Andrew Belov wrote:

>On Tue, 04 Jan 2005 21:47:05 +1000, Mike O'Connor wrote:
>
>>I presume that that rate of 20KB/second for version 2.1 of the RTL8029 
>>driver was a joke?
>>    
>>
>No, it was exactly THE problem and the reason why I chose to accept 2.2
>and look into its IRQ code rather than to stay with 2.1.
>
>>What real rates did you get with it and with the new patch 2 version?
>>    
>>
>With 2.1, those 20K per second were an *average* speed resulting from
>packet mishandling on the part of driver. Furthermore, this phenonmenon
>was asymmetrical (we easily get 500K in reverse direction) and was
>accompanied by intermittent hangs, ping delays and other signs pointing
>to a packet mishandling in software.
>
>With 2.2pl2, I get as much as 1070K on incoming and 815K on outgoing
>traffic (when talking to another 8029, on a P4 machine). It drops to
>800K per second in both directions when we talk to a ThinkPad 770Z
>laptop (3COM 3C589 PCMCIA), and degrades further to some 350K if the
>recipient is a 486DX2-50 Siemens laptop (same 3COM card, but a poor CPU
>and only 8M of RAM). All connections are coaxial, this machine is an
>1466 MHz Athlon.
>  
>
I was having problems with T-junctions coming apart, so switched to RJ45s.

>>FWIW, I know the 8029 is slow - running the optimum[?] case ramdrive to 
>>ramdrive gave the following results here with a single large[?] file:
>>    
>>
>
>It is slow by today's standards, but not as slow as your tests indicate.
>10 MBits translate into 1M per second in "optimum" conditions (as seen
>above) and a stable 500-700K per second in real-world setups.
>  
>
>>   15532030 bytes in 83.41 seconds equates to 186213 bytes/sec (181.84
>>   KB/sec)
>>   which is definitely better than 20KB/sec
>>    
>>
>
>Well, I'd say it is "below acceptable" anyway.
>  
>
That's what I figured - but still better than sneaker-net!

>>   26/08/98  10:36      15868           0  PCIND.OS2
>>    
>>
>
>Must be one of these older versions which I'm now trying to avoid. (and
>now I finally understand why 8029s were blacklogged by Russian
>networkers shortly after their introduction.)
>  
>
>>Having now downloaded your latest patched version I'll run it against 
>>the same test here.
>>    
>>
>It would be interesting to know the result.
>  
>
I'll post it back here probably tomorrow night [AEST]

>One more observation that I had concerning 8029s is that any outgoing
>traffic is accompanied by an abnormally high CPU load, which should not
>happen for PCI busmaster cards. This may have various implications,
>including (again) a slower outbound transfer speed on older CPUs.
>  
>

Hi Andrew,

Thanks for the speedy response.

I omitted to mention that the W3C/W4 machine was an AMI Atlas II PCI 
motherboard, with 90Mhz Pentium, and 64MB of 72-pin FP Simms. the other 
end [this m/c] is a 366Mhz Celeron with 384(-8=Video)MB.  Not a lot of 
free RAM on the AMI ~19MB with W3C, ~9MB with W4FP16, before starting 
File & Print Client!

Will give it a trial next boot.

-- 
Regards,
Mike

Failed the exam for
--------------------
MCSE - Minesweeper Consultant and Solitaire Expert
--------------------
[ISP blocks *.exe, *.cmd, *.com, *.bat, *.reg attachments]
[Please use zipped versions of above]



 
Yahoo! Groups Links

 To visit your group on the web, go to:
    http://groups.yahoo.com/group/os2user/

 To unsubscribe from this group, send an email to:
    os2user-unsubscribe{at}yahoogroups.com

 Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 




---