While it would probably not be very noticeable given how much faster a high end SD card is than those old SCSI drives, I want to point out that SD card write performance in particular is extremely sensitive to format. For example, try formatting one of the SanDisk ultra high speed SD cards using HFS+ instead of FAT32 and then benchmark them. They'll go from hitting their rated speeds (45 or even 90+MB/s if you have a fast enough reader/writer) to 2.5-8MB/s depending on the reader and SD card. I discovered this initially when I was fooling around with a Sandisk extreme plus 64GB microSD and installed OSX to it. The 64GB ones are less than half as fast as a 32GB if you format them HFS+, and the 32's write at only 8MB/s once they're HFS+. Reformat it FAT32 and suddenly it is fast again. I'm not sure how much this is because flash cells are not write-aligned to file system blocks and how much is due to the flash controller not understanding HFS+, but the below discussion of sector sizes prompted me to mention it. It may be that 2.5 or 8MB/s is as fast or faster than the ancient Commodore hardware interface can manage and you'll never see a read or write bottleneck, but you might still want it to run at its full speed when you insert it into a reader on a PC to move data between platforms. Justin > On Jan 28, 2014, at 19:54, Ethan Dicks <firstname.lastname@example.org> wrote: > >> On Tue, Jan 28, 2014 at 8:29 AM, didier derny <email@example.com> wrote: >> Apparently a solution almost exist to repair the broken 9060/9090 >> >> a board: SCSI2SD permit to connect a board converting SCSI to a sdcard >> http://www.codesrc.com >> >> the commodore is not SCSI but SASI some differences exists: >> >> - no termpower on SASI (it is possible to build a custom cable with the >> termpower wire removed) >> - on the 9060/9090 the ATN and Parity Lines are not connected >> - the 9060/9090 is using vendor specific commands. > > Some of those vendor specific commands are related to low-level > formatting the HDA through the SASI bridge, so they wouldn't apply to > a more modern SCSI target anyway. Parity is something that could be > enabled or disabled in early SCSI targets. Perhaps it could be > disabled in the SCSI2SD (especially if you have to load custom > firmware in the SCSI2SD anyway. See 'sector sizes' below). > >> this board has a support for a replacement of SCSI drives in apples. > > A very middle-of-the-road platform that doesn't require any exotic > options on the SCSI2SD. > >> he could eventually add the support for SASI commodore 9060/9090 if >> someone is interested. > > Start with the fact that the D90x0 firmware uses and expects 256 byte > sectors. The same physical drive when used with contemporary PC MFM > controllers is formatted to 17 512-byte sectors per track, but in the > D90x0, it's formatted to 32 sectors of 256 bytes. > > Start with how the SCSI2SD would handle that and you will have gone a > long way to getting it to work. > > The alternative is to completely rewrite the D90x0 firmware to work > with 512 byte sectors but ignore half the data space (not a problem > with modern hardware because the max filesystem with DOS 3.0 is around > 16MB, even if you extend the geometry past the real size of a TM603S). > > We've devolved to the point where all drives act like big wads of > sectors that are perfect (no bad blocks to work around) and are always > 512 bytes per block. Things were not always so, and a lot of the > options and flexibility and extra complexity of 30 years ago has just > vanished. It comes roaring back when you want to hook a modern device > to a vintage one. It's not an unsolvable problem, but there are > specific technical issues to be overcome. Unlike a 1541, of which > millions were made, the D90x0 drives only number in the thousands, so > it's less appealing to custom-craft hardware when you might only sell > a couple dozen boards. It also comes down to "how much would you > pay?" If the answer is "$200", then there are a lot more options than > if the answer was "$50". > > Please don't take this as discouragement, but it is meant to point out > a specific technical problem which must be solved by modifying the > thing at one end of the SASI cable or the other for it to work. > > -ethan > > Message was sent through the cbm-hackers mailing list Message was sent through the cbm-hackers mailing list
Archive generated by hypermail 2.2.0.