I disagree. While I agree that Winmodems are abominatio… er… impose a subtantial burden on the CPU, that load is unlikely to strain a modern CPU. The “connection speed” is one one many fixed, preset standard speeds, that is chosen by negotiation between the local and remote modem during the handshanking phase. A Winmodem should negotiate exactly the same connect speed, even if the CPU is near full utilization. An overloaded computer won’t be able to “fill the pipe” - i.e. process, compress, and error correct the data stream as fast as the negotiated connections speed would permit, but that won’t affect the “size of the pipe” or connection speed negotiated by the modems.
In simpler terms: when a CPU is burdened, the “connection speed” [a function of of the specific protocols and settings both modems agree to use] is not affected, but “throughput” [the actual measured rate of data transfer over time] does suffer. When you’re very distracted, you might not drive as fast as conditions and traffic regulations permit [decreased throughput], but the [connection] speed limit is determined by physical and social conditions, whether you can maintain it or not.
Since essentially all modems today can handle the full v.90 spec, I think the most likely explanation is the analog circuit, not the digital side.
a) if the two computers use different phone jacks in different places in the house, the difference can be caused by loose connections in the household phone wiring, corrosion in the jack, abused phone cables, etc. Try putting them in the same room and checking what speeds they achieve on the same jack. Many phone cords, in particular, are cheap and degrade over time (and with routine abuse).
b) though todays cheap standard chips and reference designs have specs that should easily allow them to achieve the maximum speed a line is capable of, we must remember that modem connections are analog telephony, and that an analog input or output stages may have been less than optimally tuned, may have drifted slightly from the factory tuning, or may even have been cumulatively damaged (e.g by line transients, overheating, rough handling in shipping, etc.) You can test this by swapping modems, and seeing what connection speeds each achieves on the opther computer
c) EM/RF interference is very real possibility, especially for slot-based modems. The EM/RF conditions inside the case of the newer computer is very different from the inside of the older one. The two modems may also be not be equally tolerant of common-mode interference in the household wiring (phone line or power) or low line levels in the ‘last leg’ connection to your house. If the interference is external, a good conditioning UPS on the power line, or a ferrite core on the phone line (placed as close as possible to the computer end of the cable), might clear enough interference to allow a better connection speed. If it’s internal (either inside the omputer case, or in he design and construction of the card), there are few good alternatives. You could try shielding an internal modem card with thin sheet metal, mesh, or even aluminum foil, but don’t come crying to me if your homemade shield accidentally shorts something, and blows your modem or motherboard.
There is one likely digital explanation, though: many or most modems made today claim to be v92/v44 compliant, but many really aren’t. In some cases, they are really “v92/v44 ready” and getting the latest manufacturer’s driver (if/when the manufacturer cares enough to issue a fully working driver) may bring the slower, but newer modem up to snuff. It’s worth a try.