From owner-freebsd-questions Mon Feb 24 13:31:21 1997 Return-Path: Received: (from root@localhost) by freefall.freebsd.org (8.8.5/8.8.5) id NAA20155 for questions-outgoing; Mon, 24 Feb 1997 13:31:21 -0800 (PST) Received: from siegfried.utmb.edu (FTP.scms.utmb.edu [129.109.59.86]) by freefall.freebsd.org (8.8.5/8.8.5) with SMTP id NAA20116 for ; Mon, 24 Feb 1997 13:30:52 -0800 (PST) Received: from beowulf.utmb.EDU (beowulf.utmb.edu [129.109.59.83]) by siegfried.utmb.edu (8.5/8.5) with ESMTP id PAA17440; Mon, 24 Feb 1997 15:27:30 -0600 Received: by beowulf.utmb.EDU (SMI-8.6/SMI-SVR4) id PAA08444; Mon, 24 Feb 1997 15:28:01 -0600 Date: Mon, 24 Feb 1997 15:28:01 -0600 From: bdodson@beowulf.utmb.edu (M. L. Dodson) Message-Id: <199702242128.PAA08444@beowulf.utmb.EDU> To: freebsd-questions@freebsd.org Subject: Adding memory slows down 486 to less than 386SX! X-Sun-Charset: US-ASCII Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Hello, I just upgraded a 486DX-25 from 8MB to 20MB memory and sequential character disk I/O to a SCSI disk is now slower than to an IDE disk on a 386SX! The machine "feels" slow, also. Not only under FreeBSD, but also under Win95 (for processes that do not cause swapping). I have included bonnie results below. The 486 has a VLB BusLogic 445C bus mastering SCSI controller, the 386SX a normal IDE builtin. Note that the 486 is 4-5X faster for block I/O, which is what I was expecting for everything. The kernels do not have bounce buffers enabled. Both machines were lightly loaded. The 486 is a NFS server, the 386SX is a NFS client (it will eventually be a router for my home network, but ppp was not running during these benchmarks). Nothing was NFS mounted. The BIOS setup for the 486 is standard as far as I can tell: 2-1-1-1 memory timing, 256KB cache was enabled, no cache waitstates. The 386 was box stock, also, as far as I know. I have no documentation on the 386 builtin controller, but BSD probes it as a standard IDE controller. I don't get any memory fault indications. Note that the 386SX is only running a 10MB bonnie file, whereas the 486 was running a 30MB. But I am talking sequential input/output, per character, rates: 82K/sec vs 55K/sec output and 87K/sec vs 64K/sec for input. And why is CPU utilization so high for the block I/O on the 486 relative to the 386SX? Any help would be appreciated. I'm at a loss. Bud Dodson Specific info: Representative bonnie benchmark results: 16MHz 386SX, 6MB, 120MB IDE (slow) disk: File './Bonnie.159', size: 10485760 . . . -------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 10 82 83.2 433 30.1 239 22.5 87 86.3 497 22.5 26.5 14.0 25MHz 486DX, 20MB, 500MB SCSI2 (reasonably quick) disk (BusLogic 445C VLB busmastering controller): File './Bonnie.210', size: 31457280 . . . -------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 30 55 97.7 1818 92.8 1055 73.5 64 98.9 1913 63.6 58.4 16.8 dmesg output for the 486DX: FreeBSD 2.1.5-RELEASE #0: Wed Sep 4 19:55:48 CDT 1996 root@nagling.611.chedworth.hou.tx.us:/usr/src/sys/compile/NAGLING CPU: i486DX (486-class CPU) real memory = 20971520 (20480K bytes) avail memory = 19075072 (18628K bytes) Probing for devices on the ISA bus: sc0 at 0x60-0x6f irq 1 on motherboard sc0: VGA color <16 virtual consoles, flags=0x0> ed0 at 0x240-0x25f irq 10 on isa ed0: address 00:40:05:26:19:c4, type NE2000 (16 bit) sio0 at 0x3f8-0x3ff irq 4 on isa sio0: type 16450 sio1 at 0x2f8-0x2ff irq 3 on isa sio1: type 16450 lpt0 at 0x378-0x37f irq 7 on isa lpt0: Interrupt-driven port lp0: TCP/IP capable interface fdc0 at 0x3f0-0x3f7 irq 6 drq 2 on isa fdc0: NEC 765 fd0: 1.44MB 3.5in ft0: IOMega tape wdc0 at 0x1f0-0x1f7 irq 14 on isa wdc0: unit 0 (wd0): wd0: 325MB (666600 sectors), 1010 cyls, 12 heads, 55 S/T, 512 B/S bt0: Bt44xC/ 0-(32bit) bus bt0: reading board settings, busmastering, int=11 bt0: version 4.21, sync, parity, 32 mbxs, 32 ccbs bt0: targ 0 sync rate=10.00MB/s(100ns), offset=15 bt0: targ 4 async bt0: targ 6 sync rate= 4.54MB/s(220ns), offset=15 bt0: Using Strict Round robin scheme bt0 at 0x330 irq 11 on isa bt0 waiting for scsi devices to settle (bt0:0:0): "IBM DALS-3540 S60E" type 0 fixed SCSI 2 sd0(bt0:0:0): Direct-Access 516MB (1056768 512 byte sectors) (bt0:4:0): "ARCHIVE VIPER 150 21247 B07:" type 1 removable SCSI 1 st0(bt0:4:0): Sequential-Access st0: Archive Viper 150 is a known rogue density code 0x10, drive empty (bt0:6:0): "TOSHIBA CD-ROM XM-3501TA 3054" type 5 removable SCSI 2 cd0(bt0:6:0): CD-ROM cd present.[324186 x 2048 byte records] npx0 on motherboard npx0: INT 16 interface -- M. L. Dodson bdodson@scms.utmb.edu 409-772-2178 FAX: 409-772-1790