uzun@sdsu.UUCP (william uzun) (12/10/87)
Well I was going to wait for the SDPI from ASDG but since this will not be around until mid 88 I have decided to plunge into a hard disk now. I went to the local computer outlet and tried the CA 2090 with a Miniscribe and was not all that pleased with the performance. I began looking through the computer shopper for faster hard disks but knew that the real problem was the File system and OS. Well I saw an Ad in another magazine for a hard disk from Supra Corp and gave them a call and was pleasently surprised. It seems that in early january they will be shipping a 30M drive with controller which Abels will sell for $675.00. It has a 65 msec access time (pretty slow) but the controller is what got to me! It formats the disk with a 1:1 sector interleave (versus about 3:1 with current software and 2090) which means the files are stored as phyically contiguous sectors for increased speed, It buffers full tracks (a great trick since it is often just as fast to read a track as it is a single sector) and stores data in RLL format (for about 3Xs the data per sector almost tripling the effective Xfer rate). The tech I talked too said they intend to support auto boot as well! Now I am in no way connected with Supra Corp. and have never even heard of them before, does anyone know of this Company? I have placed an advance order with Abels and will keep you all posted if I ever get a chance to use this hardware in the next few weeks. Anyone hear of anything better which will be available in the near future? -Roger
tdwest@lion.waterloo.edu (Terry D. West) (12/11/87)
In article <2855@sdsu.UUCP> uzun@sdsu.UUCP (william uzun) writes: > >.... I went to the local computer outlet and tried the CA 2090 with >a Miniscribe and was not all that pleased with the performance. > >It formats the disk [...] about 3:1 [interleave] with current >software and 2090... > >-Roger Note that on page 24 of the 2090 user's guide it indicates that you are allowed to soft select ANY interleave via an entry in the MountList. In fact, they suggest a value of 0 (meaning 1:1). -- Terry D. West University of Waterloo, Ontario, Canada. {backbone}!watmath!lion!tdwest
erd@tut.cis.ohio-state.edu (Ethan R. Dicks) (12/12/87)
In article <2855@sdsu.UUCP> uzun@sdsu.UUCP (william uzun) writes: > > [stuff deleted about CA2090 and drive ] >a 30M drive with controller which Abels will sell for $675.00. It has >a 65 msec access time (pretty slow) but the controller is what got to me! >formats the disk with a 1:1 sector interleave (versus about 3:1 with current >software and 2090) which means the files are stored as phyically contiguous >sectors for increased speed, It buffers full tracks (a great trick since it >is often just as fast to read a track as it is a single sector) and stores >data in RLL format (for about 3Xs the data per sector almost tripling the >effective Xfer rate). The tech I talked too said they intend to support >auto boot as well! Well, well, well... where should I begin... Sector interleave can be a good thing. Sector interleave is also largely transparent. With a 3:1 interleave, _logical_ sectors are spaced every three _physical_ sectors. According to the information sheets posted in the net for the WEDGE, the Amiga sends requests about every 65ms. If you do not support interleaved sectors, the disk will have moved the next sector away from your heads and you will have to wait for a full revolution to get that sector again. If you buffer entire tracks (altogether too good of an idea ;-), interleave is _not_ useful, because, with a standard algorithm, the disk must spin three times to get a track (assuming you want the entire track). Track cacheing is a good thing, IF you have gobs of memory to use up. As for RLL format, it does change the rate at which the data are clocked throught the HEAD of the disk (more bits on the track, same disk speed...). There is still the bandwidth of the bus sequence to deal with. It my understanding that Supra drives are nonDMA. This would explain their techniques: if you have a nonDMA drive which accesses sectors, by the time you pump a sector into the machine, the disk has spun 1/4 the way around (or more :-) and you must support a large interleave. If you buffer entire tracks at a time (in the drive controller, I hope), then the latency time goes away. If you have buffering for multiple tracks, say 10, then you should not experience head slamming when two tasks are both requesting disk access. A good controller should be DMA _and_ buffer several tracks at once, with a good cacheing algorithm ( hmm... how 'bout that dusty PeeCee over there in the corner?). The other issue is autoboot. Just _how_ are they going to support autoboot? Are they going to make you inert their ROMs for CAs ROMs? Are they going to patch KickStart? If you are going to use a KickStart patch, why not just use a KickBench disk, with about 4 commands in the startup-sequence? Once we get KS1.2, after the initial boot when KS1.2 is read in, one never need put a bootable floppy in the machine..._after the initial boot_. If we need one floppy boot anyway (something I personally do NOT mind), why contort the system. If you have a KS-in-ROM machine, you need to do a floppy boot anyway to take advantage of KS1.2.1, and you are in the same boat. With an actual ROM patch, you would need to make the harddisk a KickBench, or be in the same boat as the rest of us: inserting a floppy for some part of the boot process. How many of you have ever booted a large machine where the bootstrap was on some other medium (floppy, cassette, PAPERTAPE)? The idea of booting a machine from some form of removable medium is not a strange idea (although it is not always the best idea when you have turnips operating the computer; the IBM PeeCee would never have gotten anywhere with a boot floppy...too many turnips use PeeCees at the office every day). I realize that for many users, hard drives are a black box which sits on the side (or in the side) of the machine and is supposed to solve all the problems of the user, but do it _RIGHT NOW, FASTER, FASTER, FASTER_. When us humble user types finally get WB1.3, I think we will all notice a much faster response (like 2X-3X) from our hard drives and a similar response from our floppies (no, not inside information, just assumptions based on public knowledge). It will be at this point that the users will be able to tell the difference in speed between a DMA drive and a nonDMA drive. Right now, nonDMA drives are MUCH MUCH cheaper ($300+). My bets are with the WEDGE, but I cannot really say, since I have not tested it yet. IMHO, the WEDGE is faster than some, and cheaper than ALL harddrives (if you assemble it yourself ;-). I will indeed post a review once mine is up and running. Sorry to be so long winded, but I think that there are too many misconceived notions about hard disks. -ethan -- Ethan R. Dicks | ###### This signifies that the poster is a member in 2433 N. Fourth St| ## good sitting of Inertia House: Bodies at rest. Columbus OH 43202| ## (614) 262-0461 | ###### "You get it, you're closer."
dg2l+@andrew.cmu.edu (Douglas Phillip Ghormley) (12/12/87)
In article <2855@sdsu.UUCP> uzun@sdsu.UUCP (William Uzun) writes: > ...but the controller is what got to me! >It formats the disk with a 1:1 sector interleave (versus about 3:1 with current >software and 2090) which means the files are stored as phyically contiguous >sectors for increased speed, It buffers full tracks (a great trick since it >is often just as fast to read a track as it is a single sector) and stores >data in RLL format (for about 3Xs the data per sector almost tripling the >effective Xfer rate). The tech I talked too said they intend to support >auto boot as well! Correct me if I'm wrong, but you would use this controller instead of the 2090? What controller is it and is it compatible with the A2000 hardware? Douglas Ghormley dg2l+@andrew.cmu.edu
uzun@sdsu.UUCP (william uzun) (12/12/87)
The Supra hard disk controller (according to the information I was able to get from them) is a DMA device. The ability to read physically contiguous sectors in a bus which can support the bandwidth (like Zorro II) is a definate speed advantage. This is really only true if the FS makes some sort of effort at keeping files contiguous on the disk like the new FS will supposedly do. I agree that the new FS will give a much needed boost to HD performance but any little bit can't hurt. Considering the cost, the full track buffering and RLL format characteristics of the Controller (and the fact that according to their literature it is also a DMA device) I cannot see this controller drive combination as being in any way inferior to the 2090. But only time will tell since I cannot make any conclusive statements until I can do some benchmarks on each system. If they support the same auto boot as 1.21 (that again is what I am told by Supra) and I get ahold of 1.21 ROMS from CA when released I for one favor booting from the hard disk, of course again this is only my preference. -Roger
uzun@sdsu.UUCP (william uzun) (12/13/87)
> 2090? What controller is it and is it compatible with the A2000 hardware? > > Douglas Ghormley > dg2l+@andrew.cmu.edu The controller and a 30M hard disk are being sold by Abels supply for $675.00 Yes it is an A2000 compatible DMA controller for SCSI hard disks. It should be faster than the a2090 but I will not state that it is until I get a chance to use it and do some benchmarks. They also sell a controller by itself but the tech said it is not the same one. They (Supra Corp) are located in Albany Oregon. -Roger
uzun@sdsu.UUCP (william uzun) (12/13/87)
> Note that on page 24 of the 2090 user's guide it indicates that > you are allowed to soft select ANY interleave via an entry in the > MountList. In fact, they suggest a value of 0 (meaning 1:1). > That figures, all of the 3rd party drive manufacturers assured me you were limited to 3:1 or 5:1 interleave with the CA 2090 and the current FS, oh well like I said it is still my belief that the Supra Drive will be noticeably faster, and I'll report as soon as they send me one on the results. An SDPI it isn't but I hope it gives me some speed advantage! -Roger
grr@cbmvax.UUCP (George Robbins) (12/28/87)
In article <2861@sdsu.UUCP> uzun@sdsu.UUCP (william uzun) writes: > > Note that on page 24 of the 2090 user's guide it indicates that > > you are allowed to soft select ANY interleave via an entry in the > > MountList. In fact, they suggest a value of 0 (meaning 1:1). > > > That figures, all of the 3rd party drive manufacturers assured me you > were limited to 3:1 or 5:1 interleave with the CA 2090 and the current > FS, oh well like I said it is still my belief that the Supra Drive will > be noticeably faster, and I'll report as soon as they send me one on the > results. An SDPI it isn't but I hope it gives me some speed advantage! Interleave comes in two flavors. One is where the controller can't do multi-sector transfers if the sectors are contiguous and the other is where you try to adjust the interleave so that when software requests sequential block transfers, the next block is about to appear at the read-write head. The Commodore controller *does not* suffer from the first problem, unlike some of the PC type controllers. The second issue is basically a matter of tuning, and probably doesn't make a whole lot of differece with the current file system. Controllers that offer intelligent cacheing of the filesystem data such as promised by ASDG may offer major perforance gains over the Commodore controller with the current file system, however this advantage will probably decline somewhat with a better filesystem. Controllers that provide simple track buffering may offer some improvment now, but may or may not pay off in the future, depending on whether the software does it's own track buffering. -- George Robbins - now working for, uucp: {uunet|ihnp4|rutgers}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@uunet.uu.net Commodore, Engineering Department fone: 215-431-9255 (only by moonlite)