Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!ukma!rutgers!cbmvax!steveb
From: steveb@cbmvax.UUCP (Steve Beats)
Newsgroups: comp.sys.amiga
Subject: Re: CRETIN MANOR MAIL
Message-ID: <5009@cbmvax.UUCP>
Date: 17 Oct 88 14:58:25 GMT
References: <7699@gryphon.CTS.COM> <3220@hubcap.UUCP> <10062@cup.portal.com>
Reply-To: steveb@cbmvax.UUCP (Steve Beats)
Organization: Commodore Technology, West Chester, PA
Lines: 61

In article <10062@cup.portal.com> version B 2.10.3 4.3bsd-beta 6/6/85; site cbmvax.UUCP cbmvax!rutgers!rochester!cornell!batcomputer!gould!steinmetz!uunet!portal!cup.portal.com! thad@cup.portal.com (Thad Thad Floryan) writes:
>
>====================
>``
>Re: the claims made by ******* for the ***** interface (re: "500K per
>second"), that sounds more like 500 K *BITS* (not bytes) per second.
>
>The ONLY way I could believe the (undocumented (e.g. test environment NOT
>stated, what disk, etc.)) alleged 500KBytes/second would be if the system has
>multi-tasking shut down, the interface was having its registers read in a busy
>wait loop, *AND* the disk was capable of providing sustained data transfer
>onto the bus at 500Kbytes/second.  This is NOT the way any sane person is
>going to use the Amiga for doing anything "real."
>
>The claim of 500Kbytes/second implies that a 10MByte FILE can be read into the
>Amiga in 20 seconds.  Or let's make it easier: a 2MByte can be read into RAM
>in 4 seconds.
>
>As I said before, if it CAN do that (read a 2MByte file into RAM in 4
>seconds), then this should be touted.
>
>If the Amiga (with *****'s interface) is outperforming that, then why hasn't
>this been publicized other than here in an "in passing" comment by *******?
>
>''
>====================
>Thad Floryan [ thad@cup.portal.com (OR) ..!sun!portal!cup.portal.com!thad ]

Thad,

500K per second across a SCSI interface is not tough to do at all.  Remember,
UNIX machines do all file I/O in 1K chunks which hardly makes good use of a
DMA hard drive.  The command overhead gets appreciable when you have to 
do it 1000 times to fetch a MegaByte of data.  DMA driven Amiga hard drives
show thier best performance when used with FFS, which attempts to coalesce
blocks and transfer them all in one go.  This REALLY allows DMA drives to run
at full tilt.  Maybe some (un)official timings of my own will convince you.

Setup		A2000
		2090 DMA HD controller
		CDC Wren 5 (200Mb)
		2 Meg fast RAM
		FFS partition with 100 cache buffers

Even though the partition was >30% full, I managed to get a sustained read
rate (at the filing system level) of 700K per second.  Peak read rates (when
the system was quiet I guess) were clocked at 850K per second.  Reading 2Meg
directly into memory from the device took 1.8 seconds.  Thats BETTER than a
Meg a second.  Note, that the drive was disconnecting and reconnecting every
time a seek was required so there was a fair bit of task switching and 
interrupt activity going on at the same time too.  

Now Wren 5's are pretty much the Porsches of hard drives these days.  They
are outrageously fast!  Stepping down to a more modest drive, say an 80Meg
Quantum, I could only get 900K per second at the device level.  Maybe you
should blame the drives, not the controllers.

	Steve

PS.  I hope I'm responding to the correct person, the original posting was
     a little confusing but it looked like Thad originated all of it.