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

Re: memory upgrade on LS1 ?

Expand Messages
  • James Stewart
    I m often amazed how well my LS runs on just 64mb. Granted it gets real slow running Mozilla Firefox, but considering all the other stuff I have running on it
    Message 1 of 9 , Sep 18, 2005
    • 0 Attachment
      I'm often amazed how well my LS runs on just 64mb. Granted it gets
      real slow running Mozilla Firefox, but considering all the other
      stuff I have running on it at the same time, I'm really happy with
      the performance (memory management anyway).

      I even reduced the swap partition from the original 512M to 256M and
      I usually never swap more than about 64 megs in normal use (running
      a graphical desktop through VNC, ssh, apache, ftp, webmin, DNS
      server, Full NTP time server, samba, mail server, then having it do
      something like running some bit-torrents, or normalizing a bunch of
      mp3's, or even playing some of the through a USB audio dongle) and
      even then, half the swap is just buffer/cache.

      What I'd like to speed up is USB throughput, but we need to
      re-compile the kernel to do that (take the "debug" mode out of USB).
    • haberschnasel
      ... This would be great. Increasing network throughput would be great, too. But, do you expect more than the 10% USB speed increase reported in post #3779?
      Message 2 of 9 , Sep 19, 2005
      • 0 Attachment
        --- In LinkStation_General@yahoogroups.com, "James Stewart"
        <wartstew@y...> wrote:
        > What I'd like to speed up is USB throughput, but we need to
        > re-compile the kernel to do that (take the "debug" mode out of USB).

        This would be great. Increasing network throughput would be great,
        too. But, do you expect more than the 10% USB speed increase reported
        in post #3779?
      • James Stewart
        ... Maybe. What is odd is that I could never get more the USB-1.0 (12MB/S) mode detected when using the Kurobox Kernel on my LS in the first place. I wonder
        Message 3 of 9 , Sep 19, 2005
        • 0 Attachment
          --- In LinkStation_General@yahoogroups.com, "haberschnasel"
          <haberschnasel@y...> wrote:
          > --- In LinkStation_General@yahoogroups.com, "James Stewart"
          > <wartstew@y...> wrote:
          > > What I'd like to speed up is USB throughput, but we need to
          > > re-compile the kernel to do that (take the "debug" mode out of USB).
          >
          > This would be great. Increasing network throughput would be great,
          > too. But, do you expect more than the 10% USB speed increase reported
          > in post #3779?

          Maybe. What is odd is that I could never get more the USB-1.0
          (12MB/S) mode detected when using the Kurobox Kernel on my LS in the
          first place. I wonder if the same thing happens on the Kuroboxes with
          that same kernel? In this case the 6.somthing MB transfer rate might
          be about right once you deal with all the overhead.

          I guess I need to benchmark my network performance. Since I'm using
          mostly current Samba release on my LS, It does seem faster, but maybe
          I'm a victim of a placebo effect.

          At least I know I'm running the hard drive in DMA mode and am getting
          over 26MB/S out of it (on the slow 5400RPM Samsung 120MB drive). I
          had found a post where someone got info from Melco/Buffalo that the LS
          is only capable of 29MB/S bus speed anyway, so If I could get USB
          performance of about half that I think I would be happy.

          -------

          A side note, my fan started getting noisy. Because replacement is a
          little tricky, I pulled it out and pulled the plastic sticker off the
          hub, put a drop of oil into the bearing and put the sticker back on.
          It is now nice and quiet again. I've done this trick in the past with
          these kind of cheap fans and got a lot of extra life out of them.
        • gkennydunn
          Wondering where your local disk stats came from. I ve tried a few things with varying success but now have hdparm -tT /dev/hda /dev/hda: Timing buffer-cache
          Message 4 of 9 , Sep 21, 2005
          • 0 Attachment
            Wondering where your local disk stats came from. I've tried a few
            things with varying success but now have

            hdparm -tT /dev/hda

            /dev/hda:
            Timing buffer-cache reads: 128 MB in 1.82 seconds = 70.33 MB/sec
            Timing buffered disk reads: 64 MB in 2.02 seconds = 31.68 MB/sec

            Obviously, this whole discussion belongs in a thread away from any
            memory issues which we would all like to overcome.

            an LS1 I just received a few weeks back

            --- In LinkStation_General@yahoogroups.com, "James Stewart"
            <wartstew@y...> wrote:
            > --- In LinkStation_General@yahoogroups.com, "haberschnasel"
            > <haberschnasel@y...> wrote:
            > > --- In LinkStation_General@yahoogroups.com, "James Stewart"
            > > <wartstew@y...> wrote:
            > > > What I'd like to speed up is USB throughput, but we need to
            > > > re-compile the kernel to do that (take the "debug" mode out of USB).
            > >
            > > This would be great. Increasing network throughput would be great,
            > > too. But, do you expect more than the 10% USB speed increase reported
            > > in post #3779?
            >
            > Maybe. What is odd is that I could never get more the USB-1.0
            > (12MB/S) mode detected when using the Kurobox Kernel on my LS in the
            > first place. I wonder if the same thing happens on the Kuroboxes with
            > that same kernel? In this case the 6.somthing MB transfer rate might
            > be about right once you deal with all the overhead.
            >
            > I guess I need to benchmark my network performance. Since I'm using
            > mostly current Samba release on my LS, It does seem faster, but maybe
            > I'm a victim of a placebo effect.
            >
            > At least I know I'm running the hard drive in DMA mode and am getting
            > over 26MB/S out of it (on the slow 5400RPM Samsung 120MB drive). I
            > had found a post where someone got info from Melco/Buffalo that the LS
            > is only capable of 29MB/S bus speed anyway, so If I could get USB
            > performance of about half that I think I would be happy.
            >
            > -------
            >
            > A side note, my fan started getting noisy. Because replacement is a
            > little tricky, I pulled it out and pulled the plastic sticker off the
            > hub, put a drop of oil into the bearing and put the sticker back on.
            > It is now nice and quiet again. I've done this trick in the past with
            > these kind of cheap fans and got a lot of extra life out of them.
          • Arias Hung
            ... Sorry I wasn t more clear. By stating the chipset of the gig ethernet chip I meant to imply that only the gig ethernet driver is compiled into the stock
            Message 5 of 9 , Sep 22, 2005
            • 0 Attachment
              On Mon, 19 Sep 2005, irvine_stock delivered in simple text monotone:

              >Arias,
              >The question is not if the HG uses a 10/100/100 realtek chip. That's
              >obvious. My HG uses the 8110 chip. The LS1 uses the ADAMtek 983B.
              >Maybe it is even compatible with the lower speed realtek?
              >Does anybody know the difference between a RealTek
              >8100/8110/8139/8169? Are these all bug fixes?

              Sorry I wasn't more clear. By stating the chipset of the gig ethernet chip I meant to imply that only the gig ethernet driver is compiled into the stock HG kernel. At least in every kernel config I've ever seen for the HG there has never been anything more than four ethernet compile options:

              CONFIG_NET_ETHERNET=y
              CONFIG_NET_PCI=y
              CONFIG_R8169=y
              RTL8169_JUMBO_FRAME_SUPPORT=y

              if you find an existing firmware in which a 10/100 card is compiled into a HG firmware, that would be fantastic since I would hate to see such a labor intensive mod (at least for me it would be :) ) be in vain.

              As alternative to jtag, you could get the serial console with input line working, which would allow you to load a proper kernel without a network connections ( at least our japanese colleagues seem to do it this way ).

              A
            Your message has been successfully submitted and would be delivered to recipients shortly.