philip@pescadero.Stanford.EDU (Philip Machanick) (09/06/90)
In article <1546@chinacat.Unicom.COM>, woody@chinacat.Unicom.COM (Woody Baker @ Eagle Signal) writes: > It would not be a matter of economy, but rather one of speed. Being able > to write to the disk at the full SCSI speed would ridiculously speed printing > up, as you can use the disk as a *VERY* large print buffer... It would be interesting to test this statement. In general, caching works by having a relatively small amount of fast memory, rather than a relatively large amount of slow memory. For the price of an external 20M SCSI disk, you could probably buy about 4M of RAM. I would guess in most practical applications, increasing the amount of RAM sould be a more effective way of speeding things up (RAM access time: about 100ns; disk access time: around 20ms - 200 times slower, if my arithmetic is right; not the whole story, but the right order of magnitude...). Philip Machanick philip@pescadero.stanford.edu
woody@chinacat.Unicom.COM (Woody Baker @ Eagle Signal) (09/07/90)
In article <1990Sep5.211829.4130@Neon.Stanford.EDU>, philip@pescadero.Stanford.EDU (Philip Machanick) writes: > > up, as you can use the disk as a *VERY* large print buffer... > large amount of slow memory. For the price of an external 20M SCSI disk, you > could probably buy about 4M of RAM. I would guess in most practical > applications, increasing the amount of RAM sould be a more effective > way of speeding things up (RAM access time: about 100ns; disk access time: Considering that few printers are expandable that way, those that are would see an increase in speed. Note: on most implementations the font cache is more or less static. You have no way of partitioning memory (unfortunatly). I wish that PS had a mechanism that would allow you to balance VM vs Font cache. > around 20ms - 200 times slower, if my arithmetic is right; not the whole Yes, but consider the following: I have a book that consists of about 100 or so pages. Say something like the RED book in size. The postscript code to describe these pages resides on my hard disk. Give the normal limited input buffer (again I wish there was a way to partition more memory to this as needed). it is going to take quite a while at even 19200 or centronics to transfer this. If I could dump it over the scsi port at say 5 megs per second or whatever it is that the scsi port can handle, and store it on the hard disk, then my transmission time drops to almost nothing, and while it might take an hour or two to print the book, it should only take very few min to transmit it. Increasing the memory is not going to help in this case. This is what I was refering to. :=} Cheers Woody