CF-IDE HDD on A1200 - only 2 (out of 4) partitions visible!

[mingle]

Hi,

After mucking around with the 4GB CF card in the PCMCIA adapter (and failing to get it to be recognised in my A1200)

I've now installed it inside my machine using a CF-IDE adapter.

I booted from my original WB3.1 install floppy and used HDToolBox to partition the 4GB CF card as follows:

DH0: 150MB
DH1: 1000MB
DH2: 1000MB
DH3: 1000MB
DH4: 620MB

For each of the partitions I didn't change any of the parameters from the defaults, apart from setting the 'directory cache' flag.

Then I save the changes to the drive and rebooted from the WB3.1 Install floppy, but only the DH0: and DH1: partitions were visible from Workbench.

I did a clean install of WB3.1 and still no sign of the missing partitions.

When I looked in HDToolBox again, all 5 partitions appear to be there. When I used the "format" util (in the Workbench:System drawer) all 5 partitions show up, but when I try to format them I get a "no disk in drive" error.

Do I need to tinkering with any of teh advanced setting in HDToolBox (like maxtransfer and mask?) or could it be that my CF card is just not compatible - seems unlikely since it seems to run fine with the other two partitions...

Any help or comments appreciated!

Cheers,

Mike.

[MozzerFan]

Try to partition/format it with HDInstTool.

If that doesn't work try to partition/format it under WinUAE.

BTW IIRC Directory Caching should only be used with floppy disks. Somebody please correct me if I'm wrong.

[mingle]

Hi Lloyd,

Once again, many thanks...

HDInstTools worked and all of my partitions are now visible!

The only problem I seem to have now is that the CF card is slower than the old 2.5" 4GB IBM Travelstar hard drive!

The card is a cheapy, but it's supposedly 'high-speed'. Anyway, I would expect that any CF card would be faster than an old HDD...

I did a quick test, copying the contents of my C drawer (95 files) from the Workbench partition to an empty partition...

Using the 4GB CF card it took 60 seconds to copy the 95 files.

Using the old hard disk it only took 16 seconds!

I tried changing the maxtransfer for all of the CF partitions to xFFFFFFF, but that completely mucked up the card, and I had to reformat it in my PC and start again (a minor problem).

The only difference I can see between the two drives is that the HDD was formatted "FFS Directory Cache", but the CF card was only "FFS International".

One other thing was that when I used HDInstTools, there was no file-system present on the CF card (in the RDB?). Do I need anything there - didn't seem to cause any problems when I was partitioning the drive though...

Any other way I can speed up the CF card using some of the setting when partitioning?

Cheers,

Mike.

[MozzerFan]

How many buffers are you using on your CF partitions?
You can try to increase the value to 100 buffers and see if it makes a difference.
If possible, check how many buffers your HD partitions were using, and use the same value for your CF.
Don't set them too high, or else it uses up your RAM.

You can try a max transfer setting of *edit:* 0x0001FE00 Report back if you have data corruption when you copy large files.

With the *edit* filesystems button of HDintsTool you can check if fastfilesystem is in the RDB.

[mingle]

Hi Lloyd,

Both the CF and HDD are set up with 30 buffers...

I did a bit of googling and found mention (on EAB) of setting the max transfer to 0x1FE00000, but that made no difference.

The max transfer you suggest is lower than that - is your's a better value?

I also tried setting interleaves (in the drive setup options) to "1" (was "0") which made no difference, but I was just clutching at straws!

SYSINFO shows 1.57MB/sec reads for the CF card and 2.8MB/sec for the IBM HDD...

I managed to put FFS on the RDB. I also had a quick go of SFS (but didn't set the optimum max transfer for it) and that boosted read/write times to be the same as the old HDD, but I'm a bit reluctant to use SFS, since it has a bit of a reputation for REALLY buggering-up partitions if you're not careful!

Finally, is there any benefit to using directory cache on a HDD/CF - it was mentioned that this is only really for floppy drives? I tried that too, but made no difference...

Cheers,

Mike.

[MozzerFan]

First of all, max transfer settings is a size. It has nothing to do with transfer speeds. It limits how much data can be transferred to or from the hard drive at one time.

The max transfer you suggest is lower than that - is your's a better value?

I don't know if it's better. It works for me. It's the default value that hdtoolbox from amigaforever uses when partitioning a harddrive.

As for the speed of the CF. It could also be the CF-IDE adapter. There was a guy here at amiga.org who had a cheap CF-IDE adapter. He also got lower speeds than others.
Or maybe the problem is the CF card. If possible, try to test it with other CF-cards.

As for directory-cache: I always understood that it was not necessary to use that option for harddrives. Floppy drives are really slow, so caching directories can really speed up things.

[zipper]

As for directory-cache: I always understood that it was not necessary to use that option for harddrives. Floppy drives are really slow, so caching directories can really speed up things.

I got frequent read errors as I once installed FFS-DC by mistake onto my CSPPC/SCSI3 system. And speed was miserable.

[mingle]

Hi,

No luck with any of the maxtransfer settings. The speed is still the same...

I've tried other CF cards (8MB & 512MB) with the same result.

So it could be the CF adaptor? It was a cheapy, but I've seen the same used a few times:



Can you suggest a particular model that will work better?

Cheers,

Mike.

[MozzerFan]

I made a mistake in my previous post concerning the max transfer size. I use 0x0001FE00. It's the default value of HDToolBox from Amigaforever. I corrected it in my previous post.
But like I said, entering a higher value doesn't boost the transfer speed of the drive. It's just that some harddrives have a limit of how much data they can move at one time. If the maxtransfer size is set too high, this can result in corrupted data when copying large files.

I don't know if your cf-ide adapter really is to blame.
Right now, I have a CF-IDE adapter in both my A600 and a4000, but it wouldn't be fair to compare your transfer speed to the transfer speed of those systems, because in my a4000 the adapter is connected to my cyberscsi, and my a600 only has a chipmem expansion and no fastram or accelerator.

[MozzerFan]

BTW What is the blocksize you are using on your CF?
You could try a blocksize of 512. Also try to see if changing the buffers from 30 to 100 changes things.

[MozzerFan]

I put on my eye-glasses again and read that SFS was boosting your performance.

but I'm a bit reluctant to use SFS, since it has a bit of a reputation for REALLY buggering-up partitions if you're not careful!

I'm using sfs since many years. I got sick and tired of my harddisk partitions validating when my system had crashed during a writing process.
I must say I never looked back since.
SFS never failed me. If my system crashes there are no more long validation processes and it's faster than FFS.
Since installing SFS, no partition ever got messed up, and if one ever does, there's always SFSSalv (if for some reason I forgot to make a backup).

[mingle]

Hi Lloyd,

The block size I've been using is 512bytes (seems to be the default in HDInstTools).

I've been using 0x0001FFFE for max transfer on with both FFS and SFS... It seems to be fractionally (about 2%) quicker than when I use 0x0001FE00 - wow! :-)

With regards to SFS, I've never had much luck, although I'm using it now on the CF card. I was getting random lock-ups of the system and the SFS format command seems to crash quite a lot... Maybe it's just me?

Having said that, it's much faster (on the CF card) than FFS...

My setup is a KS3.1 A1200 with Blizzard 1230 IV (50MHz '030) and 32MB of FAST RAM...

I'd be interested if you could post some timings for you CF card setups... Maybe give SysInfo a go, although I think that uses some sort of raw reading of data, as there was no change at all when I used FFS or SFS...

Also, what type of CF-IDE adaptor are you using?

Cheers,

Mike.

[MozzerFan]

With regards to SFS, I've never had much luck, although I'm using it now on the CF card. I was getting random lock-ups of the system and the SFS format command seems to crash quite a lot... Maybe it's just me?

When I use the SFSFormat command, it just asks me to press enter, but when the format is finished, it never says so.
I notice it's finished, because the volumelabel of my disk-icon changes and the harddisk-led stops.

 

I'd be interested if you could post some timings for you CF card setups

On my a4000/060 with cyberscsi and scsi-ide adapter I get ~7.5 MB/s. I have devided the Sandisk Ultra II 4GB CF in 4 partitions and on the last one I get ~8 MB/s.
When I had a harddisk in it, I got ~5.7 - 6 MB/s.

I can't use SFS on my A600, because it needs a 020 processor or higher. I use FFS on my A600 and get speeds way under 1 MB/s. But like I said, it only has a 68000 processor and no fastram. I also use a fairly old Toshiba 128 MB CF card.

Also, what type of CF-IDE adaptor are you using?

I'm using this one:


I bought it at Conrad.com. It's rather expensive (though shipping was free), probably much more than I should have paid for a CF-IDE, but I wanted to play it safe. Some users who bought one via Ebay had problems with the Led not working or other problems. Soldering is not one of my skills.

[stanlong]

its probably the brand of cf card your using
i had many probs with three brands and switched to
sansdisc probs evaporated

[Thomas]

I've been using 0x0001FFFE for max transfer on with both FFS and SFS... It seems to be fractionally (about 2%) quicker than when I use 0x0001FE00 - wow! :-)

There must be something else which causes the speed difference, because both values result in the same amount of data being transferred in one chunk. Data is always transferred in units of blocks. One block has 512 bytes (0x200). If you divide either MaxTransfer value by 512 bytes, the result is the same in both cases: 255 blocks per chunk. The remainder is dropped.

Bye,
Thomas