pete@octopus.UUCP (Pete Holzmann) (05/03/88)
In article <613@mccc.UUCP> pjh@mccc.UUCP (Pete Holsberg) writes: > >Has anyone found that the controller has a big effect on HD performance. > I have a Micropolis 1335 (71MB formatted, 28ms) drive with a DTC 5280 >controller in a 386 box, and the Infoworld benchmark says that it is a >dog! CORETEST 2.4 and PC Mag Benchmark suite 4 agree that it is a 28 ms >drive! Steve Satchell (IW) has yet to answer this question about what >the IW test does that might make this combo look so bad. Perhaps one of >you has a thought?? Thanks. The controller has little to do with seek times. It should also be noted that for many applications, average seek time has almost nothing to do with performance, either! For most applications, DATA TRANSFER RATE is much more important. This is affected by: - interleave supported by controller - data buffering in controller - head switch time of drive - track-to-track seek time of drive - drive data format (MFM, RLL, ESDI, SCSI in increasing speed order). If you deal with lots of short accesses to widely varying data areas (e.g. random lookups in a database with thousands of short records), then the seek time will be important. For everything else, including program load time, sequential data search, etc., the data rate is much more important. CONTROLLERS Your absolute fastest possible data rate will be had with a synchronous SCSI controller running at full speed with a SCSI drive. I haven't seen one yet. Next is an ESDI controller/drive running full speed. CAST makes controllers that run full speed (> 1 MB/sec on Coretest!); Adaptec is supposed to (I'm testing that later this week). OMTI controllers do not get full performance out of ESDI: they 'only' go around 700K/sec. I haven't tried the WD ESDI controllers, but if their MFM/RLL performance is any indications, I doubt they would do very well. Next is RLL full speed combo. Both Adaptec and OMTI versions go full speed 1:1. Adaptec is a little faster (less overhead/better buffering) under most circumstances. If you have a very fast bus (10MHz 0 wait state) though, the Adaptec controllers actually slow down! If you have a slow AT, the Adaptec is definitely best: it gets full RLL speed (> 700K/sec on Coretest) on 6-8 MHz AT's. OMTI gets 660-700K/sec on 8+ MHz AT's. MFM running 1:1 interleave is next, getting about 1/3 less throughput than RLL (around 450K/sec). Adaptec and OMTI make good MFM controllers. Certain WD flavors can do 1:1. DTC controllers don't. If a controller can't handle 1:1 interleave, you lose big very quickly. 2:1 interleave cuts your data rate in HALF! (There's no such thing as 1.5:1). DRIVES The only really tricky question is: "Which drives are good RLL drives?" (Well, a second part is 'which controllers can do RLL best on non-RLL drives?' Answer: Adaptec has better window margins (see #1 below for definition) than others. Not a big deal in general, though.) The answers: 1) The Seagate ST-225 and ST-238 are *NOT* good for RLL. They are barely good as MFM drives! (Their electronics and platter material just can't handle the narrow 'window margins' (technical term meaning 'how accurate must the data signal timing be?'). 2) Miniscribe drives, whether certified for RLL or not, make prettty good RLL drives. Their electronics/platters were *designed* for RLL. Some are not officially RLL certified, which simply saves Miniscribe from some warranty claim costs. 3650's work just fine! 3) Maxtor drives seem to work just great, even though they are not certified. Also note that the 1140 is really a 2190 with a different label: It is advertised as having only 918 cylinders, but really has 1224. $1795 at Fry's + an RLL controller gets you 240MB of disk space! 4) Anybody have experience with other brands? Pete -- OOO __| ___ Peter Holzmann, Octopus Enterprises OOOOOOO___/ _______ USPS: 19611 La Mar Court, Cupertino, CA 95014 OOOOO \___/ UUCP: {hpda,pyramid}!octopus!pete ___| \_____ Phone: 408/996-7746
pjh@mccc.UUCP (Pete Holsberg) (05/05/88)
...The controller has little to do with seek times. It should also be noted ...that for many applications, average seek time has almost nothing to do ...with performance, either! Did I say anything about seek times? The IW test measures performance in a yet-unpublished way. ...For most applications, DATA TRANSFER RATE is much more important. This is Agreed. I changed the interleaving from 3 to 2 and data xfr went from 165 KB/s to 250 KB/s. ...CONTROLLERS ... ...Your absolute fastest possible data rate will be had with a synchronous ...SCSI controller running at full speed with a SCSI drive. I haven't seen one True, but that doesn't address my problem. I have an ST-506 drive. ... 4) Anybody have experience with other brands? I've been running a Rodime 203E with an early Adaptec RLL controller with excellent results for nearly two years. However, no one should use any HD with an RLL controller that is NOT certified for RLL formatting. ... OOO __| ___ Peter Holzmann, Octopus Enterprises Your treatise on HDs was undoubtedly very informative, but my purposes would have been served better had you answered my questions. :-) Pete
pete@octopus.UUCP (Pete Holzmann) (05/05/88)
In article <618@mccc.UUCP> pjh@mccc.UUCP (Pete Holsberg) writes: > >[me]... 4) Anybody have experience with other brands? > >I've been running a Rodime 203E with an early Adaptec RLL controller >with excellent results for nearly two years. However, no one should use >any HD with an RLL controller that is NOT certified for RLL formatting. I am certain that you shouldn't be so certain about this absolute prohibition :-) ! I'll repeat it in a different way: If you have a Seagate non-RLL drive, *stay away* from RLL. If you have other drives, your luck will be much better. In particular, the Maxtor and Micropolis drives have good enough designs to always be used with RLL, despite what their marketeers would lead you to believe. IF you are selling the drives under warranty, you probably can't get away with using non-RLL under RLL conditions, since the mfg won't back you up. If you are the user, and the drive fails, it won't be the RLL's fault! If you start getting lots of data errors, *that* could be due to RLL, but all you have to do to fix it is re-do the low level format using non-RLL controller. Manufacturers *frequently* sell high-end equipment with low-end stickers on them. Why? It is cheaper to make a single product than two different ones. You sell it into two different markets, making more $$ from the people that can afford to pay more (and think they are getting more). Maxtor in particular seems to be fond of this. Two examples, one certain, one rumor that is being checked on this week: 1) Maxtor 1140 vs. 2190. The ONLY difference in the specs is 918 cylinders for the 1140 vs. 1224 cylinders for the 2190. Actually, they are identical drives, right down to the testing (and results) done at Maxtor. Confirmed by inside sources, and *extensive* use [close to a hundred drives known to be in use this way, from personal contact] 2) New Maxtor ESDI drive [don't have model handy]. The 170MB version has 7 heads, the 350MB version has 15. Rumor has it that changing a jumper on the drive will enable the other 8 heads [and platters] that have been waiting in there all along! We'll see... > >Your treatise on HDs was undoubtedly very informative, but my purposes >would have been served better had you answered my questions. :-) Sorry! I try to answer the underlying general questions whenever I think that people are probably interested. I didn't realize that I missed your particular question in my answer. Maybe you can email me the specific Q I missed and I'll try again [your original article is already gone from here]. Pete -- OOO __| ___ Peter Holzmann, Octopus Enterprises OOOOOOO___/ _______ USPS: 19611 La Mar Court, Cupertino, CA 95014 OOOOO \___/ UUCP: {hpda,pyramid}!octopus!pete ___| \_____ Phone: 408/996-7746
pjh@mccc.UUCP (Pete Holsberg) (05/06/88)
In article <214@octopus.UUCP> pete@octopus.UUCP (Pete Holzmann) writes: ...In article <618@mccc.UUCP> pjh@mccc.UUCP (Pete Holsberg) writes: ...> ...>[me]... 4) Anybody have experience with other brands? ...> ...>I've been running a Rodime 203E with an early Adaptec RLL controller ...>with excellent results for nearly two years. However, no one should use ...>any HD with an RLL controller that is NOT certified for RLL formatting. ... ...I am certain that you shouldn't be so certain about this absolute prohibition ...:-) ! ... I can think of no circumstance in which I would ever recommend that anyone use a drive not certified for RLL use with an RLL controller. Yep -- I'm positive. :-) ... ...If you are the user, and the drive fails, it won't be the RLL's fault! If ...you start getting lots of data errors, *that* could be due to RLL, but all ...you have to do to fix it is re-do the low level format using non-RLL ...controller. No one said it would be! If you put low octane gas in an automobile which requires high octane, and the engine knocks, do you blame the engine? ... ...Manufacturers *frequently* sell high-end equipment with low-end stickers ...on them. Why? It is cheaper to make a single product than two different ...ones. You sell it into two different markets, making more $$ from the ...people that can afford to pay more (and think they are getting more). What about profit margins? These people have to compete. ... ...Sorry! I try to answer the underlying general questions whenever I think ...that people are probably interested. I didn't realize that I missed ...your particular question in my answer. Maybe you can email me the specific ...Q I missed and I'll try again [your original article is already gone from ...here]. ... 'Sokay! I got the answer elsewhere.
Paul_L_Schauble@cup.portal.com (05/08/88)
Can anyone report experience with Newbury Data disks. These are clones of
the Maxtor 1140 and 2190 drives, supposedly manufactured under license
from Maxtor.
Pauljtara@m2-net.UUCP (Jon Tara) (05/08/88)
In article <212@octopus.UUCP>, pete@octopus.UUCP (Pete Holzmann) writes: > 4) Anybody have experience with other brands? I've had good luck with a Priam/Vertex V170 on an OMTI 8627. Just for yucks, I tried to format my old Rodime 203E which I had in D: when I got the OMTI controller. Suprised the hell out of me when not only did it format, but all the errors I had been getting with my IBM controller went away - 6 months later, I haven't seen an error on either drive! (I used to get them all the time on the Rodime, and very occasionally on the Priam.) Anyone for 650K/sec at 60msec? -- jtara%m-net@umix.cc.umich.edu ihnp4!dwon!m-net!jtara "If this is all a dream, I can't wait to see what it's like when I wake up." - Explorers
pjh@mccc.UUCP (Pete Holsberg) (05/09/88)
In article <1721@m2-net.UUCP> jtara@m2-net.UUCP (Jon Tara) writes:
...Just for yucks, I tried to format my old Rodime 203E which I
...had in D: when I got the OMTI controller. Suprised the hell
...out of me when not only did it format, but all the errors
...I had been getting with my IBM controller went away - 6
...months later, I haven't seen an error on either drive! (I used
...to get them all the time on the Rodime, and very occasionally
...on the Priam.) Anyone for 650K/sec at 60msec?
Me! Which OMTI controller are you using? I have one of the original
Adaptec RLL controllers on my 203E, and it's pretty darned nice, but yours
sounds even better!wtm@neoucom.UUCP (Bill Mayhew) (05/11/88)
At the moment, I have a Priam 330 MB drive on the IBM model 80 I am using. The Priam is a full height drive, with about 28 mS access time according to CORETEST. It draws about 2.2 amps while it is running, so it isn't for faint-of-heart power supplies. It plugged right into the indigenous ESDI controller in the mod 80, and shows a transfer rate around 768 Kbytes/sec (without IBMCACHE running). Priam's documentation is very good, and the included setup disk did everything automatically for setting up the drive. The only thing you have to do is to make a backup copy of the installation disk and then SYS DOS onto it, boot up and it does its thing. There must be one heck of a mondo positioner in the Priam drive, as the whole mod 80 shakes visibly when the disk drive is doing some heavy duty seeking. By the way, the mod 80 is the "old" 16 MHz version with IBM's own ESDI card, which of course is made by Adaptec. I had been using a CORE International drive AKA CDC Wren III in the model 80 before. The core drive was very quiet in operation, and showed about the same performance figures as the Priam drive. Kind of makes me wonder if the controller isn't the limiting factor here. The CORE drive really is a nicely put together device. The construction quality looks much nicer than any other drives I've looked at, except for some Toshiba 8" SMD/E drives (but those don't count!). I'd say that if I wanted a pretty large (~120 MB) fast, rugged disk drive, the CORE is the way to go... but of course, you get wahat you pay for. CORE isn't the cheapest game in town. I don't have personal experience with the OMTI ESDI controllers, but I have used the 5527 ST-506/412 RLL controller, and have been impressed with the performance. OMTI's controllers from visual inspection seem to be more durably constructed than Adaptec's. --Bill
keithe@tekgvs.TEK.COM (Keith Ericson) (05/13/88)
In article <212@octopus.UUCP> pete@octopus.UUCP (Pete Holzmann) writes:
< 3) Maxtor drives seem to work just great, even though they are not
< certified [for RLL]. Also note that the 1140 is really a 2190 with a
< different label: It is advertised as having only 918 cylinders,
< but really has 1224. $1795 at Fry's + an RLL controller gets
< you 240MB of disk space!
<
< 4) Anybody have experience with other brands?
<
The Maxtor 1105 has a similar story - it's warranted for 918 tracks
but will format out to 1024.
keith