32 bit transfer mode:

Discussion in 'Acronis True Image Product Line' started by tronic592001, Feb 3, 2007.

Thread Status:
Not open for further replies.
  1. tronic592001

    tronic592001 Registered Member

    Joined:
    Feb 23, 2005
    Posts:
    164
    Location:
    Wales uk
    I notice in my bios the 32 bit transfer mode is disabled, ive read on a few links that it sood be enabled.

    Will enabling this setting effect my acronis backups in anyway o_O


    :ninja:
     
  2. seekforever

    seekforever Registered Member

    Joined:
    Oct 31, 2005
    Posts:
    4,751
    It shouldn't unless it makes your machine unstable for some reason. I think this is a good reason for a test!

    Create an image and validate it and measure the time. Now change the setting to 32 bit transfer in BIOS and do the same thing again and measure the time. If the 32 bit transfer has any effect it should show up since the backup/valdiate processes are likely throttled by disk speeds if you have a reasonably fast processor. You have also now tested that it does or does not affect TI as well.

    I remember playing with this on a PC and came to the conclusion it made no discernable difference in disk speed but I don't think I flogged it as much as TI would.
     
  3. tronic592001

    tronic592001 Registered Member

    Joined:
    Feb 23, 2005
    Posts:
    164
    Location:
    Wales uk
    seekforever thanks for the reply:

    No problem works fine think its a bit faster but need more time to experiment...

    :D
     
  4. seekforever

    seekforever Registered Member

    Joined:
    Oct 31, 2005
    Posts:
    4,751
    You can probably dig the times out of the TI log if that is what you used. I have a small digital timer that I use to measure such things should I be inclined.

    I was reading somewhere that most people can't tell if a program is running faster or slower unless the difference is around 20%. People who often run the test programs are able to tell the difference if it is between 5 and 10%. Take the numbers with a grain of salt due to on my personal memory performance but the message is that small differences don't matter that much.
     
  5. tronic592001

    tronic592001 Registered Member

    Joined:
    Feb 23, 2005
    Posts:
    164
    Location:
    Wales uk
    So i gather it takes some of the load off the CPU?
     
  6. seekforever

    seekforever Registered Member

    Joined:
    Oct 31, 2005
    Posts:
    4,751
    Primarily should take some load off the PCI bus since it is now carrying its full 32 bit data on each transfer rather than only 16 bits.

    My P5LD2 has a 32/16 bit entry in BIOS. It has a SATA2 drive in it, no IDEs. I ran some tests under Windows which has AVG anti-virus, Sygate Firewall, Windows Defender, and is connected to a LAN.

    I validated a TI image for the test and the results are interesting. The PC was obviously rebooted before each test and the only program running, apart from the background stuff mentioned above, was each test program either TI9 or HDtach alone. All TI times measured with stopwatch.

    TI9 validate:
    16 bit: 1min 15sec
    32 bit: 1min 0 sec Better!

    32 bit HDtach: Burst 169.1MB/s Sequential Read 65 dropping to 35. 16 bit not done at this time.

    Reset BIOS to 16 bit:
    TI9 validate:
    16 bit: 55 seconds Wow, now it's better than 32 bit
    16 bit HDtach: Burst 166 Sequential read 65 to 35

    TI9 Validate again: 1:07 and again, 1:09 - slowed a bit!

    Reset BIOS to 32 bit:
    32 bit: 1:17 dragging its butt a bit
    32 bit HDtach: Burst 166.9 Sequential read 65 to 35
    CPU utilization 2% +/- 2% 55.4MB/s read (didn't record before)
    TI9 Validate again: 1:12 and again, 1:09 - bit faster

    Reset BIOS to 16 bit: (to get CPU utilizataion and read)
    16 bit TI validate 1:21 and again 1:14
    16 bit HDtach: Burst 165.8 Sequential read 65 to 35
    CPU utilization 2% +/- 2% 55.5MB/s read

    These numbers indicate that for this machine running XP setting the BIOS to 32 bit doesn't hurt but there are other factors that swamp any theoretical benefit. Disk caching could be affecting the numbers but I would think the first test after a boot would not see the caching effect.

    A real "test the 32/16 bit function" would probably be better run using the TI boot CD version to eliminate the effects of Windows which I presume would be more than for the bare Linux OS but my machine doesn't run that way 99.9% of the time.

    I would run the machine with the 32-bit enabled since you will feel better and it doesn't hurt.

    If anybody has some insight as to what exactly is happening I would be interested in hearing it. I first did a similar test many months ago because of variations in the speed numbers given by the XCSC checksum calculator which provides a read/calculate function to a TI validate.
     
Thread Status:
Not open for further replies.