From: Digestifier To: Linux-Admin@senator-bedfellow.mit.edu Reply-To: Linux-Admin@senator-bedfellow.mit.edu Date: Wed, 8 Sep 93 00:02:51 EDT Subject: Linux-Admin Digest #43 Linux-Admin Digest #43, Volume #1 Wed, 8 Sep 93 00:02:51 EDT Contents: POLL RESULTS, PART 4/7 (Matthew Dillon) ---------------------------------------------------------------------------- From: dillon@moonshot.west.oic.com (Matthew Dillon) Crossposted-To: comp.os.linux.misc,comp.os.linux.help,comp.os.linux.development Subject: POLL RESULTS, PART 4/7 Date: 7 Sep 1993 18:46:30 -0700 =================== HD/CONTROLLER SPEED ================== CONTROLLER DRIVE TYPE ? Western Di LO DISK Nothing available for comparison HI DISK Nothing available for comparison ? 116MB Quan HI DISK My Quantum is VERY slow (well it appears slow to me), however it is quiet (which I prefer) and the Conner is much faster but somewhat noisy. I sleep in the same room as the computer which is one reason I like a quiet drive. I can get use to a fan (constant noise) but the intermitent HD can be annoying. ? Seagate, C HI DISK a bit slow ? ? LO DISK Fast HI DISK Medium ? ST1144A SE LO DISK Well, it's 18ms, and plenty fast for what I am doing, which sometimes can include stuff that uses the disk heavily. I haven't done yet any testing like using iozone, but I've tested the disk inside DOS. ? ? HI DISK Slow! ? Seagate * LO DISK No problems. ? ? HI DISK The 'alps' is slow. mostly because of the aforementioned problem with interrupts and the 300ms delay needed. The quantum is a fair bit faster, but not noticeably. (raw data rate is around 700K/s for both, the alps drops to around 50K/s on random read/write). ? Seagate (4 HI DISK Very slow (compared with the Sun's, at work... ;-) ? Quantum I HI DISK Relatively fast - would prefer SCSI though ? ? LO DISK IDE interface is Ok for DOS HI DISK SCSI is reserved to Linux, I'm working whith UNIX (Ultrix AIX, and linux) and need lots of compilations . SCSI is a great improvement from this point of view ? ? HI DISK Fast as heck. ? Seagate LO DISK With all the cash, It does not slow me down. I use it as a Point of sales terminal I can look up 300 products per secound with a PERL (ndbm) data base. You can not pick up prouducts that fast :-) ? ? LO DISK Fast - almost no noticeable delay even loading large binaries, like emacs. Seems surprising for IDE. HI DISK Adequate. This is where I seem to pay the penalty for the slow data transfer rate (~1meg/s) of the drive. ? Western Di LO DISK No perceptions. HI DISK No perceptions. ? 2 Western LO DISK Sufficiently fast for my needs HI DISK Could be a tad faster, but acceptable CONTROLLER DRIVE TYPE ? Western Di HI DISK I don't know whether the disk controller or the disk is the bottleneck. So I can only guess very roughly that the speed is "medium". (I didn't measure it or compare it to other hardware) ? Toshiba mk HI DISK Yes it's a little lame. It's hard to use up more than 40% of the CPU while compiling with GCC. I've never run iozone/etc so I can't separate this out from disk latency / linux file system, ... ? ? HI DISK Slow for swapping (no benchmarks taken - It was all I could afford) ? quantum LO DISK not as fast as i'd like HI DISK not as fast as i'd like ? Fujitsu 18 LO DISK Normal HI DISK Very slow (but I think the problem lies in disk) ? Seagate LO DISK Fast enough. HI DISK A bit slow, especially with a lot of reading/contention. ? Maxtor LO DISK Zippy. HI DISK Reasonably zippy. I get about 900kb/sec dd'ing a partition to /dev/null. ? Maxtor & W HI DISK If I turn off turbo mode, I notice that the drive doesn't stay as active. I assume this means the controller is keeping a good pace, with the drive's ability. ? Quantum LO DISK Disk subsystem seems fine to me. ? Maxtor 200 LO DISK Fairly good. HI DISK Well, my second HD, a Connor 250 MB, is slightly faster. ? 2 * Seagat LO DISK Fine HI DISK Too slow - any mechanical device would be. Are you sure you thought this question out? ? Maxtor HI DISK The controller isn't the problem, but it's better to have two drives with a well designed splitting of the directorys. ? 2 WD Cavia LO DISK excellent for IDE and SCSI HI DISK both OK ? WESTERN DI LO DISK System responds very quickly. The WD drive has 64kb of buffer RAM which it using for reads and writes (write-back cache). The cache uses a adaptive algorithm to determine the optimal caching scheme based on recent disk usage. HI DISK Still get good performace numbers. A misc disk test which was posted some months back returned number of 1.8MB/s on reading and 2.0MB/s on writing. The test was done on a 9MB file. The write-back cache skewed the write results (no doubt). ? Quantum HI DISK Seems on the slow side but that's probably more to do with the fact that I'm swapping to file not to a swap partition. ? Maxtor HI DISK It's a reasonable speed. 12 ms access time. Using floppies, of course, slows it down a bit. ? Western Di LO DISK reasonable HI DISK reasonable CONTROLLER DRIVE TYPE ? Western Di LO DISK ACCEPTABLE HI DISK SLOWER THAN DESIRED WHEN COMPILING PROGRAMS AND RUNNING X ? DEC DSP 30 HI DISK disk controller works fine, but Linux Adaptec 1542 SCSI kernel driver has very BAD performance! ? Western Di LO DISK The VESA Local-Bus IDE hard disk controller is very fast. ? Conner Cav HI DISK Very good ? Not sure, LO DISK seems pretty speedy, but I'm not a PC expert (Don't have it handy...) Quantum LP LO DISK I haven't noticed much of a speed increase over a non VLB controller... Probably 'cos the kernel doesn't support VLB fully AFAIK HI DISK Fairly constant, regardless of disk usage... (slightly slower, but still acceptable) ? Conner WD LO DISK high speed HI DISK high speed ? one hitach LO DISK seems fine ? Conner / Q LO DISK seem to be allright HI DISK sometimes a bit slow ? Western Di HI DISK >Works OK for heavy swap, very fast for normal I/O ? Western Di HI DISK >Works OK for heavy swap, very fast for normal I/O ? Seagate LO DISK enough (never measured) HI DISK enough (never measured) ? Seagate LO DISK In light usage, I don't think the card is a bottle neck. HI DISK Heavy usage? Hmmm, I never really thought about it. I doubt the card is too much of a bottle neck. ? Quantum * LO DISK slow, barely satisfactory HI DISK sloooooooow ? ? LO DISK Fast. HI DISK Average to fast. ? Western Di LO DISK Normal. CONTROLLER DRIVE TYPE ? Maxtor LO DISK I have not done enough serious disk accessing to answer. Just running X does quite well. ?? ?? LO DISK reasonable HI DISK reasonable ?? Quantum LS HI DISK Not the fastest I've seen. At work I have a faster disk setup, but apart from that, I have no experiences. ?? 1 Maxtor a LO DISK I don't really have anyhting to compare it with although it seems a little slow to me. HI DISK As above. ?? (Generic) Western Di HI DISK seems very speedy, ~10 ms seek time ??? Maxtor LO DISK Supposedly VLB based IDE controllers can deliver twice the throughput of an ISA based IDE controller - but I have no disk access benchmarks to offer. A VLB/ISA disk benchmark would be interesting. HI DISK Swapping speed seems 'reasonable' in Linux, usually only a few seconds hold up if and while swaps occur. Seems to much more swapping with Xfree 1.3/SLS 1.03 than when I was using SLS 1.02 ??? Seagate & LO DISK Fine HI DISK Fine ??? Western Di LO DISK Pretty good HI DISK Pretty good ??? Maxstor 72 LO DISK Beats the hell out of the MFM Seagate 4096 I had in there... HI DISK See above. ??? Kalok 3100 LO DISK Fast enough, Linux uses 'spare memory' as cache so if you have 8M ... HI DISK Still good enough, used to lots of things mounted over NFS at work compares favourable to that. I am not running Xfree very much...yet ??? (perhaps Seagate) Seagate ST LO DISK no problems HI DISK swapping under X slows the machine significantly, on the other hand, I wait the hole day under X (386/25) :-( Acculogic sIDE-3 Maxtor HI DISK Heavy swapping will load the thing down. Normal use is OK. Adaptec Fujitsu, m LO DISK good HI DISK not good, because only about 300Kb/sec transfer rate )-: but no idea why ! Adaptec IBM (2x) LO DISK Same applies. Because I have two harddisks, and my root system and HI DISK See 4A. ADAPTEC Micropolis HI DISK the disk system seems fast, compared to an ST115 or the same setup with DOS Adaptec ? LO DISK excellent HI DISK excellent Adaptec 50M RLL Se HI DISK Ok. comparing with the CPU-Speed CONTROLLER DRIVE TYPE Adaptec Fujitsu HI DISK My board is very slow. So the disk speed is not so important. But on heavy load you examine waits because of the small bandwidth of the 8MHz ISA bus. Also I have the B version not the C version Adaptec released early this year. Adaptec Maxtor HI DISK Seems to work *wonderfully*. :-) Adaptec Seagate & LO DISK With linux's internal caching, very good. HI DISK Ouch... Remember, one drive is 60ms, the other 30ms seek times... Controller is fast enough, drives are not.... Adaptec 1xMaxtor, LO DISK Very good HI DISK Fairly good Maxto is quite slow, Fuji is very fast Adaptec Western Digital Fujitsu/Ma HI DISK [486/50] The difference in speed between the Fujitsu drive and the Maxtor drive is noticable when running compilations. Adaptec 1542b Toshiba LO DISK reasonably fast HI DISK ditto Adaptec 1542B Maxtor LO DISK Seems fast since everything gets cached into memory anyway. HI DISK Of course, I'd like a faster system. Seems at least as fast as any other SCSI system like it that I've used. Adaptec 1542B Fujitsu MK LO DISK Reasonably quick for an ISA-bus machine. HI DISK _Much_ quicker than the IDE I used before Linux. Adaptec 1542B Maxtar LO DISK qyucj Adaptec 1542B Future Dom TDC8 2 Seagates LO DISK The adaptec is v. fast. The FD, ok. HI DISK The adaptec is still v. fast. The FD get confused and slooooow. ADAPTEC 1542C IBM(SCSI) LO DISK SCSI is always faster than IDE HI DISK SCSI is always faster then IDE Adaptec 1542C SCSI, generic ID Quantum ID LO DISK It's OK. It is very hard to separate hard drive and system performance in terms of perceptual feel. The system compares favorably with a SPARCstation (IPX) in terms of feeling of speed. Adaptec 1742 Fujitsu an HI DISK In advanced mode the speed is 5MB/sec, but this mode isn't useable (see above). In 1542 emulation it is 1,5MB/sec. I'm NOT dissapointed Adaptec 1742 Fujitsu LO DISK Faster than the Suns at work. HI DISK Ditto. Adaptec 1742A ? LO DISK >1MB/s (limited by drive) very high HI DISK dito Adaptec ? SCSI seaga HI DISK SCSI on the 1GB drive is excellent. I process very large text files using PERL (100 MB or greater). The 486/66 and 1GB drive give me better performance than our Solbourne main frame which is heavily load with other campus users Adaptec AHA1742 HP C2247 ( HI DISK Its reasonable fast, compared with the workstation (HP9000s425) I use at office. In fact, I enjoy working on my Linux box (compiling, program development) much more than on this 425 under HP-UX! (The 425 has the same disk-drive as my 486) CONTROLLER DRIVE TYPE adaptec, ZEOS 1 Quantum, HI DISK The important thing is performance/$. Both controllers are good when rated that way. Then upgradability is important. The SCSI is Fast SCSI 2. It'll do up to 10 meg/sec on a 16 bit path, more on a VLB Fast SCSI2 controller card. But they are expensive. Adaptech (1542C) ? LO DISK Seems OK, though copying between efs2 and dos partition seems slow. HI DISK Seems OK. When swapping occurs, system seems (as to be expected) slow. Adaptech 1742 Quantum (q LO DISK Very fast. HI DISK Very fast. For example, under DOS I can record two channels of 16-bit audio at 128 KHz sampling rate continuously (this is done using a DSP board my company produces - no compression is used, the board just does acquisition). I am in the process of writing drivers for the board for Linux so that I can try it. AHA 1742 Seagate 1G LO DISK Quite fast HI DISK With CNEWS, receiving <> 2000 articles the disk io is very slow AHA1542B Fujitsu LO DISK No noticeable delay. HI DISK Depends how much thrashing I do :-). I can't really compare disk performance, as it is part of the overall speed etc. Also, comparative performance between different systems is very difficult. Appian ADI2 Conner 500 LO DISK sufficient HI DISK I know it could be better if there was a VLB driver since under DOS transfer rate goes up from 1Mb/sec to 5Mb/sec Both SCSIs are Adaptec 1542 Maxtor, Qu LO DISK Compaq MB ide: Being used as swap after 16M makes it hard to say. :) Compaq w/ 1540B: Good. Noname IDE: tolerable. Noname w/ 1542C: see 1540B (bus speed has _not_ been raised) HI DISK Compaq MB ide: no heavy use. Compaq w/ 1540B: Excellent sequentially. (* Noname IDE: "Could be better" Noname w/ 1542C: see 1540B (bus speed has _not_ been raised) BusLogic Micropolis LO DISK not being used. HI DISK very good throughput makes it very fast feeling I/O response. Bustek Fujitsu HI DISK Fast, because it's the SCSI-II drive/controller combination. Cheap Taiwan one (No-name) Maxtor LO DISK Fairly fast. Low latency times. HI DISK It slows the system a bit but I can't afford SCSI Chicony Maxtor HI DISK Under Linux it is quite good. Under 386BSD it is not impressive. ChipsTech ??? Western Di LO DISK medium (probably due to slow ide driver), fast under DOS ( ca. 3400 kB/s) HI DISK better (short seek times) Compaq Compaq LO DISK Reasonable, but not great. Compaq(??) Conner LO DISK Quite Fast HI DISK Still Fast (enough for me anyway ) DFI Fujitsu LO DISK ok HI DISK ok don't know Maxtor LO DISK Acceptable speed. HI DISK Too slow. Don't know Conner 85M LO DISK Linux's buffer cache masks controller/disk speed if only small amounts of data are being transfered. My perception is that the speed is reasonable under light loads. HI DISK When I use the disk heavily, its usually for paging. In this situation, faster controller/drives wouldn't help - what's needed is more RAM. don't really know, some no-nam Quantum 10 LO DISK quite good, don't think it's a bottleneck (386 - no further comment necessary, I think :) HI DISK Well, Again I couldn't compare to any faster system, really, and I'm still used to using a 68010-based SysV-clone with 15 Terminals, which makes even the longest swapping seem normal... CONTROLLER DRIVE TYPE don't remember (not at home no Maxtor HI DISK Not bad when building XFree86 beta releases :-) Better for X builds (and anything else) than a Sun Sparc IPC with a Wren VII (1 GByte SCSI). More than twice as fast (thanks to a large disc buffer pool :-) DTC Maxtor LO DISK Reasonable HI DISK never fast enough! DTC ? LO DISK Reasonably fast; I have not measured it HI DISK Reasonably fast; I have not measured it DTC 2270VL Western Di HI DISK Speed is drive-limited in single-user modes (eg DOS). Obviously, I might expect higher throughput with SCSI in a multiuser/multitasking environment (Linux with lots going on), but for the price? DTC 3292 Conner 540 LO DISK responsive, but I don't have numbers at hand... iozone gave surprisingly low numbers. Recent discussion about SCSI drivers indicates that there may be a bottleneck at that level. HI DISK Heavily? it's a single user system, so ``heavy'' is a kernel compile and Emacs at the same time. not appreciably different from ``light'' use DTK Western Di LO DISK latency is pretty good compared to previous HD at least HI DISK continuous transfer from drive 0 is about 1200K/sec. drive 1 is noticeably slower, as usual on second IDE drives. dunno Maxtor LO DISK Fast enough for me. Hard to describe accurately. HI DISK Not fast enough :-) The performance collapses when disk usage becomes heavy. dunno 1 x Conner LO DISK Fast. HI DISK Slow. El cheapo :) ? LO DISK Most of the times the load is light => No problems .... HI DISK BUT as the controllers are the normal dumb ones, they are to slow for heavy use. Its god enough for the old drives but its a shame with my new fast 452MB 12ms Seagate :) Dos thansfer rates approx 990kB/s Everex Seagate, M LO DISK Obviously an MF/M controller is much slower than anything anyone would purchase today. Forgot ... I forg LO DISK Who's gonna know, really? It seems fast, but the cache is doing that. HI DISK I can't get a lot done when I'm copiliing something. Things then take Fujitsu Fujitsu LO DISK Speed of actual controller remains constant (in MHz) regardless of use (if this is what you're asking about). My controller is moderately fast (for machines from that era) Fujitsu Fujitsu LO DISK Speed of actual controller remains constant (in MHz) regardless of use (if this is what you're asking about). My controller is moderately fast (for machines from that era) Future Domain This was a tes Quantum (1 LO DISK VERY VERY VERY SLOOOOOOWWW HI DISK VEEEERRRRRYYYY VERRRRYYY SLOOOOWWWW the damn thing is 16bit, what the hell do you expect compared to loalbus? :-) Future Domain FD1680 Seagate 42 HI DISK Could be better - I suspect the motherboard of limiting the transfer though. My IDE disks have moved on to a 386 system and give better transfer rate than they ever did on my 486. Only tested under DOS, I get about 950K/s generic WD LO DISK fast seek and data response make access almost instant HI DISK very good, real transfer rates in 1 - 2MB/s range make this appear only slightly slower than local disk access on my SPARC 2 or HP 735. With only 1 disk disk IDE performs nearly as well as SCSI Generic multi-IO card. Seagate LO DISK on the 386SX/16, it does not seem to be IO bound. HI DISK only slightly noticable when I have X running and it has to go to the swap partition. generic taiwanese ? HI DISK Seems ok, but pretty slow with serious swapping. generic, Adaptec see below LO DISK The Quantum and Seagate drive are fairly fast. Due to the overheating the Conner is stuck at a 300kps transfer rate, so it is limited to the humiliating task of holding a FAT partition to allow file transfer between OS/2 and Linux. HI DISK Under heavy disk use having two controllers makes a big difference. Heavy use on the Quantum for instance doesn't slow down the Seagate much. If I try accessing the Quantum while heavily using the Quantum there is a noticable slowdown (same for the Seagate). CONTROLLER DRIVE TYPE God only knows Maxstor LO DISK I have nothing to compare aginst Goldstar West. Digi LO DISK average HI DISK average Hedaka Seagate HI DISK Speed is acceptable. I'm not interested in specs: If I do a 'make' and the job is done in an acceptable amount of time, I'm satisfied. This is my personal benchmark, the only one that counts for me. integral (WD chipset) WD "Caviar LO DISK Acceptable. HI DISK Acceptable. jap-special Quantum LO DISK I have not done any benchmarking with it. It is Fast Enough (tm) Kouwell Western Di LO DISK Very acceptable HI DISK Feels a bit slow, f.ex. when unpacking UseNet news batches LongShine (LCS) Western Di LO DISK Quite fast, although I haven't used it enough to give it an accurate rating. I only got the caching controller recently. HI DISK Very fast. I only recompiled my kernel once. I didn't time it, but I was very surprised at the speed. (This is with 16M RAM & 4M cache) Like I said above I haven`t been able to use it enough to accurately rat it. I however, rarely wait for the disk. (with or without cache) Media Vision Pro Audio Spectru SEAGATE + LO DISK quite good, no HW problems yet MIO-O5? Western Di HI DISK Noticably slower than a sparc2 for big compilations I think this is a disk interface problem (non-cached IDE) but I have not really looked into it. Never looked at it Quantum, I LO DISK Somewhat slow. HI DISK Thrash, thrash, thrash...slow. Very slow. No clue Western Di LO DISK Gee, how many electrons flows across the bus, that'll also help seperate Joe from power...give me a break. WD Drive - 15ms Seagate - 17ms no name Quantum an LO DISK controller is fast, efs is slow HI DISK controller is fast but efs needs to be remade ( waiting for efs2) no name 1x Quantum LO DISK I don't have access to other controllers to compare performance. no name Seagate ST HI DISK normal for this type of el-cheapo controller. Fortunately Linux has No-name Seagate LO DISK Good to excellent. HI DISK Difficult to say, since peak disk activity occurs due to swapping in X. Can't really separate the two effects at this time. no-name connor and HI DISK just fine No-name Segate LO DISK Seems very speedy. HI DISK zcat 3megfile|tar xvf - speeds along Swap memory, obviously, is slow. CONTROLLER DRIVE TYPE no-name from Taiwan Fujitsu LO DISK seems quite reasonable compared to DOS or other Unices HI DISK it's not too bad when there is no page swapping going on, but once this Noname Seagate 24 LO DISK not sensational but enough NoName Seagate LO DISK Decent, but not real fast HI DISK Fine. Swapping is rare with 16 meg Noname ? LO DISK Fast to be IDE HI DISK FAst to be IDE NoName Western Di LO DISK Acceptable HI DISK Acceptable, but when Linux (0.99.9 - 0.99.12) syncs, it's a pain. noname Western Di LO DISK fast HI DISK reasonably fast noname Western Di HI DISK Notice: Here my box in the Work is a WD-Caviar on a 486 DX/2 50 MHz At home I have a WD1170 Caviar on a 386DX 40 MHZ. But my harddisk performance at home is twice the perfonce an the 486. Why? Because the WD-Caviar is so fast, that the 25MHZ systemspeed of the DX/2 is too slow Noname WD, Micros LO DISK under DOS coretest gives 1.8Meg /s HI DISK drive is still responsive, but only if the second drive (slower) is not accessed NoName WD + Conne HI DISK I only have my company multiuser unix to compare with: Speed is Okay, actually quite fast compared to the > 100.000$ computer i work with professionally. NoName (could not determine) Western Di HI DISK Could be faster. It is with an appropriate driver under MS/DOG. I hear there are efforts going in doing 32-Bit-transfer... NONAME + ADAPTEC Maxtor IDE LO DISK Great. HI DISK Not bad. Swapping is not too fast, but I wouldn't expect better. Noname + Adaptec Conner (ID LO DISK The IDE-Controller takes a small DOS-Partition for DOS to Linux exchange, so there is no heavy access. Speed is OK. HI DISK The SCSI Part (Adaptec) is used as root partition for Linux. Heavy load espacially under X11. The controller works fine. Noname and Adaptec Drive 0 is LO DISK Very good on the adaptec, fair on the IDE. HI DISK good on the adaptec, poor on the IDE. noname DC-660 (Tekram??) Fujitsu 26 LO DISK meets expectations HI DISK meets expectations Noname, I guess. Conner+NEC HI DISK Access time is at about 17ms for both drives. It's OK for most jobs, though I would not say it is fast. Loading of big X apps takes a while, but this is not only a problem of the drive, but also of my CPU and ISA board. I can't figure out what you expect as an answer to this question. none Conner HI DISK It's just ISA IDE. They're all the same. OEM Conner LO DISK faster than some Sun3s I've used. on board seems to b LO DISK hard to tell ... CONTROLLER DRIVE TYPE orchid? Western Di HI DISK Speed is acceptable. Promise Seagate LO DISK Pretty fast, latency seems good HI DISK Throughput is around 1Mb/sec, so not as good as SCSI, but I was after maximum compatibility Promise Adaptec ? LO DISK Both disks are very fast, load makes little difference (but then there is only one device on the SCSI bus at the moment). The IDE controller is *QUICK* HI DISK see above Seagate (ST22R) Maxtor LO DISK Mediocre. Remember, the drives are 40ms drives. HI DISK Poor. Single threaded access and slow access times cause thrashing. Seagate ST01 Maxtor (6 LO DISK Speed is good. HI DISK Speed sucks - only an 8bit interface to motherboard some cheap Korean brand Western Di LO DISK adequate HI DISK let's just say that SCSI would be an improvement... Some taiwanese thingy :-) Quantum HI DISK Not bad at all. I get around 1.0 - 1.4 Megs per second. Amazing. something Taiwan Adaptec Maxtor A72 LO DISK ??? Throughput IDE/Maxtor with Checkit v3.0 under DOS: IDE/Maxtor: 500 KB/sec (Coretest: 650) SCSI/Quantum: 620 KB/sec (Coretest: 820) HI DISK Speed is ok, but the Quantum PD 425S steps the head very loud. Super I/O (generic Taiwan) WD LO DISK No idea. I get good throughput, but I don't know how much is the cont. vs. the drive (it has 64k cashe). Taiwan clone Seagate ST LO DISK Average. The card isn't slow but it doesn't have buffer or cache memory to improve the performance past the drive's abilities. HI DISK Average. The card isn't slow but it doesn't ahve buffer or cache memory to improve performance past the drive's abilities. Taiwan-o-clone (don't know/car Western Di LO DISK Why, not very fast, of course. HI DISK Too slow without disk cache. Trantor Seagate 11 HI DISK The 1006V-MM1 has a track buffer, giving good MFM performance. The Seagate and Trantor SCSI boards have no intelligence, require the CPU to do some amount of manual polling, and give poor interactive performance, allthough throughput isn't that bad. The IDE controller is a couple of PAL's and floppy controller IC, and shouldn't have any affect on IDE performance. UltraStor ? LO DISK fast HI DISK fast UltraStor 12F/24 Fujitsu LO DISK Extremely fast; has a 32k track cache and it's highly integrated. I'm told it supports 24mHz drives and as such very high xfer rates. HI DISK Feels like the cache gets overloaded when accessing huge amounts of (separate, distinct) data spread throughout the disk. Still extremely fast, but the HD seems to be the bottleneck. UltraStor 34F ? LO DISK Quite speedy HI DISK Quite speedy, given the load conditions UltraStore 14F Rodime & C LO DISK No trouble. Never even notice it. HI DISK It is slow when doing memory is used up and disk writes are forced (but before I installed .99pl8 it was *much* slower). It's workable though (I don't get too annoyed at <10 second pauses :-). umc chipset & adaptec Maxtor & q LO DISK good enough on both controllers HI DISK it's a bit slow on the scsi (but the disk is only scsi-1) UNKNOWN Unknown LO DISK Seems very fast (has some on board cache) HI DISK Seems fairly fast unknown Seagate LO DISK very fast HI DISK medium-fast CONTROLLER DRIVE TYPE unknown Conner LO DISK Acceptable HI DISK OK Unknown Quantum LP LO DISK Quite fast, at least much faster than the Micropolis 1355 on my Sun-3/80. find / -print really is fast. Unknown - very generic Western Di HI DISK I do a lot of compiling and the HD system is very peppy. I use Sun's at work and drive speed is a little slower than those - it appears that way - I have no hard numbers on it. WD one's a se LO DISK It's not Annoyingly slow HI DISK It's a clunker! WD Magnetic P LO DISK Old, 28ms drives are _always_ slow (especially on slow computers) HI DISK See above Western Digital Micropolis LO DISK Fairly fast. Micropolis has a 18ms seek time. But 1 to 1 interleave HI DISK Still not bad. Very useable even under heavy load. Western Digital Fujitsu LO DISK Very fast. HI DISK Slight degradation, drive caching and read-ahead tends to keep perceived speed constant in most situations. Biggest indication of heavy use is increased head movement noise. Large head movements (eg copying from partition to partition) cause a more significant slowdown. western digital maxtor, se HI DISK better get a shitload of core, cause it is SLOW even with 16-20mb on my high end box, syncs are deadly RLL and IDE drives both take the load much better, with IDE again being the fastest Zeos, built into MB Western Di LO DISK Fine. HI DISK Too slow... Zeos, built into MB Western Di LO DISK Fine. HI DISK Too slow... END OF PART 4 ------------------------------ ** FOR YOUR REFERENCE ** The service address, to which questions about the list itself and requests to be added to or deleted from it should be directed, is: Internet: Linux-Admin-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.admin) via: Internet: Linux-Admin@NEWS-DIGESTS.MIT.EDU Linux may be obtained via one of these FTP sites: nic.funet.fi pub/OS/Linux tsx-11.mit.edu pub/linux sunsite.unc.edu pub/Linux End of Linux-Admin Digest ******************************