The $99 dollar SSD has arrived

[halfmanhalfamazing]

I am no expert, but I think the problem you may be encountering is that most compact flash do not support UDMA in fixed disk mode.

A compact flash has 3 different modes: I/O, Memory, and True IDE. The last mode is the one that makes a Compact Flash appear to be an IDE fixed disk.

The problem is that the standard states that a CF card must have an True-IDE mode, but it does not state that it will support UDMA in True-IDE. So your extreme card is fast in other modes, but bottlenecked by running in PIO mode when it is simulating a hard disk.

Unfortunately, vendors like Addronics don't seemed to do a good enough job of specifying which device will work in UDMA Fixed Disk mode. So far, the only card I think will work are Industry Compact Flash that states they support UDMA Fixed Disk mode. One such card would be Transcend Industrial CompactFlash, but I don't think there is a 266x or 300x version. In any case, this is just a guess on why you can't go into DMA mode.

Thank you for your reply, I had read a little bit about the three modes of compact flash drives but you laid it out nicely.

I'm pretty sure that that's not directly the problem though. I've tried many distros at this point, and most will not recognize the card at all, some see it and lock up. Suse is one of the few that I've used in the past few months that works properly, albeit with slower transfers than I should be getting. I've been testing out the suse 10.3 betas, I'm pretty sure that my issues are all driver related. It wasn't until beta 2 that YAST would load up properly, and only one of the available 5 ide drivers at boot time would work properly.

As I've posted previously, HDPARM does report this as supporting UDMA mode 3.(I think. it might have been 4)

As far as the adapter I'm using, there's nothing to it. All it is, is essentially a series of short wires that allow the CF to plug into the IDE cable. There's no bridge chips or anything, meaning that there's a direct connection from my IDE motherboard adapter to the CF card. The adapter doesn't technically 'support' anything any more than a standard IDE cable does. It's just wires.

Some of the CF adapters out there do have bridge chips, as they connect to serial ATA.

I think that most of my issues are primarily driver related, and I think that it revolves around those three modes of compact flash. I've been trying to get ahold of an intel based system to do some testing with to see if that makes any difference. One of the links I recently posted, a guy has the same card as me and is getting much higher transfers.

And of course Michael said that a phoronix review is coming at some point, still waiting for that too.

[Thetargos]

It has some nice buffered reads numbers.

[joshuapurcell]

Here is the same command output ran on my old Thinkpad T40:

joshua@laptop:~/work/wrigley$ sudo hdparm -Tt /dev/hda

/dev/hda:
 Timing cached reads:   1644 MB in  2.00 seconds = 822.06 MB/sec
 Timing buffered disk reads:   96 MB in  3.04 seconds =  31.61 MB/sec
joshua@laptop:~/work/wrigley$ sudo hdparm -i /dev/hda

/dev/hda:

 Model=HTS421280H9AT00, FwRev=HA3IA70S, SerialNo=HKA371AKC8BM6M
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
 BuffType=DualPortCache, BuffSize=7528kB, MaxMultSect=16, MultSect=off
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=156301488
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 
 AdvancedPM=yes: mode=0x80 (128) WriteCache=enabled
 Drive conforms to: ATA/ATAPI-7 T13 1532D revision 1:  ATA/ATAPI-2 ATA/ATAPI-3 ATA/ATAPI-4 ATA/ATAPI-5 ATA/ATAPI-6 ATA/ATAPI-7

The drive is nothing special... I believe it's a 4200RPM 60GB drive, which is the default that came with the laptop.

[Thetargos]

What are the hdparm -Tt results?

Currently for one of my ATA disks it is:

hdparm -tT /dev/sda

/dev/sda:
 Timing cached reads:   938 MB in  2.00 seconds = 468.36 MB/sec
 Timing buffered disk reads:  116 MB in  3.01 seconds =  38.51 MB/sec

And it is a Seagate 80Gigs drive:

 hdparm -i /dev/sda

/dev/sda:

 Model=ST380023A                               , FwRev=3.33    , SerialNo=3KB125H9            
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
 BuffType=unknown, BuffSize=2048kB, MaxMultSect=16, MultSect=?16?
 CurCHS=4047/16/255, CurSects=16511760, LBA=yes, LBAsects=156301488
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 udma3 udma4 *udma5 
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: ATA/ATAPI-6 T13 1410D revision 2:  ATA/ATAPI-1 ATA/ATAPI-2 ATA/ATAPI-3 ATA/ATAPI-4 ATA/ATAPI-5 ATA/ATAPI-6

For an ATA-100 drive in UDMA5 mode, I've always thought it was pretty slow... Compared to my other 80Gig S-ATA Seagate:

hdparm -i /dev/sdd

/dev/sdd:

 Model=ST3808110AS                             , FwRev=3.AAH   , SerialNo=            5LR4A3D2
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
 BuffType=unknown, BuffSize=8192kB, MaxMultSect=16, MultSect=?16?
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=156301488
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: Unspecified:  ATA/ATAPI-1 ATA/ATAPI-2 ATA/ATAPI-3 ATA/ATAPI-4 ATA/ATAPI-5 ATA/ATAPI-6 ATA/ATAPI-7

hdparm -tT /dev/sdd

/dev/sdd:
 Timing cached reads:   986 MB in  2.00 seconds = 492.35 MB/sec
 Timing buffered disk reads:  204 MB in  3.01 seconds =  67.85 MB/sec

But from what I can make out of these tests you've run, it would seem that CF has a clear advantage when it comes to overall latency.

[paulsiu]

Probably because most compact flash does support UDMA in True IDE

I am no expert, but I think the problem you may be encountering is that most compact flash do not support UDMA in fixed disk mode.

A compact flash has 3 different modes: I/O, Memory, and True IDE. The last mode is the one that makes a Compact Flash appear to be an IDE fixed disk.

The problem is that the standard states that a CF card must have an True-IDE mode, but it does not state that it will support UDMA in True-IDE. So your extreme card is fast in other modes, but bottlenecked by running in PIO mode when it is simulating a hard disk.

Unfortunately, vendors like Addronics don't seemed to do a good enough job of specifying which device will work in UDMA Fixed Disk mode. So far, the only card I think will work are Industry Compact Flash that states they support UDMA Fixed Disk mode. One such card would be Transcend Industrial CompactFlash, but I don't think there is a 266x or 300x version. In any case, this is just a guess on why you can't go into DMA mode.

[halfmanhalfamazing]

I found this:

http://www.opensubscriber.com/message/[email protected]/6818048.html

in one of the postings it suggests that an IDE cable being too long may affect performance. I'll have to get my hands on a female CF/IDE adapter that will allow me to eliminate that.

http://www.newegg.com/Product/Product.aspx?Item=N82E16822998003

He's using the same card as I am and getting much higher transfers. I'll have to snatch an intel based system and give this a try..............

[Kano]

With a new pc with usb boot support you can directly use USB sticks for linux. Little ones with 1/2 GB are enough to run a Linux livesystem (compressed), bigger ones can hold even a hd install. The install onto usb is a bit tricky, but possible. The sticks that work have usually a 512 byte sector size, one 2 GB stick I have got has 2048 sector size - that does not work. There is a simple check for it:

blockdev --getss /dev/XXX

Older pc can be used to boot from a cd with kernel/initrd stored on it and root fs on usb. And there are much faster usb sticks than 6 mb/s...

[Syzygies]

It looks like we will be doing a comparison with a CF converter at Phoronix shortly. Just waiting on the CF card(s) to arrive.

I'm hoping that CF can improve Linux memory swapping perfomance.

Here are some links re: DMA support:

IDE Compact Flash Adapter with Mounting Plate

http://www.acscontrol.com/Index_ACS.asp?Page=/Pages/Products/CompactFlash/IDE_To_CF_Adapter.htm

Which Compact FLASH cards support DMA?

http://www.acscontrol.com/Index_ACS.asp?Page=/Pages/Products/CompactFlash/IDE_To_CF_Adapter.htm

[halfmanhalfamazing]

I was thinking USB pen drives, till I found this thread.

On the other hand, USB pen drives would save on the conversion
hardware, and wouldn't matter after booting if one then used a RAM
disk.

You're in a whole different territory than I am.

If you're running a cluster, then they would all be a slave to the primary computer, correct?

You'd only need a high performance setup on the primary computer.

I'd probably just go with pendrives, you can get them for 10 dollars each(256mb) off of pricewatch.

Glad I could offer you a possible alternative though. :-)

[Syzygies]

Beowulf cluster?

This is a beautiful thread.

The Intel Desktop Board D201GLY is a powerful, almost complete
computer for $70, aimed at Intel's "next billion customers."

http://www.intel.com/products/motherboard/D201GLY/index.htm

As others have also noted on the web, it cries out to have dozens
assembled into a cheap supercomputer, e.g. a Linux Beowulf cluster.
Just need memory and a drive, per board, supply them all power, and
network them through a switch.

I was thinking USB pen drives, till I found this thread.

On the other hand, USB pen drives would save on the conversion
hardware, and wouldn't matter after booting if one then used a RAM
disk.