s Write

I got 22MB/s Read, 23MB/s Write
https://drive.google.com/open?id=1EVwxQ9...Qvy2iRjwmX


3- Micro Center class 10 32GB 

https://drive.google.com/open?id=1wNf4c7..._PAR4l3-18


4- HardKernel 8GB EMMC

https://drive.google.com/open?id=1byXvrG...36C-KbnYLc



Below the same micro-sd tested with Debian Stretch 64Bit, Kernel 4.9.0-5-amd64 on my Dell XPS 13 9360

1-Samsung 32GB EVO Plus Class 10 Micro SDHC with Adapter (MB-MC32GA/AM) --- 95MB/s Read, 20MB/s Write

I got 24.3MB/s Read, 60.4MB/s Write
https://drive.google.com/open?id=1W5Qkm5...ioDnPP3Sii


2- Samsung Pro Plus 64GB MicroSDXC Memory Card--- 95MB/s Read, 90MB/s Write

I got 62.8MB/s Read, 58.8MB/s Write
https://drive.google.com/open?id=1OToQrw...xtyfx0Gg-i


3- Micro Center class 10 32GB 

I got 4.3MB/s Read, 30.9MB/s Write
https://drive.google.com/open?id=1DfiwGc...x0nrep9Jo5


Could someone perform some tests on their Rock64 using dd with my same options that I used and share their results?

I would love to understand if the problem are the Rock64's Linux distros not yet optimized or is my Rock64's micro-sd controller that is defective.

Thank you in advance.

[xalius]

No surprise there, the SD card interface on Rock64 operates at 3.3V so it only does SD high speed which is limited to 25MB/s. If you want fast flash media use eMMC which runs at 1.8V and supports HS200...

[z4v4l]

No surprise there, the SD card interface on Rock64 operates at 3.3V so it only does SD high speed which is limited to 25MB/s. If you want fast flash media use eMMC which runs at 1.8V and supports HS200...

How do you know this? I mean I also want to know, but the advertised "open-source" Rockchip crappual doesn't have anything SD interface related.
According to the leaked "confidential" rk3399 manual floating on the Internet, SD interface is capable of doing

Up to 832Mbits per second data rate using 4 parallel data lines (SD’s SDR104
mode)

I believe, rk3328 has the same controller.

[dkryder]

i think the rock64 page on the pine64 website has some documentation that includes a technical reference and a rk3328_box_ref_v10_20170224.pdf which on page 4 shows that vcc_io that goes to tf [sd] card is 3.3v and emmc is 1.8v. as zalius said the sd card is locked at 3.3v and can not switch to SDR104. anyway, if you want to know about the rock64 spend some time reading the documents. i have not looked at the 3399 but it could be that it has both 3.3v & 1.8v in ,(or maybe just 1.8V)and therfore can do the SDR104.

[soichiro]

i think the rock64 page on the pine64 website has some documentation that includes a technical reference and a rk3328_box_ref_v10_20170224.pdf which on page 4 shows that vcc_io that goes to tf [sd] card is 3.3v and emmc is 1.8v. as zalius said the sd card is locked at 3.3v and can not switch to SDR104. anyway, if you want to know about the rock64 spend some time reading the documents. i have not looked at the 3399 but it could be that it has both 3.3v & 1.8v in ,(or maybe just 1.8V)and therfore can do the SDR104.

Thank you all for all your inputs and info. I did really appreciate it.
One last thing.... Could any of you please perform some test with dd on their micro-sd card?

So, if the Rock64 does not support SDR104, what is the purpose of adding this micro-sd controller? I know that virtually everyone on this planet has a micro-sd, but still, I believe that is just a waste of space on the PCB if the Rock64 does not support SDR104.

[dkryder]

you might also want to look at page 9 of the box_ref , it shows U800C/U800D as the emmc/tf/sd card line out. sdcard is 4bit/emmc 8 bit. as far as why companies do things the most common is economics. but it looks to me like the emmc/sdcard controller is on the rockchip itself and not a separate chip. so being that 3328 is one of the economy models of the chip line it would not make much sense to have the same speeds as the 3399 which is right at the top.

[z4v4l]

i think the rock64 page on the pine64 website has some documentation that includes a technical reference and a rk3328_box_ref_v10_20170224.pdf which on page 4 shows that vcc_io that goes to tf [sd] card is 3.3v and emmc is 1.8v. as zalius said the sd card is locked at 3.3v and can not switch to SDR104.
anyway, if you want to know about the rock64 spend some time reading the documents. i have not looked at the 3399 but it could be that it has both 3.3v & 1.8v in ,(or maybe just 1.8V)and therfore can do the SDR104.

Yeah, you have not looked, but you are pointing others spend more time reading the documents. As I've said Rockchip didn't provide them.
But for you, to not post BS, spend some time and read the whole quote below. before teaching others, this is from SD specification, page 33. Specifically regarding UHS-I cards OP asked.

After power cycle, card is in 3.3V signaling mode. The first CMD0 selects the bus mode; SD mode or SPI mode. 1.8V signaling mode can be entered only in SD mode. Once the card enters 1.8V signaling mode, the card cannot be switched to SPI mode or 3.3V signaling without power cycle. If the card receives CMD0, card returns to Idle state but still work with SDR12 timing. UHS-I is provided in SD mode but not in SPI mode.
As higher bus speed requires low level signaling, UHS-I adopts 1.8V signaling level for SDR50, DDR50 and SDR104 modes. Still card is supplied with 3.3V by the host and 1.8V signaling level for SDCLK, CMD and DAT[3:0] lines is converted from 3.3V power line. To avoid voltage mismatch between host and card, signaling level is changed by voltage switch sequence at the initialization. The host and card communicate using ACMD41 whether host and card support 1.8V signaling mode. Support of 1.8V signaling both host and card means UHS-I can be used. CMD11 invokes the voltage switch sequence. The card enters UHS-I mode and card input and output timings are changed (SDR12 in default) when the voltage switch sequence is completed successfully. (Refer to Section 4.2.4 for more detail.)
Only 4-bit bus mode is supported in UHS-I except CMD42. If the card is locked, host needs to unlock the card by CMD42 in 1-bit mode and then needs to issue ACMD6 to change 4-bit bus mode. Operating in 1-bit mode is not assured.

For the OP, looks like the driver in linux on Rock64 doesn't do all the work needed to support higher speed modes of your cards. SD controller is not to blame. And yes, it is inside of the SoC.
But again who would say how it is? Who did see RK3328 TRM part 2? I didn't. Maybe the controller there doesn't support UHS-I modes? Ask rockchip.

[dkryder]

i think the rock64 page on the pine64 website has some documentation that includes a technical reference and a rk3328_box_ref_v10_20170224.pdf which on page 4 shows that vcc_io that goes to tf [sd] card is 3.3v and emmc is 1.8v. as zalius said the sd card is locked at 3.3v and can not switch to SDR104.
anyway, if you want to know about the rock64 spend some time reading the documents. i have not looked at the 3399 but it could be that it has both 3.3v & 1.8v in ,(or maybe just 1.8V)and therfore can do the SDR104.

Yeah, you have not looked, but you are pointing others spend more time reading the documents. As I've said Rockchip didn't provide them.
But for you, to not post BS, spend some time and read the whole quote below. before teaching others, this is from SD specification, page 33. Specifically regarding UHS-I cards OP asked.

After power cycle, card is in 3.3V signaling mode. The first CMD0 selects the bus mode; SD mode or SPI mode. 1.8V signaling mode can be entered only in SD mode. Once the card enters 1.8V signaling mode, the card cannot be switched to SPI mode or 3.3V signaling without power cycle. If the card receives CMD0, card returns to Idle state but still work with SDR12 timing. UHS-I is provided in SD mode but not in SPI mode.
As higher bus speed requires low level signaling, UHS-I adopts 1.8V signaling level for SDR50, DDR50 and SDR104 modes. Still card is supplied with 3.3V by the host and 1.8V signaling level for SDCLK, CMD and DAT[3:0] lines is converted from 3.3V power line. To avoid voltage mismatch between host and card, signaling level is changed by voltage switch sequence at the initialization. The host and card communicate using ACMD41 whether host and card support 1.8V signaling mode. Support of 1.8V signaling both host and card means UHS-I can be used. CMD11 invokes the voltage switch sequence. The card enters UHS-I mode and card input and output timings are changed (SDR12 in default) when the voltage switch sequence is completed successfully. (Refer to Section 4.2.4 for more detail.)
Only 4-bit bus mode is supported in UHS-I except CMD42. If the card is locked, host needs to unlock the card by CMD42 in 1-bit mode and then needs to issue ACMD6 to change 4-bit bus mode. Operating in 1-bit mode is not assured.

For the OP, looks like the driver in linux on Rock64 doesn't do all the work needed to support higher speed modes of your cards. SD controller is not to blame. And yes, it is inside of the SoC.
But again who would say how it is? Who did see RK3328 TRM part 2? I didn't. Maybe the controller there doesn't support UHS-I modes? Ask rockchip.

ok, genius. we are all waiting for you to tell us why the rock64 can not do SD104 mode.  and since it seems your comprehension is lacking i said i have not looked at 3399, since i was referring to 3328. if you have some time look at the document i mentioned the rk3328_box_ref_v10_20170224.pdf . it should be obvious to anyone that can follow a line diagram what is going on there. please, tell us, how can the tf /sdcard reach SDR104 over 4 bit line? because unless rockchip messed up the diagram that's what it is. also, there is no buck/step down at the sdcard to drop voltage to 1.8v, so tell us, how does a 3.3 vcc line drop to 1.8. it seems even without the documentation  you bitch that is missing you have it all figured out, so, please do share.

[z4v4l]

I see you like self-embarrassing.

ok, genius. we are all waiting for you to tell us why the rock64 can not do SD104 mode.  and since it seems your comprehension is lacking i said i have not looked at 3399, since i was referring to 3328. if you have some time look at the document i mentioned the rk3328_box_ref_v10_20170224.pdf . it should be obvious to anyone that can follow a line diagram what is going on there.

So far, the only "obvious" thing is your total lack of understanding on subjects you are trying to teach and patronize others.

please, tell us, how can the tf /sdcard reach SDR104 over 4 bit line?

Because this only line is enough to stop messing around you and leave you ashaming yourself alone.
I mean really? Why do you feel the need to teach people when you have NO F&CKING CLUE about what you try to teach? How much lines do you expect? Didn't you know that SD operates with 4-bit (line) data bus at maximum? Jeez. By the way the quote I showed for you and you ignored, mentions 4-bit bus mode several times!

because unless rockchip messed up the diagram that's what it is. also, there is no buck/step down at the sdcard to drop voltage to 1.8v, so tell us, how does a 3.3 vcc line drop to 1.8. it seems even without the documentation  you bitch that is missing you have it all figured out, so, please do share.

"Us" who? Ignoramuses like you? But I don't know how else to show it for "you" that "you" finally got it. I bolded the quote and made it in color and suggested to read the whole quote. Everything is answered there. Still you persist on being a laughingstock. Keep on. Without my help anymore. I remind you, I just asked Xalius how he knows about supported modes for this SoC and you showed up teaching me to read wiki. If you had no knowledge about the question, it would be better for everyone if you just were silent there.

To moderators. This guy seems is OK with self-ashaming, but his poor advices and patronizing makes a bad service for the community. I am almost sure, now the OP has been solidified in his wrong beliefs that this is a SoC/board design flaw, whereas it is not.

[xalius]

Can you please calm down and just have a look at the schematics for Rock64... the SD-Card interface has a fixed 3.3V supply, so it will only support 3.3V modes, the eMMC has a fixed 1.8V supply, so it will support 3.3V and 1.8V modes (the lower speed modes also work on 1.8V). To support dual voltage modes, you need hardware support for dynamic adjustment of VCC_IO which is not trivial to implement for u-boot and Linux which is why you don't see that configuration a lot on single board computers...