Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Ever since installing Windows 7 SP1, I have noticed that all programs that display my network transfer rate have been exactly 4x higher than they actually are.

For example, when I download something from a high-bandwidth web site or through torrents with lots of sources, the download rate indicated is is ~5MBps (~40Mbps) even though my Internet connection has a maximum of only 1.5MBps (12Mbps). It is the same situation with the upstream bandwidth: the connection maximum is 64KBps, but I’m seeing up to 256KBps. I have tried several different programs for monitoring bandwidth throughput and they all give the same results. I also tried different times and different days, and they always show the rate as being four times too high.

My initial thought was that my ISP had increased the speeds (without my noticing), which they have done before. However, I checked my ISP’s site and they have not increased the speeds. Moreover, when I look at the speeds in the program actually doing the transfer (eg Chrome, µTorrent, etc.), the numbers are in line with the expected values at the same time that bandwidth monitoring programs are showing the high numbers.

The only significant change (and pretty much the only change at all) that has occurred to my system since the change was the installation of SP1 for Windows 7. As such, it is my belief that some sort of change exists in SP1 whereby software that accesses the bandwidth via a specific API receives (erroneously?) high numbers while others that have access to the raw data continue to receive the correct values.

I booted into Windows XP and downloaded some things via HTTP and torrent and in both cases, the numbers were as expected (like they were in Windows 7 before installing SP1). I then booted back into 7SP1 and once again, the numbers were four times higher than possible. Therefore it is definitely something in SP1 that has changed how local bandwidth is calculated/returned.

There is definitely something wonky with Windows 7 SP1’s network speed calculation.

I tried Googling this, but (for multiple reasons), have had a difficult time finding anything relevant.

Has anybody else noticed this behavior? Does anybody know of any bugs or changes in SP1 that could account for it?

share|improve this question
Boot into Linux and get some reference numbers. – surfasb Dec 4 '11 at 0:19
@surfasb, I’ve been spending a lot of time in Windows 7 this week. I don’t have *nix (or care to), but tomorrow morning, I’ll boot into XP to see if it happens. (I’d consider uninstalling, testing and reinstalling SP1, but that’s far more work and time than I care to invest in Microsoft at this point—and with my burgeoning cold at that.) – Synetech Dec 4 '11 at 0:26
Reference numbers would help a lot. This isn't exactly a common troubleshooting problem and anything you can do to narrow down the problem helps everyone. – surfasb Dec 4 '11 at 0:41
> This isn't exactly a common troubleshooting problem That is why I asked; to see if others are experiencing the behavior. – Synetech Dec 4 '11 at 1:11
That wasn't the reason I mentioned the lack of frequency, but ok. – surfasb Dec 4 '11 at 1:13

Perhaps the ISP speeds were fluctuating so that your speeds were 5MBps/40MBps during the file transfer, but later the speed dropped when you did your other test.

Since you suspect a windows phenomenon instead of a network phenomenon, I recommend isolating the phenomenon as much as possible. I would take the internet out of the equation by doing a speed test over a LAN. I can't imagine the windows service pack affecting the protocol of the network adapter. But isolating factors will help determine the cause of what you are experiencing.

A while back I had a bad ethernet cable which had noise. This cable caused a segment on the LAN to fluctuate between 100 Mbps and 1.0 Gbps protocols, depending on the noise at the time. It literally fluctuated back and fourth. After I replaced the cable I was at 1.0 Gbps on the lan all the time. Perhaps a segment in your home setup is alternating between 10 Mbps and 100 Mbps because of a bad connection or bad cable. That's something else to isolate. Perhaps your internet speed was increased a long time ago, but a bottle neck on your LAN is preventing you from enjoying the full speed your ISP is providing.

share|improve this answer
My ISP does have a “speed boost” feature whereby some transfers (HTTP, not torrent) will be sped up beyond the speed limit if local network usage is low (but then only for a few MB). However, that is NOT the case here. When I said “at the same time”, I meant it literally. That is, I can see the speeds in Winbar’s bandwidth module and DuMeter’s window at the same time. Further, the inconsistent speeds are not just a momentary thing, but sustained throughout the transfer, even if it lasts for several minutes. Finally, this behavior has persisted daily for almost a week now. – Synetech Dec 3 '11 at 23:00
Why do you think Windows is causing this instead of the network? – steampowered Dec 3 '11 at 23:05
Because like I said, 1: two softwares giving different values, 2: neither software having been changed, 3: SP1 being the only change since this behavior began, and (4): it is conceivable that one of the methods of measuring bandwidth have been changed others remain the same. – Synetech Dec 3 '11 at 23:18
up vote 0 down vote accepted

I figured it out. Under Windows 7, each and every network interface and every variation thereof is listed. In Windows XP, only enabled interfaces are listed, and then, only the primary entry for those. An example helps clarify the issue.

On my system, under Windows XP, Winbar’s network interface list lets you choose from the following to track:

3Com Gigabit LOM (3C940)
MS TCP Loopback interface

Under Windows 7, Winbar lists these:

3Com 3C940 Gigabit LOM Ethernet Adapter
3Com 3C940 Gigabit LOM Ethernet Adapter - ProtoWall Miniport
3Com 3C940 Gigabit LOM Ethernet Adapter - ProtoWall Miniport-WFP LightWeight Filter-0000
Bluetooth Device (Personal Area Network)
Bluetooth Device (Personal Area Network) #2
Bluetooth Device (RFCOMM Protocol TDI)
Bluetooth Device (RFCOMM Protocol TDI) #2
D-Link DFE-538TX 10/100 Adapter
D-Link DFE-538TX 10/100 Adapter
D-Link DFE-538TX 10/100 Adapter - ProtoWall Miniport
Microsoft ISATAP Adapter #2
Microsoft ISATAP Adapter #3
RAS Async Adapter
Realtek 8180 Extensible 802.11b Wireless Device
Realtek 8180 Extensible 802.11b Wireless Device - ProtoWall Miniport
Software Loopback Interface 1
WAN Miniport (IKEv2)
WAN Miniport (IP)
WAN Miniport (IP) - ProtoWall Miniport
WAN Miniport (IP) - ProtoWall Miniport-QoS Packet Scheduler-0000
WAN Miniport (IPv6)
WAN Miniport (IPv6) - ProtoWall Miniport
WAN Miniport (IPv6) - ProtoWall Miniport-QoS Packet Scheduler-0000
WAN Miniport (L2TP)
WAN Miniport (Network Monitor)
WAN Miniport (Network Monitor) - ProtoWall Miniport
WAN Miniport (Network Monitor) - ProtoWall Miniport-QoS Packet Scheduler-0000
WAN Miniport (PPPOE)
WAN Miniport (PPTP)
WAN Miniport (SSTP)

Notice how the D-Link adapter, the Protowall filters, some uninstalled ones (e.g., “… #2”), and several other miniports and such are all listed even though a lot of them are disabled while others are not even actual adapters (I don’t even have a Bluetooth device installed anymore). It is the same situation with other bandwidth monitoring programs.

As such, software which normally monitors all interfaces by default (which worked in XP) is now seeing the same data going through several “adapters” but counts each one separately, hence the results being an exact multiple of the real number.

The solution is to disable all the extraneous interfaces et voilà! the numbers are in line with reality again. (This is unnecessary in browsers, P2P apps, etc. because they track the bytes that they themselves process, so they have access to the actual, raw data instead of sniffing adapters like monitoring programs do.)

This works great and is simple to fix, but unfortunately it has the drawback that some programs don’t let you select multiple, individual interfaces (i.e., check-boxes); they let you select either “All Interfaces” or a single one (i.e., a drop-list). If this is a problem, then, it would be necessary to contact the dev to explain the limitation and have them update the program.

share|improve this answer
(Yes, I know that Protowall doesn’t work in 7; that’s why it is disabled. I’ll get around to uninstalling it later.) – Synetech Apr 2 '12 at 22:50
Curious why it only started after I installed SP1 though. I suppose SP1 could have added new virtual interfaces or changed something that exposes them instead of leaving them hidden or something. – Synetech Nov 20 '12 at 17:37

You must log in to answer this question.

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