Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!mcdchg!ddsw1!karl
From: karl@ddsw1.MCS.COM (Karl Denninger)
Newsgroups: comp.sys.amiga
Subject: Re: Re: Re: A3000UX Seems Fated (Kill file alert!)
Summary: RS/6000s stink on Ethernet I/O
Message-ID: <1991Jan07.054805.13284@ddsw1.MCS.COM>
Date: 7 Jan 91 05:48:05 GMT
References: <1990Dec25.045537.13517@NCoast.ORG> <1990Dec25.234322.836@ddsw1.MCS.COM> <17091@cbmvax.commodore.com>
Organization: Macro Computer Solutions, Inc., Wheeling, IL
Lines: 59

In article <17091@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes:
>In article <1990Dec25.234322.836@ddsw1.MCS.COM> karl@ddsw1.MCS.COM (Karl Denninger) writes:
>
>>NEITHER the Amiga 3000 or the ISA machine will make a good server.  For that
>>you need a real system bus, such as VME.  
>
>The A3000 bus, like EISA and MCA, isn't all that slower than full 32 bit VME.
>VME can kick up to about 66MB/s (as I recall).  Zorro III theoretical limits
>(the numbers the VME, EISA, MCA, and NuBus people always give you) are 50 MB/s
>for single cycle, 150 MB/s for burst cycles.  Typical values depend on the bus
>master -- the A3000 bus master implementation itself hits around 20 MB/s for 
>single cycles.  The A3000 local bus (where the 68030 and SCSI DMA sit) runs a
>minimal 20 MB/s to memory, which maxes out at around 44 MB/s (on-page burst
>to Fast RAM).  CPU slot expansion memory can go considerably faster.  

Ah, but it depends on how good of a job you can do interfacing with that
bus.  The VME people have a lot of experience with it, and they do a hell of
a job.  The "PC" people, whether ISA, EISA, MCA or Amiga, all seem to be
more interested in shaving $5.00 off the price of the card than they do
performance.

>>I have no idea on the 3000UX -- but I'll bet ethernet performance will be 
>>one of the sore spots.  Hell, ISA Ethernet is faster than Ethernet on MCA 
>>systems!
>
>Hmmm, I'm sure all them folks with RS/6000 systems would be a little surprised
>by that comment.  

Oh would they?

RS/6000s use 3Com boards for Ethernet the last time I looked.  3C523s.  
They are DOGS.  I have worked with one, and it did something like 
150KB/second of real throughput doing real work (ie: serving up an FTP 
file).  That's terrible!  Sorry, but 15% of the cable bandwidth doesn't cut
it folks!

Considering that a Sun Sparcstation can do something like 400KB/sec, and a
"real workstation" like a MIPS Magnum does something like 600KB/sec (again,
these are "real world" numbers at the file transfer level) the MCA bus
systems that I've seen (of which the RS/6000 is one) leave a lot to be
desired.

I've seen servers (ie: MIPS 3260) hit something close to 750K/sec, which is
75% of the cable bandwidth and a darn good performance.  Not coincidentally,
user response times are in the "gawd that's fast" arena.

>It depends on who built the boards, of course.  The only
>Ethernet boards available now for the A3000 are Zorro II boards, which of
>course are slow, like ISA boards.  

Yep.  And as long as the makers are more interested in cost than
performance, it will stay that way.

For a fileserver, this is NOT an acceptable tradeoff!

--
Karl Denninger (karl@ddsw1.MCS.COM, <well-connected>!ddsw1!karl)
Public Access Data Line: [+1 708 808-7300], Voice: [+1 708 808-7200]
Macro Computer Solutions, Inc.   "Quality Solutions at a Fair Price"