PCMCIA on 68k Macs - Speed (in ns) & Byte or Word Mode?

[Paralel]

Does anyone know the speed, either 255 ns or 100ns for the PCMCIA standard implemented on Macs, and also whether they operate in byte mode or word mode with compactflash cards?

I have been trying to figure out the theoretical maximum transfer rate for the Rev C PCMCIA module on the 540c, but I can't really determine it without this information.

My best guess is that since the PDS it plugs into is 16-bit, and runs at 15.7 MHz, the peak data rate should be 31.4 Mbits/sec, which is approximately the maximum speed of PCMCIA 16-bit @ 255 ns in byte mode.

If it actually runs at that speed I'm going to want to get a CF card that is capable of at least 30 Mbits/second to max out the speed of the main storage media I'll be using in the machine.

 

[Elfen]

According to Wikipedia ( http://en.wikipedia.org/wiki/PowerBook_500_series#PCMCIA_.22card_cage.22 ) its 16bit word. But the last line of the "article" gets to me:

PC Card (PCMCIA) cage, 16bit, 2 Type I/II or 1 Type III cards, using a 68000 CPU to convert the PC Card protocol to PDS.

With the other laptop as far as I know, the PCMCIA is straight to the main CPU. Using a second processor for signal processing only slows down the device as I see it.

But as I see it on my PB190, 5300, and 1400, booting from the PCMCIA is twice as fast as booting from the IDE. The IDE I clocked at just over 15MB/sec and I believe is 8bit. These two reasons alone slows down a CF to boot from the IDE when compared to the PCMCIA.

 

[Paralel]

I don't see anywhere on that page it says 16-bit word mode.

 

[Elfen]

It says 16 bit, which 16 bit is a word.

As I remember it, 1 bit is a bit, 8 bits is a byte, 16 bits is a word, and 32 bits is a long word.

Quote the wiki page:

The different revisions of the PCMCIA module were released by Apple to accommodate the developing PCMCIA standard. These modules are difficult to find, and the RevC module is in particular demand because it alone works with 16-bit WiFi cards.

So the Rev C Cage works in 16bits.

 

[Paralel]

http://en.wikipedia.org/wiki/List_of_device_bit_rates

If you look under "Portable" you'll see the PCMCIA is listed as operating in 16-bit, and has two different modes, byte mode, and word mode, which has absolutely nothing to do with the fact that it is 16-bit.

 

[Elfen]

Interesting, I did not know that.

In the Old '486 days, the older '486 PCMCIA Slots were 8bit while Pentiums introduced the 16bit PCMCIAs. And in those days, though you could not get a 16bit card into an 8bit slot, you can get an 8bit card into a 16bit slot and but it will run in 8bit mode. I thought it was this mode you were speaking of. My old '486 Toshibas keep reminding me of this as the many 16 but cards I own does not fit into their slots.

The behavior however, as I stated above, booting a CF on PCMCIA is a lot faster than booting it off the IDE. About twice as fast.

 

[Gorgonops]

In the Old '486 days, the older '486 PCMCIA Slots were 8bit while Pentiums introduced the 16bit PCMCIAs. And in those days, though you could not get a 16bit card into an 8bit slot, you can get an 8bit card into a 16bit slot and but it will run in 8bit mode. I thought it was this mode you were speaking of. My old '486 Toshibas keep reminding me of this as the many 16 but cards I own does not fit into their slots.

Say what?

PCMCIA was designed from day one as 16 bit, although the standard does accommodate falling back to 8-bit mode when used on lightweight hosts. (For instance, an 8088-based machine like a Poquet PC or HP Palmtop.  Something tells me you're confusing the difference between "standard" PCMCIA (which, again, is 16 bit with semi-optional 8-bit fallback, implemented by a signal line on the bus and otherwise basically a superset of ISA) and "32 bit CardBus", which is technically a completely different bus (a flavor of PCI) implemented on a backwards-compatible connector as being "8bit vs. 16bit". The keying on CardBus cards *is* different to prevent them from being plugged into a machine equipped with a "standard" PCMCIA controller and, yes, most 486 laptops don't support CardBus while most Pentium laptops do.

(To support CardBus the system's chipset has to support PCI, which is why CardBus cards likewise don't work on any Mac laptop prior to the 3400. Very few 486 laptops were PCI based, most used some combination of ISA and VESA Local Bus for their internal peripherals.)

Easy way to tell if you're looking at a CardBus card: almost *all* of them have gold-colored metal plates on top of the bus connector. For plain PCMCIA those plates are either plain metal or absent entirely.

 

[Paralel]

Looks like I was able to answer my own question, the PDS the PCMCIA module hooks into is 16-bit @ ~16 Mhz, so it has a maximum transfer rate of ~32MB/s, not including latency, overhead, etc... which is more than even the fastest PCMCIA 2.1 PC Card mode rate of 20 MB/s, so the limiting factor will mainly be the PCMCIA transfer rate, as long as I get a card that is 20 MB/s or faster.

I am looking forward to testing how fast the PCMCIA transfer rate really is with some benchmark programs.

 

[Gorgonops]

If the 68030 bridge side of the PBX can handle a transfer every cycle I'll eat my hat. Burst mode on the '030 is a word every other cycle and that's usually only applicable to RAM, so realistically the upstream bandwidth of the PCMCIA controller is probably closer to eight MB/s, not 32. (Coincedentally,that's about what ISA is capable of.)

 

[Paralel]

Good to know. Still, light years ahead of the 1.5 MB/s on the SCSI bus.

The one thing I've never been able to figure out is why they use an '030 bus, but an '040 processor?

 

[Elfen]

Paralel, do you find the speed on a PowerPC PowerBook's PCMCIA to be the equal or superior to a 68K PowerBook's PCMCIA?

From what we are seeing where, a PowerBook 190 is a bit slower than PowerBook 5300ce; with a hypothesis that a PowerBook 5300c/cs being equal. This would mean that a PowerBook 68040 at 66/33 MHz has the same bus speed as a PowerPC 603e on the 5300c/cs but the 5300ce having a slightly faster data bus.

Incidentally, a PowerBook 1400 117/133 seems to have  slowed down bus while the 1400 166 seems to have a slightly faster bus, but how would they compare to a 68K PowerBook 520 - 550 or a PowerBook 190?

 

[Paralel]

I never had the card cage when the PPC upgrade was still in my hands, so I have no idea. I will be able to test the Rev C. card cage speed with a PB550 CPU.

 

[Paralel]

From the benchmarking I've done on the 540c, the PCMCIA is no faster on reading than an internal SCSI -> CF adapter, but the writing is faster, at least twice as fast, possibly faster on larger files (larger than a floppy or two)

I'm fairly disappointed. I know the CF card I have can read way faster than that, so it must be something else choking the speed somewhere.

 

[Elfen]

Something is choking the speed on the PCMCIA on your 540. The 520/540/550's are the only Powerbooks with a PCMCIA slots that was part of an upgrade package and not built in, and as such has its own CPU controlling that on top of the Powerbook's expansion module and are indirectly patched with buffers onto the data bus. In other Powerbooks the PCMCIA are tied directly to the Powerbook's expansion module chip and directly to the data bus.

But the CF on the 520/540/550 should be a lot faster than a hard drive through the PCMCIA. I do not trust a lot of benchmarking software, even though they measure things that says how things are. I do notice that the CF is at least twice as fast than a hard drive on the IDE BUS. Boot Times, Application Load Times, Application Quit Times, and ShutdownTimes is seconds on a CF when compared to a hard drive on the IDE Bus. The CF is almost twice as as fast as that on the PCMCIA than on the IDE BUS. This should be reflected as similar with the SCSI Bus.

My work is on the PB 190/5300/1400/etc. - all Powerbooks with IDE Buses. Using the 190 as a standard, it boots from Smiley Mac to Desktop in under a minute with System 7.6.1 on the PCMCIA - about 45 seconds. It takes the hard drive over 3 minutes for it to load up. With the CF on the IDE Bus, booting it about a minute and a half, still faster than the hard drive. But benchmarking software would says, "Blah Blah Blah CF Drive loads a 5K file 100X in over three minutes..." which is not what it is reflected in the real world actions.

Think; 5K X 100 is 500K, which it takes the benchmarking software to load in thee minutes. But it boots up a 1 - 4 meg System Files, who knows how many extensions, and finally a 500K to 1.2meg Finder before it gets to Desktop in under a minute. So who is lying? Sometimes it is better to sit there with a stop watch and see how long it takes from Smiley Mac to Finder, Application Load, Application Quit and Shut Down than to let some program tell you otherwise.