Date: Mon, 13 Dec 1999 20:33:45 -0700 From: Brett Glass <brett@lariat.org> To: Terry Lambert <tlambert@primenet.com> Cc: dscheidt@enteract.com, noslenj@swbell.net, chat@FreeBSD.ORG Subject: Re: dual 400 -> dual 600 worth it? Message-ID: <4.2.0.58.19991213200556.0473c1e0@localhost> In-Reply-To: <199912140040.RAA27620@usr08.primenet.com> References: <4.2.0.58.19991210230453.046806e0@localhost>
next in thread | previous in thread | raw e-mail | index | archive | help
At 05:39 PM 12/13/1999 , Terry Lambert wrote: >I will let you in on a "secret": SCSI drives cost more because >that's what the market will bear, based on their performance >characteristics relative to IDE. Unfortunately, I don't believe that the price/performance ratio of UltraSCSI is anywhere near that of Ultra-66 ATAPI. >They cost the same to manufacture; it doesn't matter what mask >you use to burn your 1 square inch ASIC. Which is the problem. You're being charged a premium for hardware that's very similar, due to lower volume. And SCSI has higher command latency than IDE. SCSI drives usually make up for this with tagged command queueing, hidden elevator seeking, and larger on-drive caches. Sometimes this is a clear win, but sometimes it is not. The ideal thing would be a hybrid: a drive which supported the full SCSI command repertoire but didn't have the overhead of selection, arbitration, bus settling time, signal deskewing, etc. I would do this by making the one drive per ATAPI cable act like a SCSI device that was always selected, eliminating the bus selection phase altogether. The interface would be cheaper, because straight TTL is much less expensive than the terminators and transceivers needed for SCSI. It'd be more energy-efficient, too. And it'd have a higher peak transfer rate, because SCSI is limited by having to handle the more varied transmission line characteristics that come from an 8-foot or 16-foot cable with multiple irregularly spaced taps. The "SCATAPI" drives would use a 1 meter cable -- ALWAYS 1 meter, even if you could get away with less. Fold it up neatly if it's too long. No taps, 28 AWG conductors, controlled impedance, and twists in the signal lines all the way. Peak speed ought to reach 132 MBps easily. This just happens to be the capacity of 32-bit PCI. A later generation could move up to AGP speeds and run off the motherboard chipset's AGP circuitry. CAM would work with no modification. --Brett To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4.2.0.58.19991213200556.0473c1e0>