home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.amiga.hardware
- Path: netnews.upenn.edu!dsinc!scala!news
- From: dave.haynie@scala.com (Dave Haynie)
- Subject: Re: Reports from CeBit
- Sender: news@scala.scala.com (Usenet administrator)
- Message-ID: <1996Mar22.182920.5880@scala.scala.com>
- Date: Fri, 22 Mar 1996 18:29:20 GMT
- Reply-To: dave.haynie@scala.com (Dave Haynie)
- References: <Pine.BSD/.3.91.960320130638.11700B-100000@sos.sos.net> <4ishv7$jfr@kaon.kuai.se>
- Nntp-Posting-Host: gator
- Organization: Scala Computer Television, US Research Center
-
- In <4ishv7$jfr@kaon.kuai.se>, flex@kuai.se (Anders Karlsson) writes:
- >In a message of 20 Mar 96 Pete and Vicki wrote to :
- >
- > >> EPP Parallel port
- >
- > PaV> Oh boy, that'll be handy for all those stupid devices like the
- > PaV> DITTO... hanging off a parallel port, giving poor
- > PaV> performance...
-
- Uh, the speed problems with something like a Ditto are due to the
- ditto, not its location on a parallel vs. SCSI port. Cheap tape drives
- are slow, period. That's why a Ditto costs $150, and a fast 8mm tape
- costs $1000.
-
- >That EPP par-port has performance in the range of 500KB-2MB per second.
- >I don't consider that crap acctually..
-
- It's much better than the current Amiga parallel port. Not surprising,
- really, given the Amiga parallel port is just a _slower_ C64 user
- port using the standard PC parallel pinout. The 8520 is just a
- modified 6526, a design from the 70s.
-
- > PaV> Well, why bother with SCSI, when you can have EIDE right? EIDE
- > PaV> stinks! The performance is lacking,
-
- Real EIDE is faster than fast synchronous, 8-bit SCSI-2. EIDE tops out
- at around 13.5MB/s, FS SCSI-2 at 10MB/s (that's A4091 and A4000T),
- regular synchronous SCSI at 5MB/s (that's A3000 on down).
-
- >it doesn't support enough units,
-
- That's generally true, though many EIDE controllers support two EIDE
- buses, with two devices per bus. That's not as good as SCSI, but it's
- more reasonable.
-
- >There is a new BIOS available for PC's that gives EIDE the possability
- >of 8 attached HD's,
-
- Nonsense -- the selection of disks on the IDE/EIDE bus is a hardware
- control feature. You can, in some circumstances, split the bus in
- two. There's a hack for the A4000 that takes its one IDE bus and
- splits it into two separately addressed buses. That's fairly easy,
- based on the way the A4000 implemented IDE, but it's far less likely
- to work on a PC, where the IDE bus is driven in limited I/O space,
- rather than via memory mapping. On any given IDE bus, you can have one
- master device and one slave device, period. This is a feature of the
- devices themselves; there is no intelligence in the IDE bus itself,
- it's all in the devices (historically, IDE evolved as a subset of the
- PC-AT bus -- the idea was that you would put the hard disk controller
- in the drive, and just run the AT bus to the drive).
-
- >and a stunning top performance on HD's at 200MB/second.
-
- Bunk. The official limit on EIDE is 13.5MB/s. You might be a tad
- faster than that with selected drives, but not significantly so. The
- 16-bit datapath, TTL signalling, the quality of the cable, and the
- layout of the IDE signals on the cable all set upper limits on what
- you can manage with the protocol.
-
- >(I know, there isn't any drive that does that today.)
-
- There isn't as PC with a CPU to EIDE connection that fast, regardless
- of whether the drive, cabling, etc. would support it. Keep in mind
- that the 32-bit PCI bus, which was designed very carefully to support
- high speed transfers, is limited to a peak of 132MB/s.
-
- >So EIDE can be an alternative to SCSI pretty soon.
-
- Like I said, EIDE is already faster than SCSI, until you go to 16-bit
- or Fast-20 SCSI. In either case, it's faster than any single drive
- available today, though of course bursts from drive cache may actually
- run at the full speed.
-
- > PaV> IDE/EIDE just DOES NOT CUT IT! Aaaaargh, you'd think they'd
- > PaV> learn after the A4000/A1200.
-
- >The A4k/A1k2 HD-interface was a common IDE interface. Not EIDE.
-
- In fact, the A4K/A1200 IDE interfaces are the slowest form of plain
- IDE. Generic IDE can go faster, EIDE faster than that. Realize them
- for what they are: kludges. The IDE bus was a requirement for the
- A4000 because management forced it to be, and it was added as well as
- it could have been considering it wasn't planned for in the A3000
- architecture (upon with the A4000 is based). The IDE in the A1200 was
- necessary to support some kind of hard disk, in a low-end machine, for
- as little incremental cost as possible. That's what IDE is for. And it
- still is: EIDE can give you better-than-SCSI performance for little
- additional system cost. That's why every PClone uses it, few ship with
- SCSI even through SCSI is superior to EIDE as far as the device
- supports go.
-
- >EIDE has better performance and can have 4 devices instead of 2.
-
- IDE and EIDE both support two devices per bus. It was common to have
- one IDE bus in IDE systems, and it's recently become popular to offer
- two EIDE buses in EIDE systems. But nothing has changed the basic
- design: one master, one slave per bus.
-
- Dave Haynie | ex-Commodore Engineering | for DiskSalv 3 &
- Sr. Systems Engineer | Hardwired Media Company | "The Deathbed Vigil"
- Scala Inc., US R&D | Ki No Kawa Aikido | info@iam.com
-
- "Feeling ... Pretty ... Psyched" -R.E.M.
-
-