Subject: Re: Fastest disk tranfer rates
To: Bruce Martin <brucem@cat.co.za>
From: Mark Brinicombe <mark@causality.com>
List: port-arm32
Date: 11/29/1999 20:36:34
On Mon, 29 Nov 1999, Bruce Martin wrote:

> I am testing hard disk transfer speeds on an arm32 (EBSA285) system under
> NetBSD with the Promise Ultra 33 IDE controller. Can anyone out there tell
> me the fastest transfer rates they have got, how they managed to get them,
> and what hardware they were using.

Ok here are some rates I have on different systems

All systems 233MHz SA110 + DC21285 including CATS & EBSA285

raw reads:

controller 	FCLK	PCI	disk		mod	size	rate

ALI1543		50	33	7200maxtor	UDMA2	128MB	14.9MB/s
ALI1543		50	33	5400maxtor	UDMA2	128MB	12.2MB/s
ALI1543		50	33	4000quantum	DMA2	128MB	7MB/s
PDC20262	50	33	5400quantum	UDMA2	128MB	11.2MB/s
PDC20246	50	25	5400quantum	UDMA2	128MB	8.9MB/s

ISP1040		50	25	5400ibm		SCSI	128MB	8.4MB/s

From the above rates it is clear that the disk is the limiting factor in
these transfers. All UDMA IDE discs are Ultra/33 as I don't have any
Ultra/66 disks to try on the PDC20262 at the moment.
The PDC20246 performance could be improved a bit if it upped the PCI bus
speed to 33MHZ but at the moment on that system the PCI clock is derived
from the FCLK.

For comparison with Curt's figures here is the same benchmark using the
5400maxtor disk from the above table.

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
          100  4241 85.7  9022 63.5  2386 36.6  3830 88.9  7483 67.0  85.8 10.6

Hope these figures are of use.

Cheers,
				Mark