Disable Self-Tuning TCP

I need to disable this or, at the very least, take control of it. It's made doing anything on the internet a miserable, lengthy process. I had no trouble with Tiger and nothing but trouble with Leopard. It's throttling my downstream on both my Macs to useless levels.

If I can't disable or modify this I'm going to go back to Tiger to get things done. I own a business. I cant have this time wasting nonsense and I need a solution now. Much of my work is on the web. I'm paying for high speed access and getting low speed throughput.

The DNS settings are in place, no firewall setting improves this, I've reset all the Airports. Nothing changes. Downloading anything crawls to low speed but ONLY on the Macs.

Does anyone have a clue? Can I archive and install back to Tiger?

iMac G5 2.0 + MacBook 2.0, Mac OS X (10.5.1)

Posted on Nov 22, 2007 8:24 AM

Reply
37 replies

Dec 1, 2007 12:01 PM in response to mreckhof

Glad to see the hours I've been tinkering with this have not been a snipe hunt. Here's another bit to tinker with and confirm/deny:

Set up the router for a WPA2 network and disable SSID broadcasting.

In Network add the preferred connection and select WPA2 Personal.

In TCP/IP allow IPv4 and v6 to automatically set.

In DNS add the DNS servers and router address, add the search domain.

Back at Network panel deselect "Ask to join...".

Click Apply, then lock it.

In Security, click Firewall, select Set Access, then Advanced.

Select "Enable Stealth..."

Here's where the fun begins. In theory Airport and the router are now blind. The router is not broadcasting an ID, the computer is not acknowledging it's presence to unknown downstream pings. The computer should only connect as manually configured. Airport will turn on and may or may not connect, maybe kick a fuss about WPA2 and throw back false errors. Flick it on and off a few times and it connects and stays a bit. Until Power Management kicks in. You may have to reboot for this to work. You may also have to clear any other Airport related entry in the Keychain.

There's so many things not right about this scenario. So far 6 hours of up time and all is working. Airport is still off a few points from maximum downstream but it's a heck of a lot better than the Airport Base connections by a huge margin. The Dlink is staying, the Airport bases are going.

Dec 1, 2007 8:02 PM in response to Ken.

I'll give that a shot and see if I get the same thing.

Are you using the Broadcom 43xx chipset? (kextstat | grep -i airport)

I wonder if everyone having wireless problems are using this chipset.

The firmware in the current driver is 4.170.25.8.2. The good news is, this is NOT the latest version of firmware. So there might be hope in an update. =)

Checking to see what Tiger is using (strings /System/Library/Extensions/IO80211Family.kext/Contents/PlugIns/AppleAirPortBrcm 4311.kext/Contents/MacOS/AppleAirPortBrcm4311 | grep -i Broadcom) ... (have to ask a friend that has it)

Dec 1, 2007 8:54 PM in response to mreckhof

Same Broadcom chipset and firmware version in my iMac G5.

I may have stumbled onto something. The setups have been relatively stable for hours. There's a little stumble as the computers wake up from a prolonged sleep but since they can only search the preferred list they eventually lock on.

Speed testing is showing much improved and consistent results, comparable to Tiger. Wired is about 5% faster than wireless and the variation from test to test is also about 5%. I'm not seeing the progressive speed drop and degradation with the Dlink that I saw with the Airport routers.

For the record, this setup did not work with the Airport routers. Nothing restored the wireless speed or stopped the speed drop and degradation.

Message was edited by: Ken.

Message was edited by: Ken.

Dec 2, 2007 5:06 AM in response to Ken.

Rereading with a clearer head your message on package loss and the new protocols, I was not showing packet loss. Now this may have been masked by the protocol using the Airports and the throttled bandwidth but then the theory fails with the Dlink. I should be seeing packet loss. I'm not.

What your suggesting is a new form of CRC with a much higher overhead and a potentially unsupported protocol. Wouldn't that, in essence, create a higher CRC failure until there's a fallback? On the other hand, that could mean the new protocol and the routers were mismatched. That could be the case in that the Airport software was updated but the Airport firmware was not.

I can understand that the internet is in a constant state of flux and that changes are on the horizon. However, this particular situation presents a problematic solution. In essence, it's internet throttling. Paying for high speed service and receiving half that (and less) for no apparent (emphasis on apparent) good reason should raise all sorts of alarms.

Packet loss and networks have been around as long as networks have existed. But it's easily diagnosed and dealt with at the server and client level, be that wan or lan. To introduce an invisible and unadjustable layer is unacceptable.

Taking a leap here, this is suggesting that external bandwidth is exceeding internal bandwidth or has the potential to do so. I don't particularly buy into that since that would lead back to buffers.

What I seem to have stumbled on was NAT. Apparently Airport uses NAT, the Dlink does not or it is available and not turned on (in my case). It may be coincidence, maybe not.

We can hash this through ad infinitum but the fact remains that many people are having trouble at the network interface levels and a solution must be found to resolve this across an extraordinary range of variables. People will use what router system they chose and connect as they will to a broad range of server systems.

Dec 2, 2007 5:24 AM in response to Ken.

I probably should have used some different words than packet loss.

What I'm saying is that when everything comes to a halt, I can not get any traffic out on the air. Not necessarily over the Internet - just between my PC and its connected access point.

So that shows a fundamental configuration / incompatibility between the WPA client and server.

By turning airport off and then back on, you're forcing WPA to renegotiate and things start working again, among other things that get reset in the kernel (mtu for the interface, etc.)

I'm not sure what you mean by introducing other layers or protocols. Everything I've mentioned is already being used right now if you are connecting between a wireless client and a server using IP.

Dec 2, 2007 8:04 PM in response to mreckhof

I think were on a similar track in thinking theres an element of negotiation going astray, whether it packets or encryption or whatever.

The hard part is determining what. The only common denominator for all of us is the OS itself.

I've resolved this for myself but it should not have been a problem in the first place. What worked in Tiger should work in Leopard. My hardware did not change, only the OS.

This thread has been closed by the system or the community team. You may vote for any posts you find helpful, or search the Community for additional answers.

Disable Self-Tuning TCP

Welcome to Apple Support Community
A forum where Apple customers help each other with their products. Get started with your Apple Account.