wcarroll@encore.com (Mr. New Dad) (10/03/90)
by liggio@cunixf.cc.columbia.edu (Vincent J. Liggio): > >wcarroll@encore.com (Mr. New Dad) writes: >> > > Well, I was commenting on the Wren IV from Mr. Zamost, which is along the > PC line (unless you get cases, I suppose). The PC line is faster than 1 > Mbyte/sec. I have tested and used Wren series drives extensively. In asychronous mode, they will burst at 1 to 1.25 MB/s. This roughly corresponds to the disk-head transfer rate. Using Zoned Bit Recording, the rate at the outside of the disk will be slightly better than the inside of the platter, but this difference will be only a few percentage points, not an order of magnitude. These rates cannot be improved without spinning the disk faster. Sustainable rates will always be lower than burst rates. Synchronous drives will provide higher burst rates, but the maximum sustainable transfer rate is still governed by the disk-head transfer rate. After you factor in controller and OS overhead, you will do very well to sustain 500 kB/s. I have never done performance tests with a Mac, but I would be stunned and amazed if it could sustain 200 kB/s. So my questions remain: >>Do any products in the Macintosh line do synchronous transfers? >>How many of the products in the Macintosh line can outrun an >>asychronous drive at ~1 Mbyte/sec? >>Fast drives are nice, but they require fast controllers. And don't >>believe every marketing brochure you see. Just because you have a >>drive that's advertised for 5 MB/s and a controller that's advertised >>at 5 MB/s doesn't mean you can plug them together and they will run >>at 5 MB/s. > > Of course not, but then again there are combinations that will. You missed the point. Every drive-controller pair has a theoretical maximum transfer rate. But you will likely have to tweak both of them to acheive this rate. I have never received a drive that was shipped with defaults parameters set for maximum performance. Caching is a good example. How many drives are shipped with the cache default as on? >>And is transfer speed even the correct parameter to look at? >> If your typical disk access >>is <= 5 kb, you're spending more time looking for the data than moving >>it from the disk. > > With drive caching, this should not be a problem, because the data should > be in the cache already (depending upon the application you are using, > and other factors....) If you expect caching to hide access times, you will be disappointed. >>And to be picky: >> >>All synchronous drives will also transfer asynchronous. It's required. > > Of course everything is a subset of the older stuff, for compatability > reasons. No, it is a superset, ie., in addition to its enhancements, SCSI-2 should include all of SCSI-1. Though I am not sure this is entirely true. > But an asynchronous controller won't add anything for a > synchronous drive as speed goes. And a synchronous controller won't do anything for an asynchronous drive. The point here is that there are MANY components involved in getting a SCSI subsystem to perform at its peak. If you think you can add a Wren Runner to your Mac and see performance increase, you will be disappointed. -- William R. Carroll (Encore Computer Corp., Ft. Lauderdale FL) wcarroll@encore.com uunet!gould!wcarroll "The brain-dead should not be allowed to operate motor vehicles!" - Me