iriver H340 with 240GB drive - success!

[pwhodges]

After muttering from time to time about the possibility of this upgrade, I decided to go for it.  Here are details of what I did (but without the failed experiments).

I bought a Toshiba MK2431GAH 240GB drive from rapidrepair.com, ignoring the dire warnings about it only ever working in a particular iPod; it arrived very quickly indeed (4.5 days from order, across the Atlantic), and the courier didn't collect VAT (17.5% tax, usually with the addition of a handling charge from the courier) off me either!

I used the only ZIF adapter that fits in the H300 series (because there is no leeway at all in the disk length) - this can be got through this forum, see http://forums.rockbox.org/index.php?topic=21676.msg157028#msg157028.

I have acquired a collection of updated H300 bootloaders that various people have published briefly; the one I used is r17607M-080522, but I now have no idea who made it (you can download it here, at your own risk, etc, etc).  Of course, I updated installed the bootloader before taking the old disk out!  (Later comment) On some H300s this firmware hangs during boot.  I have one on which this never happens, and another on which it happens about one time in five; I just reset and try again.

I followed the instructions for compiling Rockbox using cygwin, which got me a working compilation in under an hour from starting.  I then added:

Code: [Select]

#define HAVE_LBA48
#define MAX_PHYS_SECTOR_SIZE 4096

to the file rockbox\firmware\export\config\iriverh300.h to make a version for the 240GB disk, and compiled that with no problem. (Later comment) This is not necessary since v3.6; these changes are now included by default.

With the bootloader updated installed and the new disk installed, the H300 went into bootloader USB mode without a problem.  Windows XP saw it as the 240GB disk it is, and the Disk Manager correctly initialised and partitioned it.  I used my favourite disk manipulation program, DFSee (not free - sorry), to format the partition in FAT32, as Windows won't format a disk of that size as FAT32. (People often recommend the free Swissknife for this, but I have only ever had it crash, on multiple systems.)

I then, without even rebooting, copied my newly compiled .rockbox folder and a few music files onto the new disk for testing.  The upgraded system rebooted as if nothing had happened, and worked fine.  It is now copying my entire music library across, and I will report when this gets past the 137GB (128GiB) boundary... (possibly sometime tomorrow!).

[edit:] This disk is loading as slowly as the Samsung 120GB I have, which took over 24 hours to fill.  I had hoped that setting the MAX_PHYS_SECTOR_SIZE parameter would help, but it seems not.  Maybe I should play with the MAX_LOG_SECTOR_SIZE parameter as well?  (I am loading using the main firmware, not bootloader USB mode, so it is my compiled version that is in use.)

Paul

[gevaerts]

the H300 uses a hardware USB-ATA bridge, so rockbox can't do much about USB speed. However, have you tried building without defining MAX_PHYS_SECTOR_SIZE? In theory that one shouldn't be needed, and might even slow things down (in rockbox itself).

[pwhodges]

Can you explain what MAX_PHYS_SECTOR_SIZE (and MAX_LOG_SECTOR_SIZE) are expected to do, then?

The reason I tried MAX_PHYS_SECTOR_SIZE is that I have a Toshiba 120GB drive and a Samsung 120GB drive, and the Samsung is much slower than the Toshiba 120GB, just as the 240GB Toshiba is turning out to be (it looks like it'll take nearly three days to load everything as it stands).  So the speed limitation is in the disk access, not the USB speed; I haven't measured it, but it's slower by a factor of 3 or 4 I reckon.  The common thing about the Samsung and the bigger Toshiba is that they both have 4kB physical sector size, so I hoped that code that recognised that might be more efficient.

I am willing to experiment a bit if anyone has any suggestions - but I'll probably wait until my music library has finished copying first...  It would be nice to get it going just right, especially as this disk is probably the end of the line for (P)ATA (all Toshiba's newer disks are SATA).

Paul

(later)
Or does having a hardware USB-ATA bridge mean that disk writing via USB doesn't actually use the rockbox disk driver at all, so we can't influence it?

[GodEater]

Or does having a hardware USB-ATA bridge mean that disk writing via USB doesn't actually use the rockbox disk driver at all?

Correct.

[gevaerts]

Can you explain what MAX_PHYS_SECTOR_SIZE (and MAX_LOG_SECTOR_SIZE) are expected to do, then?

MAX_PHYS_SECTOR_SIZE enables rockbox to use physical sectors of more than 512 bytes for transferring data to and from the disk. Most modern large disks natively use 4K sectors, but nearly all of them also have a translation layer that allows them to work with traditional 512 byte sectors. Usually, the translation layer in the disk firmware is more efficient than the large sector support in rockbox, so if the disk supports it, it's often better to not use MAX_PHYS_SECTOR_SIZE.

For USB performance, you might want to have a look at http://www.rockbox.org/wiki/UDMAonUSB

[pwhodges]

Thanks for the UDMA pointer!  Enabling that increases the read transfer rate from 13MB/s to 36MB/s burst (not achieving the drive's specified 100MB/s burst capability).  It maintains around 32MB/s for 2/3 of the disk, trailing off to 20MB/s over the last 1/3 of the disk, so it appears to be disk limited towards the end at least).

I will now see if I can manage to write that configuration to the flash; no big deal if not, though.

(later)  I have now done a write benchmark using the Samsung HS12YHA 120GB disk.  Sadly, the results show that UDMA will only have a marginal effect on writing.  The write benchmark I used (HDTune) shows that three out of four writes (at least, at the intervals tested) are at a mere 1.8MB/s, with only one of four at the full rate (13MB/s for PIO and 26MB/s for UDMA.  So it appears that we're stuck with the slow writes (for comparison, a Toshiba 30GB disk writes uniformly at 3.6GB/s 3.6MB/s)

Paul

[evilnick]

Can I get one of those Toshiba 30GB disks if they really do write at 3.6GB/s? 

[pwhodges]

  I even edited the mistake - but I must have forgotten to save it!

(later)

So now I've found a workaround for the slow writing!  My disk is currently writing at the full PIO speed (I can't enable UDMA using this machine, as the programs don't appear to work on 64-bit Windows XP).  All I did was use device manager to enable write caching for the iriver drive.  I guess the caching changes the way stuff is buffered into the disk somewhere. 

The copying is now going as fast as I originally expected it to (it will be finished in a couple more hours, I reckon).  While this is going on, Windows Explorer is suffering severely, with lots of delays and long hang-ups - but as soon as the big copy is finished I'll turn the write caching off again anyway, as it would be mad to leave it enabled for the removable disk.  Anyway, I get the feeling that the best has been achieved now that can be with this hardware.

Paul

[Serenity]

Sorry for the bump, but would your procedure work on the MK1231GAL drive that's only supposed to work on the iPod Classic?  Or is that wishful thinking?

As much as I want it to, the MK2431GAH won't fit in my H320 unless I get the deeper backplate from an H340, but then it won't fit in any of my cases...and so it goes.

[pwhodges]

I expect it will work.  The vendor I bought the 240GB from has warnings all over that it cannot be used anywhere other than an iPod, and that's clearly wrong. 

(I suppose I should allow the possibility that they updated the drive firmware, and it was true for an earlier version - but I have no evidence of this.)

Paul

[TexasRockbox]

The iPod Classic uses a CE-ATA drive.  the MK2431GAH utilizes ATA-4.

[pwhodges]

True; thanks for the correction. That'll be why they say it only works in the iPod Video, then: http://www.rapidrepair.com/shop/3119-hard-drive-disk-mk2431gah.html.  They also say it doesn't work in a PC or other players, which is simply wrong - which was my point.

Paul

[sordup]

Not sure if anyone is checking this thread anymore but:

I've got some questions about this upgrade. I have just made the mistake of purchasing a Toshiba MK1231GAL 120gb drive which definitely is not compatible (not even, apparently, with Windows). Can't even initialize the disk in Windows disk manager. I don't quite understand why previous drives from Toshiba worked fine, and you say this later drive also works, and yet the 1231 will not work. Strange to me, but, anyway...

Has the process outlined in the above post continued to be successful for your H340? Is there anything not mentioned in the post that would be necessary to get that same drive up and running for my own H340?
And, is it still necessary to recompile the Rockbox bootloader for the large drive? I read that Rockbox ver. 3.6 is "large size compatible", so to speak.

Thanks for your time.

[Ste-]

I used this drive yesterday in my iAudio X5, and the process was simple connect the drive to the converter, put it in the device. Connect the device, I partitioned it using gparted for a full fat32 partition. and it worked. currently filling the device now and now problems.

Big drives are supported a long time now. Thanks to the work of torne I think it was.

[torne]

3.6 supports large drives, but not every target's bootloader has had a new release since then. The h100/h300 bootloader hasn't been updated in a while because the current code in svn appears to brick players, and I don't think that has been fixed yet, so while the h100/h300 builds of Rockbox will handle a large drive just fine, I don't think the bootloader will. Compiling your own bootloader is not recommended because it's known to be broken and there is no way to recover a bad bootloader install without special hardware. There are some old bootloaders around which have been patched with large drive support which predate whatever happened to break it, though, like the one mentioned by the original poster in this thread.