S
Sabahattin Gucukoglu
Hi all,
I have a Realtek RTL8139 fast ethernet board in my Intel 1.7ghz Pentium
IV/P4-7266A box, amethyst, connected to a 10/100mbit auto-negotiating
switch from Netgear, which runs Windows 2000 and wants to do things on the
net faster than about 15KByte/s. It starts out at about 200kbyte/sec when
downloading stuff from remote hosts, which is about right for my 2mbit/sec
ADSL downstream, but gradually, slowly, falls down to 15kbyte/sec. It
negotiates, according to the network connection's status window, at
100mbit/sec. The router and proxy to the outside world for Amethyst is
Bloodstone, which runs Linux 2.6.14/Squid and uses a Realtek Rtl8169
gigabit chip. When Chalcedony, a Windows XP box (but soon to be Win2k if
this problem doesn't affect it) with an AMD64 3400+/Asus VSEDX and a
Marvell Yukon gigabit chip onboard is connected through the same switch and
to Bloodstone, it sees no transfer issues from Bloodstone, going flat out
at a good 11 mbytes/sec when transfering using rsync directly from
Bloodstone with entirely new data from Bloodstone and whole file mode of
transfer (E.G.: it's all cool).
I've looked around on the net, and it seems other people have had similar
issues with entirely Windows setups between these two chips, so I'm ruling
out seriously buggy drivers on either side, since one or the other of them
clearly operate correctly on their own. Nach, I've updated drivers for all
of the hardware on all boxes, AFAIK.
I've looked into my IRQs because they seem to be a common cause of this
sort of thing (all the PCI boards, monitor, sound, and that lot are on IRQ
11 which includes the ethernet), and tried to disable all unneeded stuff to
free resources. It didn't work, because of ACPI and IRQ sharing, which
Windows seems to obey no matter what you try to do to set the values by
hand (BIOS, device manager won't let you go near them). The network
adaptor could never sit on an unoccupied IRQ. It was only due to
devcon.exe that I was able to recover from the last lot of experimentation,
because somehow my keyboard driver wouldn't load after disabling the USB
controller (no, of course it wasn't a USB keyboard). Thank god telnet
service was running, I just so happened to have pscp.exe and unzip.exe over
there and devcon.exe was a WinZip self-extractor! Lucky shot, but I vote
for devcon (or something similar) be included in future releases of
Windows - I've heard stories of people turning off their monitors
(not
an issue for me, I'm blind). The sound card was crackling when my waveout
speech synthesiser used it, but I updated the drivers and it works fine now
(onboard, Realtek ALC101 AC97 audio - Windows Update keeps insisting,
rather annoyingly, that it's either a C-Media or an Analog Devices, neither
of which is true), but I don't think that's anything to do with this
problem.
I can't see any other way of proceeding for the moment. Does anyone here
have any sort of ideas what could be going on? I know Realtek chips are a
bit of a bother, but they're what I have (and besides, I don't believe
they're *that* bad - they've always served me until now and there's no way
my processor can take only 15kbyte/sec!).
Any help much appreciated! I've been doing this investigation for over a
week and can't go on without sorting it out (if you think I'll transfer
over 32gbytes at 15kbyte/sec, well, you're wrong ). Thanks in advance
for your help.
Cheers,
Sabahattin
I have a Realtek RTL8139 fast ethernet board in my Intel 1.7ghz Pentium
IV/P4-7266A box, amethyst, connected to a 10/100mbit auto-negotiating
switch from Netgear, which runs Windows 2000 and wants to do things on the
net faster than about 15KByte/s. It starts out at about 200kbyte/sec when
downloading stuff from remote hosts, which is about right for my 2mbit/sec
ADSL downstream, but gradually, slowly, falls down to 15kbyte/sec. It
negotiates, according to the network connection's status window, at
100mbit/sec. The router and proxy to the outside world for Amethyst is
Bloodstone, which runs Linux 2.6.14/Squid and uses a Realtek Rtl8169
gigabit chip. When Chalcedony, a Windows XP box (but soon to be Win2k if
this problem doesn't affect it) with an AMD64 3400+/Asus VSEDX and a
Marvell Yukon gigabit chip onboard is connected through the same switch and
to Bloodstone, it sees no transfer issues from Bloodstone, going flat out
at a good 11 mbytes/sec when transfering using rsync directly from
Bloodstone with entirely new data from Bloodstone and whole file mode of
transfer (E.G.: it's all cool).
I've looked around on the net, and it seems other people have had similar
issues with entirely Windows setups between these two chips, so I'm ruling
out seriously buggy drivers on either side, since one or the other of them
clearly operate correctly on their own. Nach, I've updated drivers for all
of the hardware on all boxes, AFAIK.
I've looked into my IRQs because they seem to be a common cause of this
sort of thing (all the PCI boards, monitor, sound, and that lot are on IRQ
11 which includes the ethernet), and tried to disable all unneeded stuff to
free resources. It didn't work, because of ACPI and IRQ sharing, which
Windows seems to obey no matter what you try to do to set the values by
hand (BIOS, device manager won't let you go near them). The network
adaptor could never sit on an unoccupied IRQ. It was only due to
devcon.exe that I was able to recover from the last lot of experimentation,
because somehow my keyboard driver wouldn't load after disabling the USB
controller (no, of course it wasn't a USB keyboard). Thank god telnet
service was running, I just so happened to have pscp.exe and unzip.exe over
there and devcon.exe was a WinZip self-extractor! Lucky shot, but I vote
for devcon (or something similar) be included in future releases of
Windows - I've heard stories of people turning off their monitors
(notan issue for me, I'm blind). The sound card was crackling when my waveout
speech synthesiser used it, but I updated the drivers and it works fine now
(onboard, Realtek ALC101 AC97 audio - Windows Update keeps insisting,
rather annoyingly, that it's either a C-Media or an Analog Devices, neither
of which is true), but I don't think that's anything to do with this
problem.
I can't see any other way of proceeding for the moment. Does anyone here
have any sort of ideas what could be going on? I know Realtek chips are a
bit of a bother, but they're what I have (and besides, I don't believe
they're *that* bad - they've always served me until now and there's no way
my processor can take only 15kbyte/sec!).
Any help much appreciated! I've been doing this investigation for over a
week and can't go on without sorting it out (if you think I'll transfer
over 32gbytes at 15kbyte/sec, well, you're wrong
). Thanks in advancefor your help.
Cheers,
Sabahattin