how get max transfer speed using 1Gbps nics?

  • Thread starter Thread starter Geoff
  • Start date Start date
G

Geoff

Hello

I have 2 PCs connected via a Netgear DG834G v5 router/modem using cat5
ethernet cables..

One with Windows 7 Home Premium, the other with Windows XP Pro.

I have just installed a 1Gbps network card on each but the speed of
file transfer between the two PCs does not seem to have increased.

It takes 90 seconds to transfer a 1GB file from the Windows 7 PC to
the XP Pro PC.

Any suggestions as to how this speed might be increased?

Cheers

Geoff
 
Hello

I have 2 PCs connected via a Netgear DG834G v5 router/modem using cat5
ethernet cables..

One with Windows 7 Home Premium, the other with Windows XP Pro.

I have just installed a 1Gbps network card on each but the speed of
file transfer between the two PCs does not seem to have increased.

It takes 90 seconds to transfer a 1GB file from the Windows 7 PC to
the XP Pro PC.

Any suggestions as to how this speed might be increased?

Cheers

Geoff
Thats because the DG834G only has 10/100 Mbps on it.

So even though you upgraded the cards in your pc's, you are still feeding it
thru a non-gigabit router.
 
Thats because the DG834G only has 10/100 Mbps on it.

So even though you upgraded the cards in your pc's, you are still feeding it
thru a non-gigabit router.

Ah OK - a new router or I suppose a 1Gbps switch?

Cheers

Geoff
 
Geoff said:
Hello

I have 2 PCs connected via a Netgear DG834G v5 router/modem using cat5
ethernet cables..

One with Windows 7 Home Premium, the other with Windows XP Pro.

I have just installed a 1Gbps network card on each but the speed of
file transfer between the two PCs does not seem to have increased.

It takes 90 seconds to transfer a 1GB file from the Windows 7 PC to
the XP Pro PC.

Any suggestions as to how this speed might be increased?

Cheers

Geoff

Simplify the setup and retest.

1) Verify both computer NIC interfaces are operating at 1000BT. If
you happen to be using a 4 wire Ethernet cable on one of the computers,
it will respond by setting up a 100BT connection (which uses 4 wires).
If all eight wires are inside the cable, then the cable can support 1000BT.
I have at least one Ethernet cable in the house, with only four wires
(it came with an older computer). The connectors in that case, still have
the eight pins, but the thinness of the cable is a giveaway.

2) Connect the two computers directly, removing the DG834G from the picture.
That is to remove any doubts about the network box. To do this, you can
set up addresses manually on each computer, like 192.168.1.1 and 192.168.1.2
or whatever is appropriate. With the computers connected directly to each
other, you might not have the convenience of DHCP in the setup, to assign
addresses automatically.

When I was doing your experiment a few years back, I used ICS. This allowed
both computers to reach the Internet, while I was testing things. The gigabit
segment is assigned private addresses like those two examples, when you
do that.
"ICS" GbE
Internet ---- broadband_modem ------ computer_#1 ---------- computer_#2

ICS stands for Internet Connection Sharing.

3) Use a protocol known to have good performance. When I was doing my benchmarks,
I used FTP. That means running FTPd on one machine, and using a regular FTP
client on the other machine. I think my WinXP Pro machine had IIS as an install
option, which would be one way to get an FTP server (FTPd). But there are likely
other ways to do that.

4) On older computers, where a lot of the peripherals are limited by the PCI bus,
you can replace the disk in the test, with a RAMDisk. That prevents the PCI bus
from becoming the bottleneck. This turns a portion of available system memory,
into a file storage device. When I did my experiments, it was with some other
software, and the RAMDisk was pretty small by modern standards.

http://memory.dataram.com/products-and-services/software/ramdisk

RAMDisks are so fast, they remove the storage device as an issue.

http://img196.imageshack.us/img196/8694/hdtunedataram2gbabove.gif

Once I'd done all of that, I was getting 40MB/sec. A quick search revealed, that
Win2K was considered to have a limitation in how fast you could go (which later
OSes have corrected). That's why I feel you have a shot at doing better than that.

The purpose of my testing, was to see how well the network protocol stack in my
OSes could do. One of the computers had an excellent hardware architecture, with
a 266MB/sec bus to the Ethernet chip. So I was pretty confident it wasn't the weak
link. The other side had an integrated GbE LAN, and it seemed to be a reasonable
design as well (it shouldn't have created any limit like I was seeing).

So try those tests first. You may end up concluding, it's an OS issue (protocol
stack is limiting). Or the introduction of the DG834G is the issue. Or it could
be cabling - if the cable is suspiciously thin, then it might be the four wire
type.

There are other features, such as "Jumbo Packet Support", where a 9KB packet
can be used instead of the regular 1.5KB packet. But with the speed you are
seeing right now, this probably isn't a factor. I'm guessing something in the
current network path, is still running 100BT. (11.1MB/sec fits inside 100BT.)

http://en.wikipedia.org/wiki/Jumbo_frame

For some reason, I can't figure out how to check the 100BT versus 1000BT
thing in Windows. If I use Everest, and go to Network : Windows Network,
it says "Connection Speed 1000MBps". I can't remember where you find the
equivalent of that information, in Windows itself. It wasn't in my
Device Manager entry for the NIC. You can use Device Manager to restrict
the PHY on the NIC, to a lower speed artificially, and so that is another
remote possibility - a Device Manager setting is "shooting you in the foot".

http://majorgeeks.com/download4181.html (the old free version of Everest)

Paul
 
Thats because the DG834G only has 10/100 Mbps on it.

So even though you upgraded the cards in your pc's, you are still feeding it
thru a non-gigabit router.

And experience tells me that even if you change the router you will be
lucky to get even half the throughput that 1Gbps might suggest. Changing
over to CAT-6 or at least CAT-5E cable, making sure that every connection
is perfect and that the cable twists aren't disrupted unduly, that every
cable is as short as possible, and that extra connections (such as
wallplates and patch panels) are avoided will help you get the maximum
possible from your setup.

I feel lucky when I get 400Mbps sustained throughput between my fastest PC
and my server. This connection goes through a 2-meter patch cable, a
wallplate, a rather long stretch of CAT-5E cable, a patch panel, a short
patch cable, a switch, and a 2-meter patch cable (all pretty much
unavoidable) along the way.
 
Simplify the setup and retest.

1) Verify both computer NIC interfaces are operating at 1000BT. If
you happen to be using a 4 wire Ethernet cable on one of the computers,
it will respond by setting up a 100BT connection (which uses 4 wires).
If all eight wires are inside the cable, then the cable can support 1000BT.
I have at least one Ethernet cable in the house, with only four wires
(it came with an older computer). The connectors in that case, still have
the eight pins, but the thinness of the cable is a giveaway.

2) Connect the two computers directly, removing the DG834G from the picture.
That is to remove any doubts about the network box. To do this, you can
set up addresses manually on each computer, like 192.168.1.1 and 192.168.1.2
or whatever is appropriate. With the computers connected directly to each
other, you might not have the convenience of DHCP in the setup, to assign
addresses automatically.

When I was doing your experiment a few years back, I used ICS. This allowed
both computers to reach the Internet, while I was testing things. The gigabit
segment is assigned private addresses like those two examples, when you
do that.
"ICS" GbE
Internet ---- broadband_modem ------ computer_#1 ---------- computer_#2

ICS stands for Internet Connection Sharing.

3) Use a protocol known to have good performance. When I was doing my benchmarks,
I used FTP. That means running FTPd on one machine, and using a regular FTP
client on the other machine. I think my WinXP Pro machine had IIS as an install
option, which would be one way to get an FTP server (FTPd). But there are likely
other ways to do that.

4) On older computers, where a lot of the peripherals are limited by the PCI bus,
you can replace the disk in the test, with a RAMDisk. That prevents the PCI bus
from becoming the bottleneck. This turns a portion of available system memory,
into a file storage device. When I did my experiments, it was with some other
software, and the RAMDisk was pretty small by modern standards.

http://memory.dataram.com/products-and-services/software/ramdisk

RAMDisks are so fast, they remove the storage device as an issue.

http://img196.imageshack.us/img196/8694/hdtunedataram2gbabove.gif

Once I'd done all of that, I was getting 40MB/sec. A quick search revealed, that
Win2K was considered to have a limitation in how fast you could go (which later
OSes have corrected). That's why I feel you have a shot at doing better than that.

The purpose of my testing, was to see how well the network protocol stack in my
OSes could do. One of the computers had an excellent hardware architecture, with
a 266MB/sec bus to the Ethernet chip. So I was pretty confident it wasn't the weak
link. The other side had an integrated GbE LAN, and it seemed to be a reasonable
design as well (it shouldn't have created any limit like I was seeing).

So try those tests first. You may end up concluding, it's an OS issue (protocol
stack is limiting). Or the introduction of the DG834G is the issue. Or it could
be cabling - if the cable is suspiciously thin, then it might be the four wire
type.

There are other features, such as "Jumbo Packet Support", where a 9KB packet
can be used instead of the regular 1.5KB packet. But with the speed you are
seeing right now, this probably isn't a factor. I'm guessing something in the
current network path, is still running 100BT. (11.1MB/sec fits inside 100BT.)

http://en.wikipedia.org/wiki/Jumbo_frame

For some reason, I can't figure out how to check the 100BT versus 1000BT
thing in Windows. If I use Everest, and go to Network : Windows Network,
it says "Connection Speed 1000MBps". I can't remember where you find the
equivalent of that information, in Windows itself. It wasn't in my
Device Manager entry for the NIC. You can use Device Manager to restrict
the PHY on the NIC, to a lower speed artificially, and so that is another
remote possibility - a Device Manager setting is "shooting you in the foot".

http://majorgeeks.com/download4181.html (the old free version of Everest)

Paul

Paul,

Many thanks for the detailed response!

I have just tried Everest as you mention above and I see

Connection speed 100Mbps.

One option would be to buy a low cost1Gbps switch and see what
happens!

I see

TP-Link 5-Port Gigabit Unmanaged Desktop Switch (TL-SG1005D) Plastic
Case

for £14 on Amazon...

Cheers

Geoff
 
And experience tells me that even if you change the router you will be
lucky to get even half the throughput that 1Gbps might suggest. Changing
over to CAT-6 or at least CAT-5E cable, making sure that every connection
is perfect and that the cable twists aren't disrupted unduly, that every
cable is as short as possible, and that extra connections (such as
wallplates and patch panels) are avoided will help you get the maximum
possible from your setup.

I feel lucky when I get 400Mbps sustained throughput between my fastest PC
and my server. This connection goes through a 2-meter patch cable, a
wallplate, a rather long stretch of CAT-5E cable, a patch panel, a short
patch cable, a switch, and a 2-meter patch cable (all pretty much
unavoidable) along the way.

John,

Which would be the better option - new 1Gbps router/modem or a 1Gbps
switch?

I could try a £13 1Gbps switch?!

Geoff
 
Geoff said:
I have 2 PCs connected via a Netgear DG834G v5 router/modem using cat5
ethernet cables..

One with Windows 7 Home Premium, the other with Windows XP Pro.

I have just installed a 1Gbps network card on each but the speed of
file transfer between the two PCs does not seem to have increased.

It takes 90 seconds to transfer a 1GB file from the Windows 7 PC to
the XP Pro PC.

Any suggestions as to how this speed might be increased?

You don't have a 1GB (gigabyte) network card. You may have a 1 Gbps
(gigaBIT per second). While a 1GB file might, at best, take 8 seconds
to transfer over a 1 Gbps connection, there is the overhead of adding
the protocol headers in each packet sent from source to target host;
however, it shouldn't take 10 times that to transfer the file - if the
file were entirely in memory in the source host and entirely copied into
memory at the target host - which is not the case here.

So what's the transfer speed of your hard disks at both source and
target hosts? Maybe you only get an effective 33 to 55 Mbps burst speed
(and continuous transfer speed would be less). You're not going to
transfer files over the network faster than you can read them from one
hard disk and then write them to another hard disk. The network
transfer could be a lot faster than the storage devices that end up
limiting how effective is the overall transfer rate. Despite the 4"
main service pipe into your house might deliver 200 gal/minute and the
1/2" water pipe to your bathroom might deliver 12-15 gal/minute, the
faucet might not deliver it faster than 2.2 gal/minute (ask someone
familiar with the plumbing codes to get more accurate figures but you'll
need to know supply PSI, city pump capacity, etc); these were just
examples).

You can using something like HD Tune to find out what are the
*sustained* (more important) and burst (less important) ratings for your
mass storage subsystem (hard disks, controllers, etc). Alas, the free
version only shows the speeds for reads, not writes. Say your source
hard disk has an average read transfer rate of 40 MB/s:

1 GB = 2^9 bytes
1 MB = 2^6 bytes
1 GB / 40 MB/sec = 2^9 bytes / (40 * 2^6 bytes/second) = 1000/40 sec =
25 seconds

That's just to read from the source disk. There will also be some
additional time to store the bytes into a buffer and add the protocol
headers to generate the packets sent out to the network. There's also
time to write to the hard disk on the target disk on the other host to
decode the packets and then write to that hard disk. Write speed is
less than read speed so let's say it took 30 seconds to write the 1 GB
file to the target hard disk; however, there is also overlap in the 2
separately controlled hard disks (while one is writing the other can be
reading and generating more packets). Benchmarks assume that no other
non-involved processes are whacking the CPU usage and you have
sufficient free system RAM. In other words, your source and target
hosts are assumed to be lightly loaded.

You won't get 1 GB/sec transfer with a 1 GB/sec NIC. One, it's not
gigabytes per second rated but gigaBITS per second. So multiply the
above numbers by 8 (8 bits per byte). Similarly, your 1 megabyte (MB)
file is 8 megabits (Mb). Two, that is its maximum rating under special
setup. Typically you'll get somewhere around 72% of the rated transfer
speed with multiple sources for traffic, and much less if you have lots
of other network traffic due to the time needed for conflict resolution.
Ethernet works by resolving conflicts. Ethernet is a Carrier Sense
Multiple Access with Collision Detection (CSMA/CD) protocol; read
http://en.wikipedia.org/wiki/Carrier_sense_multiple_access_with_collision_detection
and see the flow chart there. Only one host can be connected at a time
to one target host, and other hosts trying to concurrently access the
channel get a collision and are required to wait a random interval
before retrying. Multiple access is not the same as concurrent access.
That's why companies slice up their corporate network into subnets to
reduce collisions and reduce lag to improve overall throughput. We
don't know if you quiesced all other network traffic while you were
performing the file transfer. Maybe you were using a shared printer
over the network, or browsing the web, or receiving automatic updates to
Windows, Adobe Reader, anti-virus software, or something else.

Then you're going through the switch in the router. The Netgear DG834G
allows only 54 Mbps when using wireless connects (under ideal setup).
Just where did you see in its specifications that this router supports 1
Gbps bandwidth? I see only a 10/100 Mbps switch mentioned (one tenth of
what you claimed). From the manual at:

ftp://downloads.netgear.com/files/dg834g_reference_manual.pdf

it says, "The DG834G 54 Mbps Wireless ADSL Firewall Router provides
continuous, high-speed 10/100 Ethernet access between your wireless and
Ethernet devices." It doesn't say 1000 (Mbps) or 1 G(bps). Section A
lists the LAN interface as supporting 10BASE-T or 100BASE-Tx (10/100).
So now multiply the time to transfer your files by 10 from when you
previously thought you had 1 Gbps transfer speed.

You aren't delivering (reading) and receiving (writing) the bytes as
fast as you could transfer them over the network. Your 1 GB/s NIC is
faster than your, say, 40/30 MB/s (read/write) hard disks. You don't
have a 1 Gbps LAN switch in your router but just 100 Mbps. Just like
the pipe example above, you can't push 1000 gallons per minute through a
pipe that can only deliver 100 gallons per minute (and where the smaller
"pipe" won't let you increase the pressure). 100 Mbps is all you get to
push through the router no matter using NICs that are *capable* of
delivering more.

To get rid of the choke points, you'll need to get MUCH faster hard
disks (very expensive) or setup RAID and play with stripe sizing (but
probably still won't get over 400 MB/s and perhaps only 180 MB/s) in
trying to up the effective transfer speed of your mass storage system
along with getting a LAN switch (inside a router) that is actually rated
for 1000 Mbps (versus yours at 100 Mbps). Unless you want to do lots of
research on RAID setups to determine the best controller and drives that
give decent speedup, start with replacing your router.
 
Geoff said:
Paul,

Many thanks for the detailed response!

I have just tried Everest as you mention above and I see

Connection speed 100Mbps.

One option would be to buy a low cost1Gbps switch and see what
happens!

I see

TP-Link 5-Port Gigabit Unmanaged Desktop Switch (TL-SG1005D) Plastic
Case

for £14 on Amazon...

Cheers

Geoff

Hmmm. I didn't look up your router. Is this it ? This one has
four 100BT ports, and that would explain the symptoms.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833122124

You said your NIC was GbE. You need an eight wire cable with RJ-45
connectors, to connect to other gigabit compatible equipment.

Yes, you can use a switch, to connect two gigabit compatible machines.
In fact, that's how my computers are currently wired.

ADSL_modem/router ------- 4 port 10/100BT ------ to other, unimportant 100BT machines
(Bridged mode, router box \
router disabled) \
Netgear GbE switch
| |
Computer_#1 Computer_#2

That was my attempt, to avoid the bottleneck on the router (so my
situation is similar to yours, except my network setup has more
junk in it). The 4 port router is a favorite of mine, and has a
nice web interface, which is why I kept it.

The only thing to watch for, in GbE switch boxes, is the "sleep problem".
Many of the latest generation of switch boxes, put the NIC interfaces
to sleep if there is no traffic after a while. They claim it is to
save power. The only disadvantage of this, is you can't send a "Magic
Packet" through the switch, for Wake On LAN. Or something like that.
That's the only issue I vaguely remember.

In terms of brands, TP-Link is known for making cost effective
stuff. So you're doing good there.

One of the reviewer comments here, states the TP-Link supports
Jumbo Frames, if you want to test with them.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833704042

If you're on a budget, I recommend testing from computer to
computer first with a direct connection (and manually set up
IP addresses). That's to see whether there really is the potential
for blazing performance. If the computers and their new NICs
can't manage 125MB/sec under best case conditions (like with
FTP protocol and a direct computer to computer connection),
then your £14 might be a waste. If you only got 20MB/sec
versus your current 11 MB/sec, then it might be a harder
decision as to whether it's worth the trouble. Due to the level
of performance I got, the config above isn't permanent, and I move
the cables around pretty much without caring what goes where.
The GbE side of things, wasn't as fast as I expected. I think
my previous benchmarks with older equipment, were actually
a bit better.

Paul
 
You don't have a 1GB (gigabyte) network card. You may have a 1 Gbps
(gigaBIT per second). While a 1GB file might, at best, take 8 seconds
to transfer over a 1 Gbps connection, there is the overhead of adding
the protocol headers in each packet sent from source to target host;
however, it shouldn't take 10 times that to transfer the file - if the
file were entirely in memory in the source host and entirely copied into
memory at the target host - which is not the case here.

So what's the transfer speed of your hard disks at both source and
target hosts? Maybe you only get an effective 33 to 55 Mbps burst speed
(and continuous transfer speed would be less). You're not going to
transfer files over the network faster than you can read them from one
hard disk and then write them to another hard disk. The network
transfer could be a lot faster than the storage devices that end up
limiting how effective is the overall transfer rate. Despite the 4"
main service pipe into your house might deliver 200 gal/minute and the
1/2" water pipe to your bathroom might deliver 12-15 gal/minute, the
faucet might not deliver it faster than 2.2 gal/minute (ask someone
familiar with the plumbing codes to get more accurate figures but you'll
need to know supply PSI, city pump capacity, etc); these were just
examples).

You can using something like HD Tune to find out what are the
*sustained* (more important) and burst (less important) ratings for your
mass storage subsystem (hard disks, controllers, etc). Alas, the free
version only shows the speeds for reads, not writes. Say your source
hard disk has an average read transfer rate of 40 MB/s:

1 GB = 2^9 bytes
1 MB = 2^6 bytes
1 GB / 40 MB/sec = 2^9 bytes / (40 * 2^6 bytes/second) = 1000/40 sec =
25 seconds

That's just to read from the source disk. There will also be some
additional time to store the bytes into a buffer and add the protocol
headers to generate the packets sent out to the network. There's also
time to write to the hard disk on the target disk on the other host to
decode the packets and then write to that hard disk. Write speed is
less than read speed so let's say it took 30 seconds to write the 1 GB
file to the target hard disk; however, there is also overlap in the 2
separately controlled hard disks (while one is writing the other can be
reading and generating more packets). Benchmarks assume that no other
non-involved processes are whacking the CPU usage and you have
sufficient free system RAM. In other words, your source and target
hosts are assumed to be lightly loaded.

You won't get 1 GB/sec transfer with a 1 GB/sec NIC. One, it's not
gigabytes per second rated but gigaBITS per second. So multiply the
above numbers by 8 (8 bits per byte). Similarly, your 1 megabyte (MB)
file is 8 megabits (Mb). Two, that is its maximum rating under special
setup. Typically you'll get somewhere around 72% of the rated transfer
speed with multiple sources for traffic, and much less if you have lots
of other network traffic due to the time needed for conflict resolution.
Ethernet works by resolving conflicts. Ethernet is a Carrier Sense
Multiple Access with Collision Detection (CSMA/CD) protocol; read
http://en.wikipedia.org/wiki/Carrier_sense_multiple_access_with_collision_detection
and see the flow chart there. Only one host can be connected at a time
to one target host, and other hosts trying to concurrently access the
channel get a collision and are required to wait a random interval
before retrying. Multiple access is not the same as concurrent access.
That's why companies slice up their corporate network into subnets to
reduce collisions and reduce lag to improve overall throughput. We
don't know if you quiesced all other network traffic while you were
performing the file transfer. Maybe you were using a shared printer
over the network, or browsing the web, or receiving automatic updates to
Windows, Adobe Reader, anti-virus software, or something else.

Then you're going through the switch in the router. The Netgear DG834G
allows only 54 Mbps when using wireless connects (under ideal setup).
Just where did you see in its specifications that this router supports 1
Gbps bandwidth? I see only a 10/100 Mbps switch mentioned (one tenth of
what you claimed). From the manual at:

ftp://downloads.netgear.com/files/dg834g_reference_manual.pdf

it says, "The DG834G 54 Mbps Wireless ADSL Firewall Router provides
continuous, high-speed 10/100 Ethernet access between your wireless and
Ethernet devices." It doesn't say 1000 (Mbps) or 1 G(bps). Section A
lists the LAN interface as supporting 10BASE-T or 100BASE-Tx (10/100).
So now multiply the time to transfer your files by 10 from when you
previously thought you had 1 Gbps transfer speed.

You aren't delivering (reading) and receiving (writing) the bytes as
fast as you could transfer them over the network. Your 1 GB/s NIC is
faster than your, say, 40/30 MB/s (read/write) hard disks. You don't
have a 1 Gbps LAN switch in your router but just 100 Mbps. Just like
the pipe example above, you can't push 1000 gallons per minute through a
pipe that can only deliver 100 gallons per minute (and where the smaller
"pipe" won't let you increase the pressure). 100 Mbps is all you get to
push through the router no matter using NICs that are *capable* of
delivering more.

To get rid of the choke points, you'll need to get MUCH faster hard
disks (very expensive) or setup RAID and play with stripe sizing (but
probably still won't get over 400 MB/s and perhaps only 180 MB/s) in
trying to up the effective transfer speed of your mass storage system
along with getting a LAN switch (inside a router) that is actually rated
for 1000 Mbps (versus yours at 100 Mbps). Unless you want to do lots of
research on RAID setups to determine the best controller and drives that
give decent speedup, start with replacing your router.

Thanks for the detailed thoughts - I am tempted to try a cheap £13
1Gbps switch to see what that does ...

Cheers

Geoff
 
Hmmm. I didn't look up your router. Is this it ? This one has
four 100BT ports, and that would explain the symptoms.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833122124

You said your NIC was GbE. You need an eight wire cable with RJ-45
connectors, to connect to other gigabit compatible equipment.

Yes, you can use a switch, to connect two gigabit compatible machines.
In fact, that's how my computers are currently wired.

ADSL_modem/router ------- 4 port 10/100BT ------ to other, unimportant 100BT machines
(Bridged mode, router box \
router disabled) \
Netgear GbE switch
| |
Computer_#1 Computer_#2

That was my attempt, to avoid the bottleneck on the router (so my
situation is similar to yours, except my network setup has more
junk in it). The 4 port router is a favorite of mine, and has a
nice web interface, which is why I kept it.

The only thing to watch for, in GbE switch boxes, is the "sleep problem".
Many of the latest generation of switch boxes, put the NIC interfaces
to sleep if there is no traffic after a while. They claim it is to
save power. The only disadvantage of this, is you can't send a "Magic
Packet" through the switch, for Wake On LAN. Or something like that.
That's the only issue I vaguely remember.

In terms of brands, TP-Link is known for making cost effective
stuff. So you're doing good there.

One of the reviewer comments here, states the TP-Link supports
Jumbo Frames, if you want to test with them.

http://www.newegg.com/Product/Product.aspx?Item=N82E16833704042

If you're on a budget, I recommend testing from computer to
computer first with a direct connection (and manually set up
IP addresses). That's to see whether there really is the potential
for blazing performance. If the computers and their new NICs
can't manage 125MB/sec under best case conditions (like with
FTP protocol and a direct computer to computer connection),
then your £14 might be a waste. If you only got 20MB/sec
versus your current 11 MB/sec, then it might be a harder
decision as to whether it's worth the trouble. Due to the level
of performance I got, the config above isn't permanent, and I move
the cables around pretty much without caring what goes where.
The GbE side of things, wasn't as fast as I expected. I think
my previous benchmarks with older equipment, were actually
a bit better.

Paul

Paul,

I've just ordered the £13 switch - will get back with details once
received.

Cheers

Geoff
 
John,

Which would be the better option - new 1Gbps router/modem or a 1Gbps
switch?

I could try a £13 1Gbps switch?!

Geoff

A 1Gb/s switch will do the job of enabling fast data between your
computers. Basically you would have a cable from the modem/router feeding
the switch and then the switch would feed the two computers. I can't
comment on how good a £13 switch might be. That would be something like
$21-22 as a quick estimate and that does seem a low figure. From what I've
seen UK prices tend to be a good bit higher (VAT + higher import taxes?)
than those in the US so I'd be suspicious of one that is too too cheap.

Is it the TP-Link 5-port switch you're looking at? It does have good
reviews on Amazon.co.uk so it might be worth a try and with Amazon's return
policy you wouldn't be out anything if it does turn out to be crap.

I have always preferred the old-style Netgear metal-cased routers and
switches. My main switch in the equipment closet is a Netgear GS108 but I
do have a secondary GS608 which splits a single feed going to my
entertainment center so that it can go to both my HTPC and satellite box
and two NAS boxes. The second switch is a plastic 'new' type and it seems
to work fine, just not as stout. The closest thing to those at amazon.co.uk
would be the GS105 or GS605 and they do cost £20 more and might not be
worth the extra to you as a experiment fodder.
 
A 1Gb/s switch will do the job of enabling fast data between your
computers. Basically you would have a cable from the modem/router feeding
the switch and then the switch would feed the two computers. I can't
comment on how good a £13 switch might be. That would be something like
$21-22 as a quick estimate and that does seem a low figure. From what I've
seen UK prices tend to be a good bit higher (VAT + higher import taxes?)
than those in the US so I'd be suspicious of one that is too too cheap.

Is it the TP-Link 5-port switch you're looking at? It does have good
reviews on Amazon.co.uk so it might be worth a try and with Amazon's return
policy you wouldn't be out anything if it does turn out to be crap.

I have always preferred the old-style Netgear metal-cased routers and
switches. My main switch in the equipment closet is a Netgear GS108 but I
do have a secondary GS608 which splits a single feed going to my
entertainment center so that it can go to both my HTPC and satellite box
and two NAS boxes. The second switch is a plastic 'new' type and it seems
to work fine, just not as stout. The closest thing to those at amazon.co.uk
would be the GS105 or GS605 and they do cost £20 more and might not be
worth the extra to you as a experiment fodder.

John,

Yes, it is the TP-Link 5-port switch on Amazon - will let you know how
it behaves!

Cheers

Geoff
 
Back
Top