Date: Fri, 30 Jan 2004 12:30:35 +0100 From: Philip Nienhuis <[EMAIL PROTECTED]> Subject: Re: [LIB] Networking problems
Matt Hanson wrote: > > Date: Fri, 30 Jan 2004 03:41:04 +0000 > From: "Matt Hanson" <[EMAIL PROTECTED]> > Subject: Re: [LIB] Networking problems : <snip> : > >- Xircom CEM33 (10 Mbps / full duplex): 2 min 18 sec - Conceptronic > >CON100TCL (100 Mbps / full duplex): 0 min 47 sec - IBM 10/100 Etherjet (100 > >Mbps / full duplex): 1 min 02 sec (oh and yes, I got a switch in between), > > > >so the fastest effective rate I got is in the order of 1.4 >MByte/second. > >I'm afraid that's as fast as one can get on a L110, >probably slower on a > >L100 (lower CPU speed), not to mention L70. BTW It's remarkable that the > >32bit Cardbus IBM Etherjet (actually rebranded Xircom) is actually slower > >than the 16bit Conceptronic. > > Wow... that's impressive relative to what I'm getting. I'm fuzzy on the > math again though. You said the best rate you got was 1.4MB/s. I know > there are various ways to report file size, so I guess I'm just picking nits > here. But you said you transferred 79MB of data in 47 seconds with the > Conceptronic card. 79MB/47s = 1.68MB/s by my math. Oops -- my fault, I just estimated, sorry (I should have blown the dust of my HP41 and *calculate* it). > Converting to Mbits/sec: > > 1.68MBytes/s x 8bits/byte = 13.4Mbits/sec > 1.4MBytes/s = 8bits/byte = 11.2Mbits/sec > > Either transfer rate is more than double what I'm seeing. That was for the > Cenceptronic card. The IBM card weighs in at 10.2Mbits/sec... not shabby in > it's won right. Your Xircom is seeing about the same at 4.58Mbits/sec as my > Xicom & Linksys. To clarify things a bit: - I reported actual file (data) transfer speeds, in MByte/s - If talking about Mbits/s one should include network overhead (packet headers, but also network communication outside of any data transfer like broadcasting, master browser wars, messaging, etc.) - On one side I have a very fast desktop PC. I do not remember your situation in detail, but it may be that Lib-to-Lib communication suffers twice from slow CPU speed. Be advised that I am by no means a network guru, I just know (and not always for sure) about the problems and tweaks I encountered myself. BTW from OS/2 Warp tuning I know that there are a LOT of tweaks which may have much influence on actual throughput speed. As for OS/2 I know where to look around (e.g., the \IBMCOM\PROTOCOL.INI file for the network & MPTS for the NIC itself) but w.r.t. Windows it's partly in the NIC advanced setup and the rest probably buried deep down in the registry. Access to settings on Linux is again much easier. Advise 1: (probably you already did so:) check for full duplex. Advise 2: search Internet for Windows networking sites, I'm sure a lot of useful tips can be found. Perhaps Helmig's site for a start? (http://www.wown.info/) <another snip> : > >NIC brands do make a big difference, as some take 60-80 % CPU time, > >especially in 100Mbps mode, while other brands do a lot by themselves > >and leave more CPU time for other processes. Cheap NIC cards rarely > >perform as well as e.g. 3Com, Xircom, .... > >It is really a matter of value for money here. > > So are there NIC cards with onboard controllers like the Firewire amd USB2 > cards Raymond mentioned? "Onboard controller" may not be the right wording but the general idea is quite OK. The separation between low-level stuff processed by a NIC's hardware and the higher-level stuff processed by a host (e.g. Windows, incl. NIC drivers) is quite arbitrary. The more which is emulated in software, the cheaper the NIC can be (just like Winmodems), but at a price of higher CPU load. And then again, poorly written drivers can/will also degrade performance (and not just a little bit). I suspect that's the cause of the IBM Cardbus NIC performing less than a plain 16bit NIC in my case. (BTW speaking of drivers, did you apply the network upgrade thingy for Windows 98? I believe it is called DUN14-98.exe or so, search Microsoft.com for the proper one (MS has recently extended support for Win98 & ME up to 2006 so it should still be there). Maybe that will help a bit, and as a lot of bugs have been solved it is an advisory upgrade anyway.) Philip ************************************************************** http://libretto.basiclink.com - Libretto mailing list http://www.silverace.com/libretto/ - Archives -------TO UNSUBSCRIBE------- Reply to any of the list messages. The reply mail should be addressed to: [EMAIL PROTECTED] - Then replace any text on the message's subject line: cmd:unsubscribe --------TO UNSUBSCRIBE DIGEST------ Do above but with this on subject line: cmd:unsubscribe digest **************************************************************
