1

I just got 150 mbps cable modem service.

I have a Netgear R6300v2 router, a Motorola SB6141 cable modem, and a Netgear GS605v4 switch.

I have a Sony VAIO laptop that when hooked up to the modem directly, gets ~150mbps on speedtest.net. When it's hooked up to the router it gets ~150mbps, when hooked up to the switch it gets ~150mbps.

I have a Windows 7 PC with a Realtek PCIe GBE Family controller built into the motherboard. This PC will only get about ~50 mbps when hooked up to the same switch that the laptop gets good throughput for. Also, I have a Gateway computer running Vista that is also getting ~50 mbps when tested.

Why are the two PCs limited to that speed and what settings can I look for? I tried disabling a bunch of options in the driver for green ethernet and powersaving. And I removed Auto Negotiate and just set it to 1g full duplex.

What else am I missing that lets the Sony laptop utilize the full speed but limits the two PCs?

Update: Tried a Windows 8 Lenovo laptop connected to same switch with same cable and it's getting ~150mbps as well.

This is all with the same cord/cable as well.

3
  • My experience with generic NICs is similar to what you're experiencing. I either always get a motherboard with a KILLER NIC or Intel NIC and usually get the real speeds. Can you try to transfer files from your Windows 7 PC across the network to see what kind of speeds you're getting? You may also want to try and disable IPv6 to see if that helps.
    – kobaltz
    Commented Oct 4, 2013 at 5:22
  • TeraCopy shows I'm transferring files from router's attached USB drive to the Win7 local drive at 20 MB/s. And back to router at same speed. Commented Oct 4, 2013 at 6:26
  • I just realized the 20MB/s might be USB speeds. File transfered to the Win8 laptop from Win7 PC in question seemed to be about 75MB/s most of the time. Commented Oct 4, 2013 at 6:40

1 Answer 1

0

When you tried the Windows 7 PC and the Gateway box with Vista, did you use the same Ethernet cable as you used for the VAIO and the Lenovo? You might have an old 10/100 Ethernet cable that only has two pairs (4 wires) connected.

Gigabit cables require all 4 pairs (8 wires) to be connected, and which pins are in which twisted pairs matters, so you can get the pinout right but have the which-pins-are-in-which-twisted-pairs wrong, and be unable to send gigabit signals reliably.

Rather than introduce the potential variability of Internet connection speeds into the mix, I would test with a TCP performance test tool like IPerf instead of using Speedtest.net. I also prefer IPerf over file copy tests because file copy tests introduce disk, filesystem, and file service protocol (SMB) overhead and latencies.

I'd plug two machines via known good Cat5 or better, properly wired cables into the Netgear switch (and make sure there's basically no other traffic going through the switch at the time), and then run iperf -s on one PC and iperf -c <IPAddrOfOtherIPerfPC> on the other, and see what performance I get. I'd also look at what TCP window size Windows picked for me (IPerf will show this as the test starts). If it's not at least 37 KiB, I'd manually add -w 128K to both the IPerf server and client command line arguments. This forces the TCP window size to 128KiB, which should be plenty. If you get in the high 930's of megabits per second, you're in good shape. There's no excuse for any modern 1GHz or better PC to get less than 900 megabits per second of IPerf TCP throughput on GigE over properly wired Cat5 or better cabling.

6
  • Same cable. I edited original to make sure that is clear. I will try the iPerf thing too. Commented Oct 4, 2013 at 6:52
  • iperf -c from slower Win7 pc to faster Win8 PC, shows 64.0 Kbyte TCP Window Size and 628 Mbits for a 748 mb transfer. Commented Oct 4, 2013 at 7:01
  • The -w 128k yielded 1.04gb transfer at 889 Mbps Commented Oct 4, 2013 at 7:09
  • Okay, so locally, you can get decent speeds when your TCP window size isn't getting in the way. Hmm, but your 150 mbps cable service is probably only 150mbps on download, so you want to run the IPerf server on the slow machine and the client on the fast, so that you see how fast the slow machine can receive data.
    – Spiff
    Commented Oct 4, 2013 at 7:53
  • It might also be interesting to hook one of your slow machines directly to the back of the cable modem (you may need to fully power-cycle your cable modem for it to notice the change), and see how it does on Speedtest.net without the Netgear boxes in the way.
    – Spiff
    Commented Oct 4, 2013 at 7:54

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .