My Kingston Class 6 8GB SDHC card has worked fine in the internal slot for months now, though it did have some problems in the external one.

I just purchased a new 4GB Sandisk (class 2) SDHC card and wanted to share some interesting test results... As someone might remember I do have two tablets and I also have 8GB Patriot Class 4 card that I purchased a while back. I did some comparison testing of write and read speed on them on different slots, different firmware (patched and non-patched) and in different N800 units.

First of all, it seem "Sandisk feature" that it does not want to operate in high speed (48MHz) mode. It reverts back to 24MHz bus speed on my patched N800. While this was slightly disappointing... it doesn't appear to be as bad as it sounds... read on...

I didn't use the dd or any other "built in' speed test, but I used a 319MB video file in my PC that I wrote to N800 and read back to PC using USB cable. I timed the time to transfer with a stopwatch. Both cards were formatted to a single FAT32 partition using 2k block size.

Test results with N800 having patched kernel:
Card: Ext/Int: Write_time: Read_time:
8GB E 188s 25s
4GB I 132s 43s
8GB I 185s 25s
4GB E 135s 42s

And here are the results of N800 having non-patched kernel. (The firmware is the one that is planned to be released next Tuesday... )
Card: Ext/Int: Write_time: Read_time:
8GB E 195s 40s
4GB I 132s 42s

I didn't swap the cards in the slots on the second unit, since it didn't make any difference in the results (excluding error margin) on the first unit... Just being lazy there...

So, the findings. Surprisingly the Sandisk Class 2 card outperform the Class 4 Patriot card in write speed!!! Based on the dmesg logs it is running at 24MHz speed. So, in other words, the bus clock speed is not that important as I was thinking it was... Likely the memory card timings play a big(ger) role.

The high speed mode seem to help only on read speed.

And, as if you interpreted the results of my second tablet, you can make educated guess if the new firmware will support high speed mode...

Oh, and last, both cards have identical write speed (60 seconds to write 319MB file) when used with an external SDHC USB reader/writer. So the N800 is still very slow when compared to a dedicated reader/writer... Anyone planning to do a lot of file transfers to/from your memory cards, I highly recommend to purchase external reader/writer.

UPDATE:
I just rerun the test when using external USB reader. It seems that the memory card speed gets slower when there is more data in it? After some 400MB worth of MaemoMapper files are stored on the card, the write took 66 seconds, and the read back took 31 seconds. The same test on the Patriot 8GB Class4 card (having approximately the same amount of data) took 60 seconds to write and 20 seconds to read.

Hmm... is the Oct 2nd update available somewhere for early testing?

Milhouse fanoush
I see, control values in registers are really differ to acheive same result.Really asymmetric prop's.Amazing! :\

penguinbait
One moment, but with SDHC problem is in 2 parts.One is SDHC in Linux kernel is not ideal.Other is that cards are capable of dying just because power dropped.This is against SDHC specs so card manufacturers are IMHO responsible for releasing flawed products here as well.Of course everyone wants to be first on market.As for me, looks like Transcend and A-Data did not tested their cards well enough under different circumstances, especially taking into account that real world differs from ideal lab conditions.Or why Kingston does NOT dies?I'm played with my Kingston SDHC 8Gb Class6 a bit and still not able to kill it.Works ok for me.Imho if card can be killed by methods which are within spec it is hard to trust such card at all.But about QA I'm agree, it should be more powerful.Nokia IT OS 2007 3.2 (without skype) is much more stable than IT OS 2007 4.x (with skype).Surely, there is lots of room for QA job.I wonder why Nokia does it so poorly.

It's obvious that you guys have never worked in a QA department before.

Actually I was wrong with details about 2.8 voltage, below is the code. While external slot indeed is more fine grained and has wider range, for 2.8-3.0V range it should produce same voltage. I'm not sure what voltage SDHC cards request but you can see it in kernel log when inserting the card.

Exactly. Cards should not die no matter what. This is bad design. They should simply disconnect from the bus or return error but not become permanently damaged. If this is specific to some brands it may be that all such cards use same SDHC controller chip inside with some buggy hardware or firmware. On the other hand, powering off the slot in the middle of write when going to power saving mode is not good idea either (if this is what Nokia firmware does). Let's wait for updated kernel source. I'm quite interested where the bug is and how they fixed it.

Indeed. I wrote a very lengthy, very detailed and very accurate explanation of how product QA works a while back here (it is also duplicated on maemopeople.org). It seems a few could benefit from reading it.

In addition, QA is a broad term. To paint all of Nokia QA badly with a broad brush is disingenuous. If one segment does not work, that does not automatically impugn others.

Finally, the point made a few posts above about the complexity of this issue is highly relevant. I made the prediction very early on that this would likely turn out to be a combination of causal factors. I take no satisfaction in being correct, but that fact remains: no blame can be assessed to one camp-- especially given the diversity of experiences recounted (I have YET to have problems with any card over 32 meg, and there were mitigating factors with that particular episode).

The fact that I never had this problem in the six months or so since I got my tablet shows that this kind of issue would have been difficult for your typical overworked, underappreciated QA department to find in a month or two. (I have no idea how long phones are in QA at Nokia for, just a guess.)

I'm with texrat and zerojay on this.

As for QA finding all kind or problems in advance.. that won't happen _anywhere_. Where I work we suddenly started to get problem reports on something that had worked flawlessly for more than half a decade on a huge number of installations. Turned out there was an old bug deep down there. However, nothing ever triggered it before. Ever. It must have been tested and verified ok literally millions of times, under varying conditions and on lots of platforms. Then suddenly.. boom. At least in our case it was easy to find then, but that's not always the case. The most difficult situation is when hardware, software and 3party equipment/software interoperation is involved.

Really? Actually, QA is my prophecy . And as QA I have to admit that firmware 4.2007 (this one with skype) has slightly worse quality than "default" firmware (version 3.200 IIRC). Sorry but that's true, at least it's my experience. I also do not know if Nokia extensively tested SDHC support but if QA processes adjusted properly, this bug not seems to be very hard to catch in test environment. Just buy bunch of cards from different vendors and play with your n800 devices and these cards a bit. Actually I'm surprised such big corporation has failed to perform proper testing.

Also as QA I can admit at least the following...

- Firmware with Skype detects my fingers poorly, slightly worse than 3.200 did. I'm unable to set up sensitivity to level where it detects my fingers as good as 3.200 did. Attempted to use settings slider, it did become better but still worse than in older firmware. But Nokia calls this improved screen sensitivity. Huh?

- Language swithing in on-screen keyboard is a total crap! If you're not US\UK citizen you'll have hard times. You have to use MENU to switch layout. And this menu is quite hard to navigate with fingers, even in full screen keyboard. Switching between Russian and English is a real pain in the ***** and you have to do this painful process FREQUENTLY if you're speaking two languages. I'm personally NEVER seen worse layout switching implementation... argh, wait a bit, Nokia mobiles use same ***** switching method. It is interesting if someone from Nokia corp writes in 2 different languages. Looks like it is not...

- Russian full screen keyboard layout is far from perfect. What a ***** idea to move some chars from main keyboard to extra tab! Hey guys, come on: move some chars from YOUR native language to additional keyboard tab, then try to use it. You will get the idea how stupid you were with this step. Annoying!

- Sound is sometimes getting interrupted under load. Looks like sometimes sound subsystem can't handle data in a REAL-time fashion. This even affects boot-up sound.

- Battery indicator sometimes performs strange in recent firmware with Skype. After heavy load it can immediately drop charge from "full" to "almost empty". You thinked battery full? Nope! Surprise! :P.

- Also, looks like low battery measurements somehow ineffective. Tablet can be able to boot into the OS and then ... immediately getting powered off (due to low battery). Without preforming proper shutdown sequence...

- Worse, NOLO boot loader not seems to check battery charge before proceeding with flashing. So it is up to user to ensure that battery full before flashing device. Actually boot loader should at least warn software that battery is too low or maybe even refuse flashing operation (at least without special confirmation). Not looks like this implemented. Also probably it is a NOLO loader who should refuse to boot OS if it detects battery low condition. But while NOLO part is poorly crafted it's also proprietary so nobody can fix this stupidity .

- Also NOLO does not allows to read flash image backup so if user does not likes update result he has to reinstall all from scratch anyway. Backups are not an answer. On PC you can use disk imaging software to backup your HDD to image file and once smth bad happens, return to working state with one-shot process.With n800 this is not an option.You can't dump kernel\file system backup copy easily. Especially if device no longer boots.

- Sometimes there is strange issues with tray area. Icons behave strangely a bit. For example connection indicator may display something but not a real connction status. Very rare issue, seems to occur when changing several connections in short time.

- Sometimes when you're using alarm clock tablet may occasionally hang on alarm event. I was not able to reproduce this well enough but seen couple of times.

- Built in player is slow as jerk in decoding and handles very few video formats. Mplayer is a way faster in video decoding and eats much more videos. Are there worse players on market? Nokia's players are crap, both in phones and n800 as well.

- And no, you can't just "File -> Open" in player and play desired video file. You have to use File Manager instead (surprise, yep, video player can't open video on it's own). What a ***** logic. Nokia's player imho one of the worst I ever seen.

- File manager may take a while to navigate on card with bug directory tree with lots of files.Imagine 20 folders, each with 500 photos, some subdirs, etc. Now try to browse it. What? It's so s-l-o-w in file manager!

- The same but executed over bluetooth to the phone's cars is SO slow and unstable that bluetooth link can die while "Updating..." popup appears for a minute or so. Effectively kicking you from device before you can really browse device's files. So sometimes you'll be unable to browse such directory tree over bluetooth at all.

- When you're about to view info on memory use on such card via control panel, info dialog uses CPU and RAM extensively while scanning card. Then all crashes to the hell due to out of RAM. At very best only according app crashes. But due to global RAM outage sometimes whole system may become unstable...

- Wi-Fi AD Hoc mode setup is hidden so nice in connections wizard that virtually no users can complete Wi-Fi Ad-Hoc setup without some external hints. Lots of users thinks n800 is unable to use Ad-Hoc mode at all

- Some mechanical issues. Joystick is not the best thing ever. Try to play Doom if you do not think so. So you'll have chances to change your mind . Also external SD slot cover seems to be unreliable. I think it is possible to make metallic cover for decent 400 euros. But huh, cheap plastic instead in 400 euros priced device. Also, I wonder what a ***** has decided to hide USB jack so it can't be accessed easily? Pretty uncomfortable. And what? No USB charging? Phew, thank you, hardware designers! Ever tried to remove battery? No, do not use screwdriver, scissors or something. Try to do this with bare fingers, yep? Got some fun, no? .Also buttons like [+][Fullscreen][-] and others are far from 3.5 mm jack. Cool. Ever tried to switch track in player or adjust volume with these buttons? Pretty uncomfortable since they are far from 3.5 mm jack so you can't adjust this in the pocket, duh. You have to take device to the hands.

- There is lots of other funny issues as well

Argh, no panic, that's just small n800 ***-kicking from honest QA's view. If you think n800 suxx, relax. I can find something funny in [your favorite device or program here] as well, so it will look even worse for public

P.S. as for me: I'm personally would be VERY HAPPY to work for Nokia as full time QA employee on their Linux tablets to improve their quality if they will ever need some extra QA crew for Linux internet tablets (from my opinition, they do but their opinition can differ, he-he). Virtually I'm almost dreaming about this since I'm tired from working on proprietary projects and really preferring to spend my time working on open projects instead. While I'm used DOS and Windows for more than 10 years, I did occasionally discovered that I'm really prefer Linux for myself and that I do like open source apps and open source ideas. But sadly it looks like it is not an option here in Russia to apply and I'm not willing to move anywhere else, sorry. Why full time is better than just filing bugs on maemo? Because of free time issue as anyone can easily guess. As for me it is somewhat sad to spent lots of time on proprietary projects if there is chance to work on open ones so much.