IDE numbers

IDE numbers

[Bruce M Beach]

   Hello All

  I'm just sending these numbers because they are interesting. I bought
  a couple of SCSI drives(seagate) a while ago  so I thought I would try
  a simple performance test. I copied  a 9.2Gb file (using G=1024^3) and
  timed it in the following manner:

  1) Partition to Partition(IDE)  hda1 to hda3  # wd drive
  2) Partition to SCSI drive hda2 to sda1       # wd to seagate
  3) SCSI drive to SCSI drive sda1 to sdb1      # seagate to seagate

  and got the following numbers

  time cp TEMP.tar ... hda3->hda1   49m18.320s ~  3,339,200.9 bytes/s
  time cp TEMP.tar ... hdc1->sda1   51m16.493s
  time cp TEMP.tar ... sda1->scb1    5m41.388s ~ 28,936,063   bytes/s

  At first I thought the IDE numbers were a little bit slow but thought
  the transfer must read and write the 9.2Gb so maybe I should double
  the rate and after all the 33 Mhz is just the IDE bus rate. I was suprised
  at the IDE -> SCSI transfer time except the the SCSI transport(adaptec) is
  on another PCI bus and maybe there is significant latency across the
  bridge.  The real surprise was the SCSI -> SCSI and the ultimate result
  of this little experiment was that I bought 2 more SCSI drives.

  I looked around a little bit to see if dmesg had anything to say
  and found the following lines:

  ...
  Real Time Clock Driver v1.10e
  Non-volatile memory driver v1.1
  block: 128 slots per queue, batch=32
  Uniform Multi-Platform E-IDE driver Revision: 6.31
  ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
  PCI_IDE: unknown IDE controller on PCI bus 00 device f9, VID=8086, DID=248b
  PCI: Device 00:1f.1 not available because of resource collisions
  PCI_IDE: chipset revision 2
  PCI_IDE: not 100% native mode: will probe irqs later
      ide0: BM-DMA at 0xffa0-0xffa7, BIOS settings: hda:DMA, hdb:DMA
      ide1: BM-DMA at 0xffa8-0xffaf, BIOS settings: hdc:pio, hdd:pio
  hda: ST360021A, ATA DISK drive
  hdb: 16X10, ATAPI CD/DVD-ROM drive
  ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
  hda: 117231408 sectors (60022 MB) w/2048KiB Cache, CHS=7297/255/63
  Partition check:
   /dev/ide/host0/bus0/target0/lun0: p1 p2 p3
  ...

  What 'PCI: Device 00:1f.1' is I couldn't determine except for a binary
  in /proc/bus/pci/00/1f.1. The PCI slots share resources across
  several bus's, (i.e. Slots 1 & 2 share with the SCSI trasport)
  and the only anomaly is that due to the onboard video having only 4Mb
  ram and there being no AGP slot, there is a PCI video card
  on bus 3.
  Please cc any comments to me.

  Bruce
`

Re: IDE numbers

[Willy Tarreau]

On Thu, Aug 08, 2002 at 12:19:22PM +0000, Bruce M Beach wrote:
 
>   time cp TEMP.tar ... hda3->hda1   49m18.320s ~  3,339,200.9 bytes/s

slow because it's the same disk, so the head constantly move back and forth
between hda1 and hda3. You are limited by seek time.

>   time cp TEMP.tar ... hdc1->sda1   51m16.493s

are you sure it was "hdc1" ? I can't see it in your dmesg below.

>   time cp TEMP.tar ... sda1->scb1    5m41.388s ~ 28,936,063   bytes/s

seems correct to me, I had the same numbers with 7200 RPM seagate drives too.

Regards,
Willy

>   PCI_IDE: not 100% native mode: will probe irqs later
>       ide0: BM-DMA at 0xffa0-0xffa7, BIOS settings: hda:DMA, hdb:DMA
>       ide1: BM-DMA at 0xffa8-0xffaf, BIOS settings: hdc:pio, hdd:pio
>   hda: ST360021A, ATA DISK drive
>   hdb: 16X10, ATAPI CD/DVD-ROM drive
>   ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
>   hda: 117231408 sectors (60022 MB) w/2048KiB Cache, CHS=7297/255/63
>   Partition check:
>    /dev/ide/host0/bus0/target0/lun0: p1 p2 p3

Re: IDE numbers

[Mike Dresser]

>   hda: 117231408 sectors (60022 MB) w/2048KiB Cache, CHS=7297/255/63

what does a hdparm /dev/hda say

and as well, hdparm -Tt /dev/hda?

IDE numbers

[Bruce M Beach]

  Thanks all for your messages

  Willy Tarreau wonders why I have hda1 in the letter and hdc1 in the
  dmesg and suggests that the speed is related to the head movement.
  Mike Dresser wonders what 'hdparm /dev/hda' and 'hdparm -Tt /dev/hda'
  have to say and Marian Jancar wonders from
  'ide0: BM-DMA at 0xffa0-0xffa7, BIOS settings: hda:DMA, hdb:DMA'
  that the kernel doesn't attemp UDMA anyway.

  As for "hda1 and hdc3", I constantly change things around and
  the dmesg that I sent in was made 2 days before the letter. In that time
  hdc1 became hda1. I wouldn't be surprised to see that a transfer from one
  drive to another did not make any speed increase. I have another drive
  and will try it. Now as to what hdparm says, we have:
     1) /dev/hda:
         multcount    =  0 (off)
         I/O support  =  0 (default 16-bit)
         unmaskirq    =  0 (off)
         using_dma    =  0 (off)
         keepsettings =  0 (off)
         nowerr       =  0 (off)
         readonly     =  0 (off)
         readahead    =  8 (on)
         geometry     = 7297/255/63, sectors = 117231408, start = 0
  and
     2) /dev/hda:
        Timing buffer-cache reads:  128 MB in  0.27 seconds =474.07 MB/sec
        Timing buffered disk reads:  64 MB in 21.24 seconds =  3.01 MB/sec
  In the second group we get a salesman's number, a realistic number
  and in the first group there are some interesting entries indicating
  that something may be wrong.
        I/O support  =  0 (default 16-bit)
        using_dma    =  0 (off)
  Now as to why the kernel doesn't use UDMA anyway, I too am surprised
  because I always believed that no matter what was in your machine the
  kernel would always, under all circumstances give an optimal configuration.
  Since the thought pleases me, I'll continue to believe it and attribute
  this case to some obscure third stream phenomena.

  Now a switch from 2.4.18 to 2.4.19 stops my USB mouse from working and also
  gives a new dmesg that gives a clear indication of the IDE problem:

       ICH3: IDE controller on PCI bus 00 dev f9
       ICH3: detected chipset, but driver not compiled in!
  So reconfiguring with:
       Intel PIIXn chipsets support  and
       PIIXn Tuning support
  gives us from dmesg
       ide: Assuming 33MHz system bus speed for PIO modes;
       ICH3: IDE controller on PCI bus 00 dev f9
       PCI: Device 00:1f.1 not available because of resource collisions
       ICH3: (ide_setup_pci_device:) Could not enable device.
 also we have the following

K2.4.18: ide0: BM-DMA at 0xffa0-0xffa7, BIOS settings: hda:DMA, hdb:DMA
K2.4.19: ide0 at 0x1f0-0x1f7,0x3f6 on irq 14

  and

   sh-2.05b# hdparm -d1 /dev/hda1
             /dev/hda1:
              setting using_dma to 1 (on)
              HDIO_SET_DMA failed: Invalid argument
              using_dma    =  0 (off)
    sh-2.05b# hdparm -c1 /dev/hda1
              /dev/hda1:
              setting 32-bit I/O support flag to 1
              HDIO_SET_32BIT failed: Invalid argument
              I/O support  =  0 (default 16-bit)
  hdparm -c1 /dev/hda1 worked before PIIXn support was
  compiled in and would give
       Timing buffered disk reads:  64 MB in 21.24 seconds =  5.51 MB/sec

  Finally some pertinant information on the Machine

       Processor: Dual 2.2Ghz Xeon P4
       Chipset:   Intel E7500
                    MCH:    E7500
                    ICH3-S: 820801CA
                    P64H2:  82870P2
  Regards Bruce