I ran this test as a goof which turned out to be a shocker.

Two systems.

Compaq Deskpro 2000:
* O/S: Microsoft Windows 98SE (Special Edition)
* CPU: Intel Pentium 166MHz MMX
* RAM: 64MB @ 66MHz
* Modem: Internal 32-bit PCI Lucent LT WinModem V.92 (56kbps)
= max sustained throughput for WinModem 8.2kbps

Custom built:
* O/S: Linux distro's
* CPU: AMD AthlonXP 3000 (2171MHz) @ 333Mhz
* RAM: 512MB @ 400MHz
* Modem: Internal 32-bit PCI Lucent LT WinModem V.92 (56kbps)
= max sustained throughput for WinModem 5.6kbps (10.1 with FC3 for one ftp site only: ftp.freebsd.org)


Same phone line cord with no y-adapter, same modem. No IRQ conflicts. I made sure there was no sharing of IRQ's at all. The Linux drivers were either the precompiled linmodem drivers, the deb install ltmodem driver or source code I compiled myself with the ./build_module;./ltinst2./autoload syntax. Kernels: 2.4.* & 2.6.*



Even with the custom built machine reniced for the processes, the best that the Linux distro's could get never matched the wimpy CPQ sys with Win98SE. If a LT WinModem with it's 8.26 DSP is not a real hardware modem, then how come Linux can't beat Win98SE considering the hardware advances?

Could it be that the winmodem preforms better in windows? Man who would have thought. It preforms better because it is not hardware based and the linux driver is not as efficent as the windows driver. Since its a win modem it uses proprietary software to control the modem, and with out chip specs you will usually have a loss in preformance with a thirdparty/hacked/linux non proprietary driver. I have a modem blaster isa card that preforms better in linux, but it is a hardware modem. Id say sell the one you have aand buy a hardware modem.

And so the performance of my native FC3 product to ftp.freebsd.org is a result of a non-compatible Linux driver. Ok, gotcha.