[mrtew]

I just signed up with WOW for their 4Mpbs service but I can't get over 2Mbps on either of my TiBooks (a 667 and a 1gz.) either plugged in thru the ethernet wire or thru my old graphite AirPort. They guy from WOW came out and replaced the modem and wires etc and it won't go any faster even though the Dell laptop that he had pulled 4.1Mbps easily in my house. He says it must be the Mac operating system that is incapable of going full speed. This can't be true can it? Please tell me that this isn't true... is anyone out there getting 4Mbps cable internet with a PowerBook and 10.3.2? I'm considering a PeeCee :-]

[tooki]

I know it's not a limitation of Panther, because at school (where they have, to put it mildly, insane internet connection speeds) I have achieved actual download speeds of 8MB/sec (that's 64Mbit/sec!) using Safari.

Internet Explorer is appreciably less efficient, and will reduce your effective download speed.

The other thing to consider is this: browsers don't measure the speed the same way! The best way to test your download speed is to use a Java-based speed tester, or to actually take a large test file (large is important!) and download it, but time it yourself, and then calculate the speed based on the total download speed of the whole file. Repeat on the PC.

tooki

[larkost]

a) What are you using to measure bandwidth? Most of the common methods used to measure this are Java applications that have been tuned for windows version of Java, they tend no to perform well on any other OS. File download using IE also only measures post-compression speed, which is very misleading.

b) Did you measure the PC through the AirPort, or raw... ie: could the AirPort be the limiting factor?

c) Mac's are definitely capable of use more than 4Mbits/s... I have seen very impressive numbers out of Gigabit ethernet connections.

d) Even if you were not pulling 4Mbits/s... you rarely use even 1% of that bandwidth while using your computer. 99% of the experience has nothing to do with bandwidth.

[mrtew]

Originally posted by tooki:
I know it's not a limitation of Panther, because at school (where they have, to put it mildly, insane internet connection speeds) I have achieved actual download speeds of 8MB/sec (that's 64Mbit/sec!) using Safari.
The other thing to consider is this: browsers don't measure the speed the same way! The best way to test your download speed is to use a Java-based speed tester, or to actually take a large test file (large is important!) and download it, but time it yourself, and then calculate the speed based on the total download speed of the whole file. Repeat on the PC.

Cool! Thanks! I sure am glad to know it's not the OS anyway. I'll try the file download thing. How big is necessary? Is 4MB enough? Should I go for 10MB or more?

Originally posted by larkost:
a) What are you using to measure bandwidth? Most of the common methods used to measure this are Java applications that have been tuned for windows version of Java, they tend no to perform well on any other OS. File download using IE also only measures post-compression speed, which is very misleading.

b) Did you measure the PC through the AirPort, or raw... ie: could the AirPort be the limiting factor?

c) Mac's are definitely capable of use more than 4Mbits/s... I have seen very impressive numbers out of Gigabit ethernet connections.

d) Even if you were not pulling 4Mbits/s... you rarely use even 1% of that bandwidth while using your computer. 99% of the experience has nothing to do with bandwidth.

Yeah I was using those Java download tester sites with both Safari and Camino... no Explorer here. I do get the same results thru the ethernet and the AirPort. He was using the ethernet for the PC but said that the holdup isn't the Airport because I was getting the same numbers either way. You are exaggerating with that last statement aren't you. "Rarely use even 1% of the bandwidth"??? If that were true it would be almost impossible to tell the difference between 4Mbps and a 56k modem! I can easily tell the difference between between 1 and 2 Mbps... or at least I think I can. If what you say is true paying extra for 4Mbps is a big waste... right?

[larkost]

I'm not really exaggerating about the normal use of bandwidth. The exact same thing applies for the processor on your computer.

Think about web surfing: lets say that an average web page is about 50Kilobytes of data (50 * 8 = 400Kb... this is a bit high for the average page hit, but will do for the conversation), and that it takes 2 seconds to download. So... in two seconds with a 4Mb line (1024 * 4 = 4096Kb) we have used 10%. But lets say that you spend an average of 10 seconds on a page (this is much too small, but I am being generous and more than accounting for bouts of quick page surfing, looking for what you are looking for).

So now our per page stats: 400Kb and 12 seconds on a 4096Kb/s connection.

12s * 4096Kb/s = 49152Kb

400Kb / 49152Kb < 1% (.8138%)

So in this example we are using less than one percent. If you start downloading .iso's or divix files, or gaming, then you will use a much better percentage of your bandwidth. But most people use the web and email most of the time, and this is where I am basing my 1% estimation on.

[mrtew]

Originally posted by larkost:
lets say that an average web page is about 50Kilobytes of data (50 * 8 = 400Kb... this is a bit high for the average page hit, but will do for the conversation), and that it takes 2 seconds to download. ...

Well, I see the flaw in your argument already! If you have a 4mb connection it won't take 2 seconds to download! It will take a hundredth of a second. Right?

Anyway, i don't download games etc, but I do collect Movie trailers, and MP3,s and huge desktop pictures etc. so I'm sure I will notice very much the difference.

Anyway back to the original question... I have tested my connection with many other sites including those without Java apps, and by timing it and by watching the speed in MenuMeters etc and it looks like I am stuck just below 2MB. It's definitely slower than my old 2MB server which always tested over 2MB with the same sites and setup just last week and nowhere near the twice-as-fast that I am paying for. Besides screaming at the company does anyone have any more suggestions, observations, advice, etc?

[cybergoober]

Hmm… you could try to throttle your bandwidth settings:

Perform your speed test(s)
Open Terminal
Type

sudo sysctl -w net.inet.tcp.recvspace=65536
sudo sysctl -w net.inet.tcp.sendspace=65536
sudo sysctl -w net.inet.tcp.delayed_ack=0

Run your speed test(s) again.

Notice an improvement?

If so you may want to add those lines to the end of your /etc/rc file (minus the 'sudo') so that the end of the file shows:

SystemStarter -gr ${VerboseFlag} ${SafeBoot}

sysctl -w net.inet.tcp.recvspace=65536
sysctl -w net.inet.tcp.sendspace=65536
sysctl -w net.inet.tcp.delayed_ack=0

exit 0

Now, I'm not quite sure if things changed in that respect with Panther, but I remember doing this in Jaguar and seeing a rather noticeable difference. I had actually forgotten about this until I saw this thread.

I believe there's actually an app called Broadband Optimizer that does this for you. And general-purpose utility apps like Xupport have this feature.

[OreoCookie]

Originally posted by tooki:
I know it's not a limitation of Panther, because at school (where they have, to put it mildly, insane internet connection speeds) I have achieved actual download speeds of 8MB/sec (that's 64Mbit/sec!) using Safari.

Internet Explorer is appreciably less efficient, and will reduce your effective download speed.

The other thing to consider is this: browsers don't measure the speed the same way! The best way to test your download speed is to use a Java-based speed tester, or to actually take a large test file (large is important!) and download it, but time it yourself, and then calculate the speed based on the total download speed of the whole file. Repeat on the PC.

tooki

8 MB/s is awesome. The best that I ever got was 5.5 MB/s. Connected via 100 MBit ethernet. The only restriction is that the bandwidth is shared by many computers

Here in Japan I got 22 MBit DSL (the fastest one you can get is 100 MBit!), the top download speeds are in the range of 1.5-2 MB/s (not MBit!). Still ok

And I got an iBook 800 …_

[ginoledesma]

Maximum throughput I've gotten on my iBook right now is 6.1MBps using FTP over a switched 100Mbps network. So its definitely not the OS, but you can definitely tweak the settings to improve performance. There are many "Bandwidth Optimizer" tips/scripts/programs out there for the Mac. Download one and see which gives you a setting that's the best you can eke out.

On a special note, however, SMB throughput has never exceeded 2MBps. Which is why for huge files I use FTP (at least it resumes).

Oh, I'd also recommend a nifty menubar app called MenuMeters. It can display the current network throughput among others. Very convenient to know your performance.

[larkost]

mrtew: No matter how fast your connection to the internet, it will still probably take 2 seconds to download a web page. You are ignoring latency, processing time, and total trip bandwidth. Not to mention you are missing the point of the whole exercise, which is the 10 seconds to read the page (hint... that is the majority of the time).

Now, in your examples:

movie trailers: Assuming that they are in some format that can be streamed, you will see a difference in the buffer time (the amount of time before it starts playing), but after that there will be no advantage to the faster connection, because trailers are not typically scaled above .5 Mbit/s of bandwidth. In fact you might not even see any advantage from most servers in the buffering, because they want to keep the bandwidth even.

MP3s: You are probably getting these from other people, and you will be much more limited by their bandwidth, and the intervening bandwidth... no advantage. Most people are on asymmetrical connections, and so this pays in in a big way. As to streaming: this is much less /second than video.

huge desktop pictures: Unless we are talking about >10MB tiff's, then this should not come into the conversation.

"so I'm sure I will notice very much the difference"... I doubt it.

[ginoledesma]

"so I'm sure I will notice very much the difference"... I doubt it.

I agree with this. Even if you have a fat pipe (E1, DS3, OC45), the setup time is still there. You'll only "notice very much the difference" if you were connected to everyone in almost the same way, like that in a local area network. Otherwise, the only clear differences you'd be getting is in sustained transfer rates when downloading from sites with similarly fat pipes.

For example, within our corporate network, we roughly average a measly 13KBps on file transfers from sites like http://mirrors.kernel.org/ and http://www.planetmirror.com/. But the same file transfer on our collocated servers at a datacenter pulls in 822KBps. However the time it takes to load a website (on similar machines) no matter where you are is roughly equivalent, if not negligible at most for the average website. The collocated server will probably pull a lead of one second of "real-world" observation, but that's just about it.

[tafticus]

Just since nobody has mentioned it, 10.3's Activity Monitor has a nifty network throughput graph.

[ghporter]

I have to make a comment here, just to put things in perspective. Network data connections are measured in bits per second, while just about everything else is measured in bytes. That means the raw speed of your 4Mbps connection is really only 500kBps (note the lowercase 'b' indicating bits and the capital 'B' indicating bytes).

Now, let's talk about overhead. Since it's a kind of cable broadband, there isn't a lot of overhead-not too much of your raw bandwidth soaked up by functions that correct errors, monitor connection quality, etc. DSL has some overhead that eats up about 7-10% (on average with good lines) of your raw bandwidth.

One thing that can hurt your speed is the size of your receive window (also known as RWIN). This just tells your computer what size the packets it sends and receives should be. If your computer (and/or router/AirPort) was set up for DSL, which needs smaller packets than cable, your receive window might be set too small. Too small a setting leads to receive packet fragmentation, which, in turn, slows you down.

Getting 50% of your quoted raw speed is not fabulous, but it isn't too abnormal. Also, remember that the quoted speed is "up to" 4Mbps. Don't give up on Macs because of the speed thing-check ALL your settings and see if something's just off where it's supposed to be.

[tooki]

Originally posted by mrtew:
Cool! Thanks! I sure am glad to know it's not the OS anyway. I'll try the file download thing. How big is necessary? Is 4MB enough? Should I go for 10MB or more?

I was thinking more along the lines of a CD image, or some other multi-hundred-megabyte file, ideally from a server as close to you as possible (= as few hops as possible).

Yeah I was using those Java download tester sites with both Safari and Camino... no Explorer here. I do get the same results thru the ethernet and the AirPort. He was using the ethernet for the PC but said that the holdup isn't the Airport because I was getting the same numbers either way. You are exaggerating with that last statement aren't you. "Rarely use even 1% of the bandwidth"??? If that were true it would be almost impossible to tell the difference between 4Mbps and a 56k modem! I can easily tell the difference between between 1 and 2 Mbps... or at least I think I can. If what you say is true paying extra for 4Mbps is a big waste... right?

When downloading files, you definitely can tell the difference. But for web browsing (which consists primarily of lots of small files), latency is hugely important, and high-speed helps latency a lot.

When you load a web page, say, www.apple.com, first your browser hast to communicate with the DNS server to resolve www.apple.com to an IP address. Then the browser sends a request to that IP address, and receives the HTML file. Then the browser must parse the HTML and load all the images on the page, for which it must once again communicate with the server, and possibly with other servers if anything is hosted on another server (which is rather common now, as sites grow large enough that special image servers are set up, as well as load-sharing servers). Each step takes a certain amount of overhead, and during that time, no "payload" (that is, web page content) is being transmitted.

This is, by the way, why satellite internet isn't really all that good for web browsing. The raw download speeds are great, but the latency is horrible, so each of the above steps can take over a second! This means that big files load fast, but web pages with lots of images can take just as long as with a modem (or longer).

tooki