DSL download speed query

[Steve_375]

I have a retired LC 3 running win2k pro with all the latest updates/patches installed. I also have a retired LD Pro running win2k both networked with hp 10/100 nic's running through a 10/100 hub. The LDPro downloads at 1mb the LC3 downloads at 0.5 mb. Both NIC's are set the same, I have tried swapping NIC's, changing ports on the hub, replacing Patch Leads...No change. Help please someone, Thanks in advance

[CA893598]

I'm going to assume here that you're measuring download speeds at the same time of day, from the same website, etc. - there are several factors outside of your control that affect the transfer rate of download.

I've run tests where six consecutive downloads of the same file from the same server within a 15 minute period have shown significant variations in speed (I was trying to determine the impact of encrytion on a wireless link) - you really need to do multiple tests and get an average of the results.

Depending on the nature of your connection to the internet backbone, there is a setting known as the RWIN or TCP receive window setting that can have a major impact on your throughput - take a look at the Navas guide to DSL tuning http://cable-dsl.home.att.net/ for more details

[Steve_375]

Thanks for the tip/link.
Both machines were tested within 2 mins of each other from my isp's site (the other being disconnected first). Someone just suggested this......

All things being equal, the size of the L2 cache would make the difference.
Depending on which version of the P3 processor, it might have had 512 KB
cache running at one-half FSB or 256 KB at full FSB of 100 MHz. OTOH,
the Pentium Pro 200 probably had 512 KB cache running at its full FSB of
66 MHz......

Any thoughts?
Any help gratefully received as this is driving me nuts. Will check that link out though.
Steve

[CA893598]

In my experience processor speed and cache size have very little impact on "download speeds", although that could depend on your concept of "download".

My concept of download is a file transfer from a server somewhere TO my system - it could be via ftp or http, but it would be going to a disk, and not to the screen for display.

If you have other users on the network using the internet link when you do your testing, this will also have an impact on your download rates.

Whilst there ARE local environment variables that COULD impact the download speed - things like disk transfer rates, disk fragmentation, and theoretically -insufficient RAM (if the system is busy paging RAM to disk, it means a whole lot more data is being written to the disk than just the file you're downloading), it's been my experience that these don't actually affect downloads from the internet.

File transfer speeds are usually limited by specific conditions - transferring a file across a LAN typically allows significantly higher data rates than a download from the internet which is usually limited by the slowest link between you and the server you're transferring the file from - if, for example, that server only has a 128k connection - ADSL for example limits the upload rates - you could have an 8Mb download rate and it would do you no good.

There are also variables that impact the internet itself that you have no control over that can change literally from minute to minute - my ISP uses submarine fiber to link to the backbone with a satellite circuit as backup, the routing is dynamic, so it is possible for individual packets in the same transfer to go via either link and of course one transfer could go via fiber whilst the other went satellite. It is this type of issue that dictates the need to do multiple tests and average the results before making a decision.

[Mark Young_2]

Hi,

It may not be relevant, but do you have the NIC in PCI slot 6? That is the recommended slot by HP.

Mark