Why is Winmodem Faster than Hardware Modem?

[arnaut]

I recently upgraded to a US Robotics "Pro Modem" a controller-based PCI modem, from a USR winmodem and was shocked to see my my connection speed drop from 42.6 to 34.6.

It seems that the winmodem, coupled to my P4 2.4, is outperforming the small on-board processor on the hardware modem. That's the only explanation I can see, since everything else is the same.

Any thoughts?

 

[skyking]

that sounds like a hardwware issue or setting issue to me. When your modem first dials in and you hear all that hash of noise, part of it is a negotiation between the two modems to determine a maximum connection speed with few or no errors.
Check to see if the modem tests out properly in diagnostics, and look right in the jack on the back for any funkiness. I have seen some wierd problems at the jack itself.

 

[arnaut]

Thanks for the suggestion, skyking. I'm beginning to think I have a defective modem too.

Everything is okay with the phone jacks, as far as I can tell. The modem diagnostic test is okay too, in Win XP.

USR tech suggested I set a minimum connection speed in the advanced setting in control panel, which I did. The AT command is this: &B0&N22&U22. This gives me a minimum connection speed of 34.6, and is high as it gets. Even so, it usually takes anywhere between 10 and 50 redial attempts to connect at this speed. By changing the numbers in the AT command, you can specify higher or lower speeds. For example, if you entered the number 23 instead of 22, you would connect at a little higher speed, and if you entered number 21 you would get a little lower connection.

 

[Lord Evermore]

That's a pretty large difference, but it's not uncommon. Oftentime one modem can connect at a higher speed, but during actual data transfers, get a lower throughput because that higher speed isn't quite as reliable. Another modem may not make an initial connect at as high a speed, but will be able to make use of the entire connection speed.

Test your throughput using a site such as http://bos.speakeasy.net (they have links to other sites on their network that may be closer to you) with both types of modems.

It's also possible that your P4 is simply able to deal with the traffic even better through software than the controller-base Pro Modem can. Past a certain point, a Winmodem is pretty much indistinguishable from a hardware modem because the CPU speed is so high.

 

[dmcowen674]

Originally posted by: arnaut
Thanks for the suggestion, skyking. I'm beginning to think I have a defective modem too.

Everything is okay with the phone jacks, as far as I can tell. The modem diagnostic test is okay too, in Win XP.

USR tech suggested I set a minimum connection speed in the advanced setting in control panel, which I did. The AT command is this: &B0&N22&U22. This gives me a minimum connection speed of 34.6, and is high as it gets. Even so, it usually takes anywhere between 10 and 50 redial attempts to connect at this speed. By changing the numbers in the AT command, you can specify higher or lower speeds. For example, if you entered the number 23 instead of 22, you would connect at a little higher speed, and if you entered number 21 you would get a little lower connection.

The key here you said it is taking 10 & 50 redial attempts, the new Modem has a physical problem with the connection, this is not a speed issue of software or Hardware Controller.

 

[arnaut]

Lord Evermore, you bring up a good point that the connection speed does not mean you will actually get fast throughput.

I have tested using a number of Internet connection speedometers. McAfee has one that seems pretty good. They are useful at telling you the general state of traffic on the Inernet, rather than your own connection. They do not, after measure your connection to your ISP; they measure your connection to their server, which probably involves a lot more hops. In the middle of the night with not much Internet traffic, you will get good numbers from them. In evening Internet rush hour, I get something like 20k.

So even if you are connected to your ISP at 42k, for example, you may get a slower throughput at any given time than if you were connected at 28k but traffic was very light.

That has nothing to do with my connection speed to my ISP, though, which is determined in large part by my modem's performance. (Although the condition of your telephone line, as well as the ISP's equipment and settings has a lot to do with it too.

Dmcowen, I don't understand your point. Establishing a connection requires a large number of attempts only when I set the minimum speed at 34.6. At lower speeds it connects quickly. But with the winmodem it usually connected on the first try at 42.6. I did not even have to specify a minimum connection speed.

The way I see it there are three possibilities:

either the winmodem is faster, in which case USR has serious egg on its face for selling a hardware modem at twice the cost that gets clobbered by a lousy winmodem. Or, there is something wrong with my modem. Another possibility may be that the ISP is simply not allowing this modem to connect at a higher speed, for whatever reason.

I'm waiting to hear back from USR on this. If it turns out that their on-board processor -- I believer they are called UARTs, whatever that means -- is simply not as good as a P4 winmodem, then they have a lot of explaining to do as to why they continue to use such an obviosly oudated and underpowered chip. There is no justification for selling a hardware modem at a price premium if its performance is subpar.

 

[Lord Evermore]

You misinterpreted my references to slower throughput even with a higher connect speed. Two modems may connect at different speeds, 46k and 40k for example, over the same phone line to the same ISP. However the one that connects at 40k may be better at the negotiation process, and better at handling any transient problems on the line, and so is better able to maintain a full throughput. The one that connected at 46k got a nice fast negotiated speed, but then may not be as able to deal with interference or static on the line, so you end up with lost data and retransmissions.

It may not be enough to cause the modem to renegotiate down to a lower speed, but it may be enough to cause noticeable throughput drops. This was a very common problem when v.90 was still a brand new thing (and before that, with K56Flex and x2). That's why I suggested testing throughput using both modems, to a reliable bandwidth test site (preferably testing at random times over a span of a day or two on each modem to get a good average). The issue of throughput I referred to is related solely to the modems and the line quality, and is unrelated to whether there might be heavy or light Internet traffic at any one time.

Mcafee's test seems to be too short to be a good speed test. It tests a single 600KB file, which even at 40Kbps throughput is only 2 minutes, and compression would reduce that even further. Speakeasy's test does several different sizes of data blocks to test.

The controller on the modem is not a UART. A UART is the circuitry that controls a serial port's input/output. A modem's drivers emulate a UART to give the OS a virtual path to the modem because that's how everything is programmed to consider for access to a modem.

If you put your controller-based modem onto a P2/400, and compared to the Winmodem on the same machine, you would probably notice the controller-based modem performing a bit better. It's not necessarily crap hardware, and USR can't be blamed for charging more money for it because it does cost them more to make it, and many people are willing to pay the extra to say they have a hardware modem.

 

[arnaut]

Thanks, Evermore.

I still do not fully understand your suggestion that a modem connected at a higher speed does not necessarily have higher throughput -- although it does make sense that one need not necessarily follow the other.

So what you are saying is that the winmodem may be negotiating a higher connection speed than the hardware modem, but may not necessarily follow through on that promise with higher throughput. I agree that that's possible, although I haven't run side-by-side tests over a range of conditions that would validate that. I will say, however, that the hardware modem does seem to feel snappier at loading graphics-heavy web pages, although I don't see why that should be, since the graphics processing ought to be handled by the graphics card, no? (I'm using a Radeon 9000.)

But my basic beef is that the hardware modem should be able to connect at the same speed as the winmodem -- and after that, any benefit that you get from the on-board controller making things easier on your system should just be additional gravy.

And why is that the winmodem should be able to negotiate a higher speed with the ISP server anyway? What factors come into play there? Is it possible that the ISP server is putting a lid on how fast a connection it's going to give me?

And I don't agree that USR would be justified selling a hardware modem -- at a premium -- that is outperformed by a cheaper winmodem. I think if its controller was not able to keep up to what the winmodem can do with a host processor, then USR needs to get a faster controller. This would be like Mercedes selling you a $100,000 S-class that gets outperformed by a $50,000 E-class. It just wouldn't fly.

 

[dmcowen674]

You tried Evermore, that's all you can do.

 

[WeeWolf]

I have a few questions/suggestions...
Have you tried a new initialization string for your modem? There are tons of samples available online for you to try.
That modem is a v92 spec modem do you know if your ISP supports v92? They are supposed to be backward compatible but if your ISP is v90 and your winmodem is v90 that may be the difference.
good luck

 

[Kwatt]

I am begining to belive the phone company equipment has a lot to do with it.

I have tried 5 different modems here. 2 hardware and 3 winmodems.
1- usr hardware modem and 1- boca research hardware modem.
1- CL ,1- ActionTec ,and 1- that has no name that i can find.

The USR and BocaResearch never got over 26.4
The Creative Labs 26.4 to 33.6
The NoName 19.2
The ActionTec 26.4 to 44 usually high 30's to low 40's.

I also took the computer and the modems to my parents house and tried them.
The NoName connected at 44 everytime.
The USR and BocaResearch 44 - 48.
The Creative Labs 33.6 usually.
The ActionTec in the 30's

My phone company is SBC.
My parents is SouthernBell.

I kept the ActionTec for myself, put the USR in my parents computer and the NoName in my sisters computer (she is in SouthernBell also).

All I can recomened is to try as many as you can get you hands on. Until you find one that works.(and then by a spare.I did)

Good Luck

Kwatt

 

[arnaut]

Thanks, Kwatt. I think the quality of your phone line, distance to phone company switch, etc. does play a big part.

I'm in a rural area of Ontario, Canada and the connection speed has never been great, usually from 26 to 32, using PC card modems on a laptop, including Xircom, 3Com/Megahertz, and an IBM unit that came with the ThinkPad. Friends in town always have much better connection speeds, anywhere from 45 to over 50.

When I added a new desktop recently, I started off with the USR internal winmodem and was pleasantly surprised to see a connection at 42.6, the highest I had ever seen. Thinking I could get even better with the more expensive controller-based modem, I upgraded and saw my speeds come right back to what I was getting with the notebook.

The thing I would like to find out is why the winmodem is getting a better connection speed, and if it is, is it really making a difference in terms of overall throughput? By taking those connection speedomneter tests, I've noticed that the main limitation is not my connection speed, but the state of traffic on the Internet. In other words, it doesn't matter if I'm coonnected at 42, if the Internet is so jammed that downloads are slowed to a crawl. I've seen speedometer readings of 14k at times. So it doesn't really matter much if my connection to my server is higher.

Still, I would like to find out why it is that the winmodem can connect at a higher speed. There must be some reason. I thought at first it might be driver related because the winmodem was a real dog with the WinXP drivers, and only started getting high connection speeds after I installed the USR drivers (non-digitally signed, btw.)

Wolf, both the hardware modem and winmodem are V92, which my ISP does support. Where can I find those initialization strings you mentioned?

 

[WeeWolf]

Honestly I have not used a modem at home for over 4 years (DSL/Cable) I just took a look around and tons of sites pop up if you search
'initialization strings' on google. You can copy one from one of those sites but I suggest you take a look at your manual and verify your modem supports all of the settings they've placed in the init string. The last time I used a modem at home it was a USR sportster 56k v90 ISA HW modem without a string I connected in NYC to mindspring @42 with the string I connected @48 and my pings went from 200 to 250 in games down to 100-150 (mainly Tribes at the time)
I was able to find a string specifically for my modem but at the time v90 was new and I had to add that to the string to get it to take advantage of it. That's why I suggest you review the settings capable in your manual.