Loading ...
Sorry, an error occurred while loading the content.

Re: [linux-dell-laptops] Re: where to put $$$ to beef up performance

Expand Messages
  • Alexandre Tessier
    ... I agree with you and David but all the unused ram is used for disk cache, thus if you have a lot of ram you will not need to read on the disk (except the
    Message 1 of 12 , Dec 1, 2004
    • 0 Attachment
      Dan Christensen wrote:
      > On Tue, 30 Nov 2004 09:20:34 -0500 (EST), "Robert P. J. Day" <rpjday@...> said:
      >
      >
      >> i'm looking at getting a new 8600 or 9200, and am wondering where
      >>the best place is to invest a few extra dollars to get better
      >>performance in doing lengthy and sizable builds/compiles. stuff like
      >>the linux kernel, large make-based software projects, that kind of
      >>thing.
      >
      >
      > I agree witih David: unless you are building truly large systems, I
      > doubt that adding memory would help much. The disk usage for a
      > 2.6.8.1 kernel tree, including .o files and the resulting kernel,
      > is about 300M, so this can all be in cache with 512M ram.
      >
      > I'm also doubtful that increasing the rpm's of the drive will help;
      > most drives are limited by the ATA bandwidth, which is 100MB/s.
      > It's true that higher rpm's will reduce seek time, but for repeated
      > compiles the source code will hopefully be cached.

      I agree with you and David but all the unused ram is used for disk
      cache, thus if you have a lot of ram you will not need to read on the
      disk (except the first time).
      Nevertheless it's a good idea to start with 512Mb and to buy more later.

      Alex.
    • Robert P. J. Day
      ... 512M would be a *minimum* for me. the issue is that dell typically has a free upgrade from 256M to 512M (a good thing), but that it comes as 2 DIMMs to
      Message 2 of 12 , Dec 1, 2004
      • 0 Attachment
        On Wed, 1 Dec 2004, Alexandre Tessier wrote:

        > Nevertheless it's a good idea to start with 512Mb and to buy more later.

        512M would be a *minimum* for me. the issue is that dell typically
        has a free upgrade from 256M to 512M (a good thing), but that it comes
        as 2 DIMMs to fill the slots (a bad thing). so upgrading later means
        having to replace one or both of the DIMMs. it's just kind of
        annoying that dell frequently makes it easy/cheap to select 512M or
        RAM, but much pricier to bump up to 1G. so i guess i'll just stick
        with 512M for now. thanks.

        rday
      • Dan Christensen
        ... I was asleep while typing and Haedn is of course correct. I m curious what the actual differences are in practice. Right now I m getting about 29.6
        Message 3 of 12 , Dec 1, 2004
        • 0 Attachment
          Haedn Thorn <lordhaedn@...> writes:

          > --- Dan Christensen <jdc@...> wrote:
          >
          >> I'm also doubtful that increasing the rpm's of the
          >> drive will help;
          >> most drives are limited by the ATA bandwidth, which
          >> is 100MB/s.
          >
          > The drives still are not nearly as fast as the
          > interfaces. try 'hdparm -tT' ... You'll see a rate
          > that does not even approach 100MB/s ...

          I was asleep while typing and Haedn is of course correct. I'm
          curious what the actual differences are in practice. Right now
          I'm getting about 29.6 MB/sec according to hdparm -t. I believe
          it's a Toshiba MK4019GAX (HDD2171) 40G, which is a 5400 rpm drive.

          (It's the fourth drive I've had in my I4150 since I got it about two
          years ago. The first three each failed after about 7 months of use!)

          Dan
        • R Stewart
          I did a little experimenting tonight with /etc/sysconfig/harddisks. For the data below, I uncommented each entry (USE_DMA, MULTIPLE_IO, EIDE_32BIT, LOOKAHEAD)
          Message 4 of 12 , Dec 3, 2004
          • 0 Attachment
            I did a little experimenting tonight with /etc/sysconfig/harddisks. For
            the data below, I uncommented each entry
            (USE_DMA, MULTIPLE_IO, EIDE_32BIT, LOOKAHEAD) one after another. I'm
            not sure how accurate these tests were, though, since DMA was enabled
            on my laptop running FC3, even if the USE_DMA line was commented out. I
            turned DMA off for the tests with FC3 with
            # /sbin/hdparm -d 0 /dev/hda

            The desktop is running FC2. The Hitachi drives both have an 8MB cache.
            The Western Digital has a 2 MB cache. This didn't appear to affect the
            cache read tests, as both desktop drives scored around 675 MB/sec for
            buffer-cache reads.

            Interestingly enough, the buffer-cache reads for the Travelstar in my
            laptop averaged slightly over 900 MB/sec.

            The data below is for buffered disk reads.

            The numbers below are for a Hitachi Travelstar 7200 RPM 60 GB drive in
            a Dell Inspiron 8600 laptop.

            * USE_DMA=0 - 2.41 MB/sec
            * USE_DMA=1 - 32 MB/sec
            * MULTIPLE_IO=16 - 38 MB/sec
            * EIDE_32BIT=3 - 38 MB/sec
            * LOOKAHEAD=1 - 58 MB/sec

            The numbers below are for a Hitachi Deskstar 7200 RPM 200 GB drive in a
            Dell Dimension 4400 desktop.

            * USE_DMA=0 - 2.48 MB/sec
            * USE_DMA=1 - 48 MB/sec
            * MULTIPLE_IO=16 - 52 MB/sec
            * EIDE_32BIT=3 - 52 MB/sec
            * LOOKAHEAD=1 - 52 MB/sec

            The numbers below are for a Western Digital 7200 RPM 60 GB drive in a
            Dell Dimension 4400 desktop.

            * USE_DMA=0 - 3.05 MB/sec
            * USE_DMA=1 - 34 MB/sec
            * MULTIPLE_IO=16 - 34 MB/sec
            * EIDE_32BIT=3 - 34 MB/sec
            * LOOKAHEAD=1 - 34 MB/sec

            Robert
            http://www.wombatnation.com/

            --- Dan Christensen <jdc@...> wrote:

            >
            > Haedn Thorn <lordhaedn@...> writes:
            >
            > > --- Dan Christensen <jdc@...> wrote:
            > >
            > >> I'm also doubtful that increasing the rpm's of the
            > >> drive will help;
            > >> most drives are limited by the ATA bandwidth, which
            > >> is 100MB/s.
            > >
            > > The drives still are not nearly as fast as the
            > > interfaces. try 'hdparm -tT' ... You'll see a rate
            > > that does not even approach 100MB/s ...
            >
            > I was asleep while typing and Haedn is of course correct. I'm
            > curious what the actual differences are in practice. Right now
            > I'm getting about 29.6 MB/sec according to hdparm -t. I believe
            > it's a Toshiba MK4019GAX (HDD2171) 40G, which is a 5400 rpm drive.
            >
            > (It's the fourth drive I've had in my I4150 since I got it about two
            > years ago. The first three each failed after about 7 months of use!)
            >
            > Dan
            >
            >
            > --------------------------------------------------------------
            > Please post your X config files in the group links or database
            > To unsubscribe, email: linux-dell-laptops-unsubscribe@yahoogroups.com
            > FAQ: http://www.whacked.net/ldl/faq
            > Yahoo! Groups Links
            >
            >
            >
            >
            >
            >
            >
            >




            __________________________________
            Do you Yahoo!?
            The all-new My Yahoo! - What will yours do?
            http://my.yahoo.com
          Your message has been successfully submitted and would be delivered to recipients shortly.