From: google@holzapfel.biz   
      
   I did two more benchmarks.   
   This time I'm on my PC 750 (PCI/MCA) @ 400 MHz again, running Linux 2.2.17,   
   using Netio 1.11 (because I don't have GLIBC3 there) and running the RX   
   direction, so I'm sending data from a modern multi-core, multi-GHz system to   
   the 9-K:   
      
   NETIO - Network Throughput Benchmark, Version 1.11   
   (C) 1997-1999 Kai Uwe Rommel   
      
   TCP/IP connection established.   
   1k packets:     6513 k/sec   
   2k packets:     5564 k/sec   
   4k packets:     5805 k/sec   
   8k packets:     5256 k/sec   
   16k packets:    6782 k/sec   
   32k packets:    7308 k/sec   
      
   As a comparison, I am using the exact same setup, but sending data to a 9-P   
   card, which is the PCI variant of our MCA 9-K.   
   It has the same PCnet ethernet chip and chip revision, accompanied by the same   
   amount and speed rating of on-board memory.   
   The Linux drivers of the PCnet (9-P) and San Remo (9-K) are 99.5 % identical   
   with the exception that the 9-K driver tunnels all control I/O accesses   
   through the ASIC, which is not speed relevant.    
   All driver parameters like RX/TX queue lengths, interrupt rate, burst   
   settings, error handling and all other configuration of the PCnet chip's   
   registers are exactly the same.   
   So from the hardware point of view, the only difference is that in the 9-K   
   case, the hardware access is tunneled through the PC 750's PCI-to-MCA bridge,   
   and then through the 9-K's MCA-to-PCI ASIC bridge.   
      
   This is how the PCI 9-P performs:   
      
   NETIO - Network Throughput Benchmark, Version 1.11   
   (C) 1997-1999 Kai Uwe Rommel   
      
   TCP/IP connection established.   
   1k packets:     11433 k/sec   
   2k packets:     11422 k/sec   
   4k packets:     11533 k/sec   
   8k packets:     11543 k/sec   
   16k packets:    11478 k/sec   
   32k packets:    11404 k/sec   
      
   So I dare to conclude that the system itself, Linux, the general driver   
   structure (no matter if 9-P or 9-K), the interrupt rate, the data rate, DMA   
   and CPU usage and of course the PCnet chip itself are generally capable of   
   saturating a 100 Mbit Ethernet    
   link under the exact same operating parameters.   
   The only difference is that the 9-K is attached to the system through the two   
   bridge chips in my case.   
      
   There might be some parameters in IBM's original AIX driver that they adjusted   
   inside the ASIC or PCnet to make those two work better together to squeeze a   
   little more out of it - but generally (see 9-P), there is no obvious   
   misconfiguration of the PCnet    
   chip.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)