One reader also commented:

" Apple's "Broadband" Tuner is for High-bandwidth *high-latency* connections. (apple page mentions FiOS high speed connections with high latency, although a FiOS user report below said it's *not* high-lacency, and that he saw no benefit from it with his Verison FiOS connection.-Mike) Cable and DSL are *not* high-latency connections. You are likely to see poorer performance from Apple's broadband tuner on Cable and DSL connections versus the default settings.
An example of a high-bandwidth, high-latency connection is satellite.
-Christopher F.
Macintosh Engineer, Platform Client Services
Infrastructure Transformation Services "

Not sure I'd call my DSL connection "low latency" (LOL) and with wireless networking in the mix (home network) here that is higher latency than a direct Ethernet connection. (Not to mention the delays due to server load/response times, etc.) But the larger buffers don't always help (see reports below) and what's interesting is that several people reported higher rates after installing/uninstalling it that they had with their original settings.

"The comments Re: what technology is 'high latency' and what is not 'high latency' is irrelevant. It doesn't matter if the connection is DSL, Cable, FiOS, Satellite, SONET, metro-Ethernet, whatever - if the time it takes to get from point A to point B is "high", the default TCP window sizes and other buffer settings aren't large enough to keep the pipes full. A OC-48 SONET link can have just as much latency as a satellite link if it's long enough, so if you tie two backup systems between NYC and San Jose to each other, you better tune or you won't get the bandwidth that you think you should.

This all has to do with the Bandwidth Delay Product in which the number of bytes of data that need to be kept in flight to compensate for latency is equal to the bandwidth of the link (or the amount of throughput I want to achieve - I don't HAVE to fill the pipe) * the amount of time it takes to get there.

Example.
I am an OS X user on FiOS. My ping time between here and www.google.com is 42ms. Not because Google sucks or they're not 'well connected' but because we can't speed up the speed of light.

For many operating systems, the default TCP window size runs someplace around 17KB (depending on OS, service packet/update level, etc.). The MAX TCP Window Size without enabling window scaling is around 64KB. With window scaling, you can use larger values.
If I put these numbers together:
Bandwidth(bits per sec)*Delay(sec)=BitsInFlight(bits)
... therefore ...
BitsInFlight/Delay=Bandwidth
... therefore ...
(17000*8) / (42/1000) = 3,238,095 bps
So unless I can increase my window size to greater than 17KB, I will not be able to transfer more than 3.2Mbps over my 5 Mbps FiOS service. If either OS doesn't support window scaling, and you were limited to a 64KB window, the max speed would be somewhere around 12 Mbps - which as long as the systems will increase their window sizes, ultimately, you will get max speed out of pretty much any home broadband system without tuning. Take the delay up to 120ms though, and you're back down to 4 Mbps even with these 64KB windows.

That's where the apple tool comes in. It's similar to Dr. TCP or regedit on Windows or good 'ol vi on unix, only it's setting static values rather than letting you change the settings. For that, you'll need to modify /etc/sysctl.conf yourself.

With the values provided by the tool, running through the math, on the same 42ms that I get to google, download rates should be:
net.inet.tcp.sendspace=131072 = 24,996,095 Mbps
net.inet.tcp.recvspace=358400 = 68,266,667 Mbps
Notice these buffers are greater than 64KB. Therefore, the OSes must support window scaling for these settings to truly help.
There is a lot more to tuning (application buffers, sizes of the payload the app is putting into the packets, etc.) than just these two values, but these will help in many cases.
In the long run you'll find that this is just a fraction of the battle. It may not be possible to tune out the latency without changes to the application itself - which is why some folks may not notice any increase in speed.
Mike "

Reader Reports
If you try Broadband Tuner and run tests before/after (preferably avg of several runs), let me know what you've seen and your system/broadband info.

(added 12/9/2005)
" My personal experience with BBT on a campus network (I'm on gigabit ethernet) as well as at home on Cox's 5MB cable broadband service is that BBT doesn't actually help...it hinders.
Safari was visibly pausing when I clicked a link as it "thought" about what it was supposed to do. As soon as I uninstalled BBT, performance returned to normal.
I guess neither system uses FiOS.
-John "

See below for a FIOS user report that said it didn't seem to help (and disputed the comment in the apple page on high latency).

(added 12/8/2005)
"I have a Verizon FIOS connection at home (fiber optic to the home, 15 Mbps down / 2Mbps up), and since Apple's description of the Broadband Tuner app specifically mentions FIOS connections, I thought I'd try it. However, FIOS is not a high latency connection -- ping times on FIOS are under 5Ms to on-net servers and no more than 10ms to any good server on a tier-1 connection.

Tested with Verizon's on-net speed test server:
http://infospeed.verizon.net/fttp

Safari 2.0.2 / Empty cache before each run (speeds in kbps)
BEFORE Installing BB Tuner
Three runs: 147843 / 156862 / 155172

AFTER Installing BB Tuner
Three runs: 156862 / 156182 / 155171

So, no difference. I'm un-installing. If BB Tuner does what it says, it is NOT for FIOS connections, since they have a very LOW latency (which is what you'd expect from a passive optical network).
Later, Mike "

What Mike is referring to is Apple's BB Tuner page comments that say:
"The Broadband Tuner allows you to take full advantage of very high speed FiOS based Internet connections that have a high latency."

This report included some 'real-world' usage comments:

(added 12/1/2005)
"We are using ADSL iinet in Australia on airport express cards. Upon installing broadband tuner we have received a very noticableand consistant speed increase.
Pages load more fluidly and heavily imaged sites like say, google images thumbnail screens load tremendously fast.

Tested on -
G5 DP 2.5 (3 gig ram),
1.67 Powerbook "15 (2 gig ram)
Emac G4 800 mhz (192 ram).
All macs running latest OSX updates, airport express cards, billion wireless modem and safari.
We are in a heavily used internet suburb where congestion is normal, it now feels like when internet broadband was rare and you were the only one using it in your suburb...Ahh the memories...
Tested it on constantly refreshing pages like Realestate.com.au, carsales.com.au, Hotmail.com and Google images.
-Adrian "

(added 12/1/2005)
"before 4100 avg
after 3100 avg
uninstall 2700 avg!!!
Ran the uninstaller and rebooted several times and I cannot get back by original speed.
There must be a way to fix this manually. :-( Do you know?
G5 dual 1.8g june04, 2.2 gigs ram.
Videotron Cable Extreme Edition ...max speed 6.5 Mbps down/900Kbs up
thanks, D. "

I don't know what your original settings were but (supposedly) uninstalling should have restored the defaults. (And I'd not rely on just some web benchmark page run to be the final word on performance... for the reasons mentioned previously here. You could also try some of the other benchmark sites to see what they show.) I'm not sure if reapplying an OS X combo update would change the settings (or help) - if in reality (not just a benchmark page) performance is less than before.

(added 12/1/2005)
"I have a fiber connection promising 10mbps up and down. I did not check with the usual bandwidth checking systems because normally they are lacking bandwidth or are simply inaccurate. So I went to an internal server and started downloading a compressed file from there.
In a nutshell; with or without BB Tuner the speed is of ca 800KBps. So no effect here.
My config is iMac iSight with airport extreme ->Belkin 56g router->Fiber. Thanks, Paolo Z. "

(added 11/30/2005) "(Mike said - But I've always wondered how consistent any internet benchmark is due to constantly changing server loads, net traffic, number of hops along the way, etc...)

Hi Mike,
Right you are, trying to get "repeatable" results from Java and/or Flash based tests at many websites is pretty difficult. I first ran into this problem some years ago when Sustainable Softworks brought out IPNetTunerX and I was beta testing for Verizon DSL and Roadstar, a WiFi ISP. With over 800 days of testing into many websites, http://www.dslreports.com/archive/?zip=2013&start=Search including the popular Broadband.com system with its tests from various websites. I realized I was beating my haed against something I couldn't control.

I now use my local server with 10, 100, and 1,000Mbps connections and uncompressable test files of various sizes to see what each browser and Java version is doing, including tuning TCP/IP via IPNetTunerX. Results are now repeatable. :-)
FWIW.
Best regards, -Rod Paine "

(added 11/30/2005) "Kernel panic with Broadband tuner 1.0
I installed it and tried to download 4 large files averaging 500-600 MB in size..was averaging 800k/sec on each file and after a minute, I got a kernel panic (black screen). I had to unplug my G5 (single CPU,1.8GHz, 1.5 gigs ram running Tiger) to make it run again. Rebooting by pressing the start button also caused a kernel panic on restart. In case it might matter, the files were being copied on a semi-full 250 GB USB2 external hard drive . My connection is 24 MBauds downstream/ 2 Mbauds upstream.
I uninstalled it today and no problem after after repeating the same process above.
cheers, Frank "

(added 11/30/2005) "Mike,
It seems the high-bandwidth/high-latency of Verizon's FiOS is the reason Apple released Broadband Tuner 1.0. Take a look at the "updated" description on Apple's download page. I would not recommend users installing if you have a high-bandwidth/low-latency connection, such as TWC Road Runner high speed online.
Thanks...Lee

About Broadband Tuner
The Broadband Tuner allows you to take full advantage of very high speed FiOS based Internet connections that have a high latency. The installer tweaks some system parameters.... "

(added 11/30/2005) "Just a heads up on some additional results on The Broadband Tuner App -
Tests ran at http://www.speakeasy.net/speedtest/

Without BB Tuner:
Dallas: 7843/690
Chicago: 7862/693
Atlanta: 7899/693

With BB Tuner:
Dallas: 7857/701
Chicago: 7863/705
Atlanta: 7891/699

All speeds in kbps down/up.
Doesn't seem to be too much of a difference here as far as the stats go. Page loads seem to be a bit snappier, though.

Running on a Sawtooth G4/1 Ghz (OWC Chip Upgrade) with a Comcast 6 Mbit down/768 kbit up connection.
JOHNb, Detroit, MI "

(added 11/30/2005) "Have a TiBook 800 on 1.5M DSL over older Airport. I don't really meet the suggested requirements for installing the Tuner, but figured it's uninstallable, so I'd see what it does.

Before Tuner, C/Net showed ~1100down over a few runs, and the broadband test over at DSL Reports showed ~1225down/240up. Installed the Tuner, C/Net dropped to ~725down, but the DSL Report test showed no change. Uninstalled - C/Net report went back up to about ~1100/ down, DSL Report stayed unchanged still. Don't know why there was no change in the javascript one from DSL Report, but the C/Net one showed a pretty marked drop with the tuner installed. I didn't test any real-world throughput, since I don't do a lot of large scale download/uploading via the net.
-bb "

(added 11/30/2005) "My setup:
Cable modem (Comcast, Pennsylvania) -> wired gigabit ethernet switch (D-Link GDS-1005)-> G5 & WinXP PC

3 runs averaged (speakeasy.net/speedtest): (uploads capped typically)
With Apple's BB Tuner: 6090 kbps down, 356 kbps up
Without BB Tuner: 6094 kbps down, 335 kbps up

For comparison, my WinXP AMD AthlonXP 3200+ box w/ Intel gigabit PCI card:
6176 kbps down, 339 kbps up
Bottom line: No difference for me.
-Cadaver "

(added 11/30/2005) " I'm not sure how others have calculated their findings on the effect of the Apple Broadband update, but I hope my data adds some clarity to the situation.
I ran the CNet speed test 10 times in a row before installing the update (after a fresh restart).Ê Then I installed the update and ran the test 10 more times.Ê Then I uninstalled the update and ran the test 10 more times again.Ê I then threw out the lowest value for each condition and then calculated a simple mean and standard deviation. Here are the results for the nine remaining trials:

Off	On	Off Again
4373	3863	3849
4391	2082	4373
2137	1927	4387
2622	3809	2021
4396	4396	2049
3053	3900	4396
2592	4406	3823
4410	4415	4410
4373	1953	1876
Mean	3594	3417	3465
Std Dev	970	1099	1136

Basically, the update had no statistically significant effect on my system.Ê For the record, my system is a 2x2 G5 tower Rev A with 3GB RAM connected to the internet via a 5Mbs cable modem connection. Hope that helps someone.
Ê Steve "

(added 11/30/2005) "My results on a G4 1.42 DP Mac OS X 10.4.3 with Airport Extreme and a Linksys Router, Comcast cable:
Before Update (Speakeasy) --- 3543 down/ 344 up
After Update (Speakeasy) --- 4237 down/ 344 up
Regards, John O. "

(added 11/30/2005) "I have used most of the ones you listed but this seems to be the best one with the most variations of testing
http://performance.toast.net/
-Gary H. "

(added 11/30/2005) "Hi Mike, The installer puts a file called "sysctl.conf" into the /etc/ folder for the system to read when it boots. Here's it's contents:
#
# Tuning network for broadband
#
# START
kern.ipc.maxsockbuf=512000
net.inet.tcp.sendspace=131072
net.inet.tcp.recvspace=358400
# END

And now my personal "tweaked" version:
#
# Tuning network for broadband
#
# START
kern.ipc.maxsockbuf=512000
net.inet.tcp.sendspace=131072
net.inet.tcp.recvspace=358400
net.inet.tcp.delayed_ack: 0
net.inet.udp.recvspace=73728
# END

I added these settings to help with FPS games and other UDP app, like video conferencing and file sharing programs. Others can do the same by doing the following a a Terminal:

# sudo pico /etc/sysctl.conf

It'll ask for your admin password then let you edit the file as any text editor. Use control-O to save and control-X to exit. Reboot and you're away!

I have since found this to be the best way to get better speeds yet. I had to remove any other third party settings first to take full effect, which I suggest to anyone else trying this. i.e. remove third parties apps "Network Optimizers" BEFORE installing the Apple one and make the changes there after. Always reboot in between installs too!!
Regards, Bobbi "

(added 11/30/2005) "1.1 Gig B@W G3, Earthlink ADSL 1260/600 (+-50) before, no change after.Ê
-Thomas C. "

(added 11/30/2005) "Positive results. 20% faster down
PB G4 17
Verizon 3.0/768 DSL
Linksys B wireless
Used Megapath speed test

Before: 2250-2494 up, 691-696 down. 5 tests
After: 2747-2840 up, 709-712 down. 5 tests
(I asked if the above was reversed - normally uploads from the user system is capped/much lower than downloads from the net-Mike)
-Larry B. "

(reports from 11/29/2005 follow)

" I have a nominal 1200 ADSL line and I am using AirPort with an iMac G5 1.8 GHz, 1 GB RAM & Mac OS X 10.4.3
Before: average 1258
After installation: 830
After uninstall: 1260
(averages of 15 measurements using http://multimedia.terra.es/herramientas/velocidad/)
Conclusion: not useful for me.
-Julian N. "

" I have Verizon FIOS fiber optic internet. (15.0 Mb/s DL - 2.0 Mb/s UL advertised) and have seen a increase in tests from the sites I regularly test at:
http://miranda.ctd.anl.gov:7123/
and
http://ciseweb100.cise-nsf.gov:7123/

Before the update I would get an average test result of:

TCP/Web100 Network Diagnostic Tool v5.2.1e
click START to begin
Checking for Middleboxes . . . . . . . . . . . . . . . . . . Done
running 10s outbound test (client to server) . . . . . 1.84Mb/s
running 10s inbound test (server to client) . . . . . . 13.32Mb/s

after the optimizer the average results were:

TCP/Web100 Network Diagnostic Tool v5.2.1e
click START to begin
Checking for Middleboxes . . . . . . . . . . . . . . . . . . Done
running 10s outbound test (client to server) . . . . . 1.95Mb/s
running 10s inbound test (server to client) . . . . . . 15.32Mb/s

Not bad for 50 bucks a month. ;) Plus I have noticed an increase in page loading performance as well.
Brian J.
Blackstar Productions
http://www.blackstarproductions.net "

" I have an iMac G5 20" Rev A 1.8 ghz with SBC Yahoo! DSL. I installed the Apple Broadband Tuner without running CNET (test) first and got something like 709 kbps (which was pretty favorable in the mid DSL range). Since I had read reports on uninstalling for additional improvement I decided to do it and see what happens. I went up to 724 kbps. Then I reinstalled and dropped to 77.1 kbps!!! Finally I uninstalled the second time and got over 1000 kbps. It seems to have settled to around 736.9 kbps. Very strange results, to say the least. I hope I don't get any more radical drops in speed...
Thanks, Chris "

Might also want to try another test page (and run several times for an avg.)

Another report of better speed test results after uninstalling. (See below for earlier reports on this.)

" went from 2250 to 3900 after UN-install. What a great application!! LOL.
Later, SCB "

" Just tried it on a Dual 2.5 G5 w/ 8 gigs of ram.
Before the installer 3825.7 kbps. I ran the installer and it dropped down to 1468.2 kbps! After I uninstalled the Tuner my connection went back up to 3926.9 kbps.
Needless to say this is one update I won't keep on my G5.
-Jack T. "

" I upgraded last month from a ADSL 0.5/0.5 to a 8/0.8 and was disappointed with my download speed after I applied this upgrade my download speed almost dubbled (My max speed should be around 6 MBit/s as I live 3.1Km (a little less than 2 miles) from my phonestation)

Test with TPTest 3.03

Before Broadband Tuner:
Receiving:
Max TCP: 1.86 Mbit/s
Max UDP: 6.73 Mbit/s
TCP/UDP: 27.7 %

After Broadband Tuner:
Receiving:
Max TCP: 3.99 Mbit/s
Max UDP: 6.65 Mbit/s
TCP/UDP: 60.0 %
-Mikael F. "

" The results from the cnet bandwidth test with 8Mb ADSL Dual g5, direct internet (no wireless)

Installed: 665
Uninstalled: 1847, 552, 64, 1831, 509
Installed: 162, 637, 634, 681, 679
Uninstalled: 693, 560, 591, 773, 564
Not conclusive I would say. Seems to be on average a bit higher with installed. Doing this from Finland over the seas...
-h "

" Just to let you know that Apple have released a broadband tuner for high latency high bandwidth internet connection, such as cable internet.

I ran it on my PB G4 15in 1.25ghz and it has radically improved the lag time that I was experiencing loading web pages and downloading files. HIGHLY recommended for anyone with a broadband connection.
(he later sent info on his net config)
I am using Airport Extreme. I haven't tested bandwidth, however web pages are very obviously quicker to load. Browsing was surprisingly slow and laggy given the good connection to the APE and the 5meg cable, and the update seems to have fixed those issues with my setup.
Daniel"

Matthew called to say he initially tried it on his Pismo (PB G3) with original Airport card (802.11b) and said it hurt performance (per Cnet bandwidth test) but uninstalling it boosted performance. (near the end of the installer there's an option to remove it and restore defaults) He later sent an email after trying it on his iBook G4 w/Airport Extreme card:

" Broadband Tuner results on iBook G4
Same as the Pismo - a major drop in performance from 3746kbps to 1430, uninstalled back above but only to 3791. so it seems to have helped more to uninstall. Go figure.
(Just for the record I asked for details on his broadband/network config-Mike)
Airport Express on Comcast cable, original Airport card in Pismo and Extreme card in the iBook G4 1ghz. Again Pismo went from 1984.4kbps to 1354 and unstaller took it to 2202.
-Matthew "

When he first reported that uninstalling helped (with higher rates than before installing it) I wondered if he'd previously used a 3rd party MTU/tweaker or if what he saw was just differences in run/run on a specific speedtest server.
However another reader said he also saw higher results after uninstalling it (not sure why if it just resets to the same original settings - unless the original settings were for dialup.) Repeated tests/avgs would be better than one test however and I've always wondered how repeatable internet tests like this are due to changing server loads, net traffic, hops along the way, etc.

" I just performed Cnet's bandwidth tests before and after installing Apple's Broadband Tuner. Like others have said, it seems to have hurt performance. Before installation, the Cnet tool reported my connection speed as 2022.8 Kbps. After installation, it dropped to 1475.6 Kbps. This is on a G5 Dual 1.8, 2.25 GB of RAM using Comcast High-Speed Cable.
One more piece of information about this test that I found mildly interesting. After uninstalling the Bandwidth Tuner, the Cnet tool reported my connection speed as 4387.1 Kbps, over twice as fast as the first test I performed before installing the Apple Bandwidth Tuner.
My system is only a couple of weeks old, and I cannot recall installing any third-party bandwidth tweakers on this system, so I can't explain where this performance boost came from.
(I asked if he repeated the test several times (to avg results).-Mike)

Regrettably, I did not perform the test multiple times *before* installing the tuner the first time. In fact, I just ran Cnet's tool several times (without installing the Broadband Tuner) and received a range of results. Their tool may not be very reliable for my location. I re-tested, this time using Speakeasy's tool at http:// www.speakeasy.net/speedtest/. It appears to provide more consistent results than the Cnet tool. I ran each set of speed tests several times and averaged the numbers.

Before Broadband Tuner Install
Avg Download Speed: 5960 kbps
Avg Upload Speed: 310 kbps

After Broadband Tuner Install
Avg Download Speed: 6049 kbps
Avg Upload Speed: 337 kbps

After Broadband Tuner Un-Install
Avg Download Speed: 6072 kbps
Avg Upload Speed: 309 kbps

Broadband Tuner seems to have a negligible affect on my Comcast Cable Modem connection.

-Brett "

" Broadband Tuner speeds up my 933 mhz G4 iBook with Airport Extreme to Belkin WAP to Belkin router to Belkin modem on Comcast.net. Was 5790 kbps download/345 upload and is now 6339 download/349 upload.
-Thomas S. "

" I'm on a T1, and ran the Cnet tests before, during (?) and after installation of the Broadband Tuner. I ran each config five times, threw out the high and low, then averaged the three remaining.
Results: Before 1302.9; During 1332.6; After 1316.8
There was no significant difference here. I suspect it only helps with higher bandwidth connections. T1 is not that fast by today's standards.
-Steve "

If anyone else tries this and tests before/after - let me know what you've seen and your system/broadband info.
Here's a page at links to several broadband speed test pages, although I know there's many others also.)