mdeale@cosmos.acs.calpoly.edu (Myron Deale) (08/20/89)
Hello, there was an article in a major electronics journal lately that suggests "the industry is moving to IPI." IPI stands for Intelligent Peripheral Interface, of course. I guess Imprimis has announced an 8in. drive w/ dual read channels capable of 6 MB/s, sustained or peak I'm not sure, and a native IPI-2 interface. Not bad :) But why go with an under-developed interface such as IPI-2 ? SCSI has a lot of software supporting it which implies it is reasonably well understood; you can get cheap SCSI protocol IC's that'll go 5 MB/s; and the prices are lessening as momentum picks up. Both IPI and SCSI have upgrades in the planning process. SCSI plans to go 40 MB/s (through "wide" and "fast") and IPI is projected to go 50 MB/s with a max cable length of 50m when ECL drivers are used. Someday IPI may go to fiber optic cabling, but that may compete with FDDI or SONET. In addtion, there's one company that markets protocol chips for IPI; they sell a set of 4-5 chips w/ an average liscensing fee above $50,000 ... and don't forget the cost of the chips. Sounds like a lot of samolians. Maybe I'm missing some key facts, but is there any advantage to design with/for IPI ? If you want high-end communication speed, like between a super' and it's "front-end I/O" handler you might plan for Hyperchannel or FDDI. If you want to hook up blinding disk bandwidth directly, then you might want a parallel disk array. Or if you want reasonable speed with mainstream products, then you might design with SCSI (and upgrades) and allow more controllers to be added by the consumer. Summary: I just don't see any significant improvement for IPI over SCSI. Perhaps what's needed is a RISC version of SCSI :^) -Myron // mdeale@cosmos.acs.calpoly.edu PS. What is CDC thinking, selling off Imprimis?
hutch@celerity.uucp (Jim Hutchison) (08/23/89)
Lets start by noting that IPI is not a bright-shiny-new interface, but a set of interfaces which have been in use by folks like Burroughs, Gould, and IBM for a while now. IPI-2 is akin to a better SMD, and IPI-3 is akin to a better SCSI. >[...] I guess Imprimis has announced an >8in. drive w/ dual read channels capable of 6 MB/s, sustained or >peak I'm not sure, and a native IPI-2 interface. Not bad :) Lock 2 heads together, and *poof* HSMD disk technology becomes 6 Mbyte/s with a new cable/electronics which send byte data instead of clock/data/control. IPI-2 disks are not limited to 6 Mbyte/s (like SCSI), they are limited by a "new" number of ~30 Mbyte/s. IPI-3 has a similar cable/driver based limit. > But why go with an under-developed interface such as IPI-2 ? Basis for this statement please? >SCSI has a lot of software supporting it which implies it is reasonably >well understood; you can get cheap SCSI protocol IC's that'll go >5 MB/s; and the prices are lessening as momentum picks up. Bulk software == well understood, maybe. Is it documentation you lack? SMD is also well understood, but people are getting tired of the low bandwidth (<3 Mbyte/s). > Both IPI and SCSI have upgrades in the planning process. SCSI >plans to go 40 MB/s (through "wide" and "fast") and IPI is projected >to go 50 MB/s with a max cable length of 50m when ECL drivers are >used. Someday IPI may [...] Good to hear they are both still moving forward. I'm a bit puzzled by the wide (32bit) and fast SCSIs. They will probably kill the SCSI market by confusing people. I personally would get miffed if I got a wide/fast interface and the *wrong* kind of SCSI disk. At the moment it would appear that IPI-3 has SCSI pegged with its ability to have a device decide what clock rate it wants for data. This allows for faster equipment in the future without causing peoples hair to friz with compatibility problems within the *standard* interface. As to the comment on single-source IPI chips being grossly expensive, I'm not to amazed. They've got the market, "why not milk it". When competition starts, prices whould drop. /* Jim Hutchison {dcdwest,ucbvax}!ucsd!celerity!hutch */ /* Disclaimer: I am not an official spokesman for FPS computing */
bruce@heather.pooh.com (Bruce Robertson) (08/26/89)
In article <564@celit.UUCP> hutch@celerity.uucp (Jim Hutchison) writes:
I'm a bit puzzled
by the wide (32bit) and fast SCSIs. They will probably kill the SCSI
market by confusing people. I personally would get miffed if I got
a wide/fast interface and the *wrong* kind of SCSI disk.
I may be remembering this wrong (no reference material handy), but I
believe that SCSI/2 allows both the target and the initiator to
negotiate the width of the transfer down from 32 bits. Therefore, you
can still use a 32-bit SCSI/2 device with an 8-bit SCSI controller,
and an 8-bit SCSI device with a 32-bit SCSI/2 controller.
--
Bruce Robertson
Hundred Acre Software, Reno, NV
Domain: bruce@pooh.com
UUCP: ...!uunet!tahoe.unr.edu!heather!bruce