>>>>Some USR modems have small amounts of non-volatile RAM (e.g. flash RAM) that is used to hold configuration settings. It may be that one or more of these settings is configured to a lower-than-optimal speed. If there is a USR utility setup program it might be able to display and/or let you adjust these settings (if they exist).
>>>>
>>>>You could check the details of the property sheets for the modem while it's in each machine; you may find a difference.
>>>>
>>>>Driver *may* have an effect, if an OEM driver is being used. Late-model versions of Windows usually support modems from most major vendors natively without any need for an OEM driver.
>>>>
>>>>Finally, you could check BIOS settings in each computer. Some innocuous-looking settings like "Plug & Play OS installed?" actually have a major effect on the entire system (former should be set to "Yes" for all Windows OSs newer than NT4) by affecting IRQ assignments etc. Again, you could compare the two nominally "identical" machines and check for BIOS differences.
>>>
>>>A sysadmin in my office pointed out what I knew but hadn't thought about, that the "57K" reported by my faster modem has to be the speed at the bus, not over the line. I asked him why I wouldn't see the bus speed on the other machine. He said that there could be a setting somewhere that is different. He suggested that I test the actual speed of these modems using one of these two websites:
>>>
>>>
>>>http://wdc.speakeasy.net/
>>>
>>>http://ciseweb100.cise-nsf.gov:7123/
>>
>>>I expect to check these things out this evening. Perhaps with a result from them, I can go back to Verizon and get them to take a closer look at our phone line, if necessary.
>>
>>OK, I thought you were already reporting hard speed numbers.
>>
>>By "speed at the bus" I assume you mean COM port speed. This should always be set to 115,200 for all modems that are not ancient.
>>
>>The V.90/V.92 standard sets (theoretical) maximum download rate at 56 kilobits per second. In practice I have seen 47-48Kbps under good conditions, I've seen magazine reviews under perfect test equipment conditions of 53Kbps.
>>
>>If you're using a protocol that compresses, such as ZModem, it could be that the value you're seeing is the effective throughput e.g. 28.5Kbps @ 2:1 compression would be 57Kbps.
>>
>>There are also 3rd-party products (and some dial-up ISPs) that use speculative link downloading and local content caching to "enhance the Web experience". These typically work by examining links on a page you're currently looking at, and downloading all or parts of them while the line is not otherwise occupied. If you happen to want to go to a page it's already downloaded, you get instant response. If you are making use of any product like that it could also report "effective throughput" higher than you'd expect on a POTS line, and different from the actual line speed.
>
>I was reporting the speeds that appeared in the tooltip over the icon in the system tray, or the speed announced by the DUN window when it connects. Since it never said anything like 115K, I imagined it was the actual speed.
>
>I don't use Earthlink TotalAccess or anything else like that. I just set up generic dialup networking using the instructions on the support page at www.earthlink.net.
>
>Those two speed test websites did not report plausible numbers. One claimed something like 23K download and 130K upload, the other said 300K and told me I had a DSL/Cable modem. Our sysadmin didn't know of any speed test websites designed for modems.
>
>I still haven't checked the BIOSes but I can do that tonight. But I also don't know now whether the speeds are really so different, or if it's just that the "fast modem" is really just reporting a COM port speed (which is probably what our sysadmin actually said). I can also check speed the simple way, by downloading the same file on both machines.

Another speed test you could try is at http://reviews.cnet.com/Bandwidth_meter/7004-7254_7-0.html

IIRC it works, as you suggest, by getting your machine to download a file from their site.