68K/early PPC with onboard AAUI 100mbps ethernet?

[Cory5412]

There has to be some reasonable explanation for Apple having pulled a slot out of the Q700/900 and all the Quadras to follow?

Notably, the 700's successors did in fact add that third slot back in.

Just because there has to be a "reasonable" explanation doesn't mean it has anything whatsoever to do with the OS identifying the onboard Ethernet device (or graphics!) as if it were a NuBus slot. My original point here was to suggest that a two slot Mac doesn't deserve the title of road apple.

Q840, 660AV, and PM6100/7100/8100 have the ethernet device as a CPU connected non-NuBus device

If you look at the block diagram on the Quadra 700, it does too, so I suppose I'm confused as to how this would shape up differently on the two sets of machines. Do the AV/PPC Macs specifically enumerate them differently or is it possible that my original speculation was correct and even the non-AV Quadras do not logically show that Ethernet is a NuBus device (which would make sense to me, because physically, it is not.)

Ultimately though, even if somebody does produce a screenshot of TattleTech or another information tool saying that Ethernet is a NuBus device, it doesn't really prove anything. We know from the block diagrams that even fi the OS presents them as such, these devices aren't actually NuBus slots, and if you look at the diagram JT posted above, we know that there's more than just six possible ID spaces for NuBus "devices" in the memory map, leading back to the point that from a logical addressing standpoint, Ethernet didn't prevent a three-slot Quadra 700.

 

[Trash80toHP_Mini]

Note, I didn't say it was a RoadApple, I said it was severely compromised.

Apple moving the PDS to align with a NuBus slot as opposed to putting in a discrete location like the Cache Slot in the IIci (the only model mentioned in that comparison) was a major limitation for the Quadras 700/900 and those to follow. That's somewhat Apples to Oranges as the IIci Cache Slot was not PDS, Apple's definition of which requires access to the outside world.

Apple addressed this limitation in the PDS implementation of the slot limited X100 NuBus PPC generation. The HPV had a dedicated backplane plate. It's the G3 accelerators for that PDS which kill a NuBus slot if the HPV Card is to be retained.  []

I've never said or implied that slot starved Macs are all RoadApples, that's a misrepresentation. What I've always maintained is that lack of slots limited the Macintosh to its niche market penetration all along. The Mac never had the perceived value of the well slotted PC, a major comparison shopping selling point. Een the Apple II had higher perceived value in that regard while it remained available as an alternative.

Whether used or unused due to the functions of several PC Cards being built into the Mac's system boards, lack of slots limited expansion and hence market penetration. Six slot Macs were the equal of eight slot PCs in function, but not in perception. Macs with three or less, the lower priced more consumer oriented machines fared VERY badly against PCs in terms of expansion capabilities, a major selling point though not of practical value to the average consumer.

 

[beachycove]

What would be the correct way to test 10BT and 10/100 nubus Ethernet speeds, anyway?

 

[Gorgonops]

That's a really good question. Did anyone ever make a synthetic network bandwidth test that ran under the Classic Mac OS?

I think the standard standby answer to that question would be to time some file transfers, but I suspect that a problem you'd run into there is that the disk subsystems in most 68k Macs are going to be significantly slower than 100Mb/s Ethernet. If I were setting up to do it I suppose I'd suggest using a Netatalk server on a modern machine (to make sure that server speed isn't an issue) using the Appleshare IP protocol as the target, and transfer files of various sizes back and forth from a large RAMdisk.

 

[Unknown_K]

One large file saved directly to a RAM drive (file has to be smaller then RAMdrive) to take the HD speed out of the equation as well as the OS overhead moving many smaller files.

 

[Cory5412]

It doesn't have to be wild, just a timed file transfer test. Pick a protocol and a file and copy it from a server to your machine. Then, install 10/100 ethernet and move the cable and copy it again, time both times.

FTP might be a good way to do it because many FTP programs keep logs of what happened, so you could defer back to the computer having timed itself.

RAMdisk based testing would be interesting, especially if you're considering generally using a network server for everything, which is one situation where it would make most sense to me to install a 10/100 Ethernet card.

 

[Trash80toHP_Mini]

I'm still just scratching my head here as to what you're arguing for. Are you interested if AAUI onboard ethernet shows up like it's "in a slot" (presumably the "missing" slot in a Quadra 900, or whatever?) in Tattletech? Because we know it's not wired that way. Sure, we agree the same software stack is used for most devices in a Mac regardless of how they're wired up... but whatever?
 

That is basically correct according to this, but it's a little more complicated then that. Because the slots *also* need to be able to be represented in a 24 bit address space along with RAM, ROM, and motherboard devices Apple elected to only *use* six of the sixteen available slices, $9 through $E (out of the total range $0-$F) for card slots. This is why a Mac II-family machine can only access 8MB of RAM in 24 bit addressing mode, the first 8 1MB "minor slots" are mapped to RAM; the ROM sits at the 8MB mark, and each NuBus slot gets 1MB, staring at the 9MB mark. (I/O is stuffed into the last page at the 15MB mark). There's just no place for any additional "standard" slots to go. In principle a PowerMac (or a 68k Mac that dispensed with 24 bit compatibility) could have more slots; presumably those NuBus expansion channels must do some magic to wedge their slots into otherwise unused standard/superslot space without providing a 24 bit window.(*)

(* AHA!)I just stumbled across this decent article from an old Byte magazine that does a good job talking about Mac II memory mapping, including a mention of how an expansion chassis can add slots with, indeed, the limitation that no 24 bit access is allowed. It looks like in 32 bit mode Apple actually assigned the first 4 "superslot" areas to RAM, for a theoretical maximum of 1GB, and used the next two "slots" for ROM and I/O. So in theory the most "superslots" you could have in a 32-bit Mac would be 9... but Apple never built a machine like that. Expansion chassis slots have to live inside the Superslot space for their host card, so they only get 16MB "regular slot" assignments. But in principle you could use all 16 IDs in that space, so, sure, why not, the most NuBus cards you could ever have in a Mac II-family machine would be 96, obtained by plugging six 16-slot expansion chassis in. But you wouldn't be able to use them under System 6.

Still haven't looked for a boot drive for TattleTeching the Q700, but it sounds like Apple would have tacked in onto the eight slots it gutted from the spec for adding functions like video and AAUI to be addressed as bits of the $0 home of the Mac OS?

Didn't find the real PDF of the SN74ACT2440 NuBus Interface Controller, but a screenshot of the scanned PDF of TI's NuBus Interface Products Data Book from 1991.



This has been a really productive tangent. There's one going on in another topic. In my NuBus in SE/30 thread pics Second Wave's ExpanseSE/30 interface card have appeared. Love the Byte info on swapping TI's off the shelf chipset into the NuBus Test Card you dug up. That controller and the two bus transceivers that go with it are on the SE/30 card.  Time to collect all this info in one place!

 

[Gorgonops]

This has been a really productive tangent.

Perhaps, but as much as it pains me to point this out it is *very* off-topic in regard to what the OP was asking about. In the future it would probably be better to spin up self-contained threads if you want to delve this deeply into subjects of this nature; thread hijacking is generally considered bad etiquette.
 

That said, might as well throw this in here:

Too tired to mess with this any more tonight, but I'd put a shiny nickel on the Quadra 700 AAUI  subsystem being very similar to their fancy AAUI NuBus Card, but without the CPU and taking up a fixed PseudoSlot ID in place of that excised NuBus Slot  .  .  .

So, I can definitively say this is a bet you've lost. According to "Designing Cards and Drivers for the Macintosh Family, 3rd Edition" the PDS slot in the Quadra family machines is set up to use the "missing" NuBus slot's address space ($E0000000 superslot, $FE000000 standard slot) as its Psuedoslot memory assignment. (It also says it's allowed to map into the top 3/4ths of the bottom one GB that the Mac II memory map usually reserves for RAM, which is interesting; this suggests it would actually be possible to build a RAM card to give yourself one GB of RAM in a Quadra 900/950. That'd be amusing, if utterly useless from a practical standpoint.) According to the table on page 375 the Ethernet controller hardware is mapped from $5000A000 to $5000BFFF with a configuration PROM at $50008000-9FFFF.



[SIZE=4pt]/ [/SIZE]






This is exactly where you'd expect it to be, nestled inside the 256MB's worth of address space marked as "I/O" on the older MacII memory maps referenced in the previous articles that were linked to.

Note something important here: the bus the Ethernet controller is connected to is technically not *either* Nubus or "PDS", IE, directly to the CPU's interface pins. Starting with the IIfx Apple added an internal I/O bus that runs at some fraction of the main CPU bus speed and has an arbitration chip that allows devices on it to request permission to steal the system bus away from the CPU. (For DMA transfers to RAM, etc.) This I/O bus is completely separate from but "coequal" to Nubus. (Documentation about this bus is on page 373 of the same "DCandCftMF" manual above.) Apple's software abstracts things so devices on this bus sort of "look like" NuBus devices, but they are most certainly not, either electrically nor in how they're managed. (They don't use formalized "slot spaces", etc. As an interesting aside, notice how the control registers for the YANCC Nubus controller is mapped here.)

Browse around this manual more and you'll see that under normal circumstances in a Mac II-family machine with a NuBus controller any access to memory addresses from $60000000 to $FFFFFFFF are mapped to the Nubus controller *unless* there's a PDS slot, in which case parts of the NuBus space get allocated to the PDS. (IE, the Nubus arbitrator is ordered to clam up in that case.) So, as speculated earlier, it would indeed have been theoretically possible for Apple to make a nine-slot Nubus computer (nine "Superslot" assignments are free. In fact, if you look at the $F100000-$FFFFFFFF allocation it looks like it *may* have built in support for up to 15 "standard slots"), but for the aforementioned 24 bit compatibility reasons you can only make six of them completely functional. But the *real* takeaway here is that the Nubus hardware *is only active* when the NuBus range is being poked or prodded. If something is mapped to the I/O bus, as the Sonic ethernet is on the Quadras, it completely has nothing to do with NuBus, no matter what Tattletech might have to say about it. Period.

 

[Gorgonops]

According to "Designing Cards and Drivers for the Macintosh Family, 3rd Edition" the PDS slot in the Quadra family machines is set up to use the "missing" NuBus slot's address space ($E0000000 superslot, $FE000000 standard slot) as its Psuedoslot memory assignment.

Actually, one small correction/doh: That $E0000000 assignment isn't actually the "missing" slot. The reason the Quadra 900/950 share the same physical slot space between the last NuBus slot and the PDS connector is because they also share that slot/psuedoslot space. The "missing" slot in the Quadras is actually $9; according to the relevant developer notes those machines map their VRAM as if it were a Nubus card in the first slot. THAT is why they only have five. None of the six-slot Mac II's had onboard video.

 

[Trash80toHP_Mini]

Perhaps, but as much as it pains me to point this out it is *very* off-topic in regard to what the OP was asking about. In the future it would probably be better to spin up self-contained threads if you want to delve this deeply into subjects of this nature; thread hijacking is generally considered bad etiquette.


This is exactly where you'd expect it to be, nestled inside the 256MB's worth of address space marked as "I/O" on the older MacII memory maps referenced in the previous articles that were linked to.

Note something important here: the bus the Ethernet controller is connected to is technically not *either* Nubus or "PDS", IE, directly to the CPU's interface pins. Starting with the IIfx Apple added an internal I/O bus that runs at some fraction of the main CPU bus speed and has an arbitration chip that allows devices on it to request permission to steal the system bus away from the CPU. (For DMA transfers to RAM, etc.) This I/O bus is completely separate from but "coequal" to Nubus. (Documentation about this bus is on page 373 of the same "DCandCftMF" manual above.) Apple's software abstracts things so devices on this bus sort of "look like" NuBus devices, but they are most certainly not, either electrically nor in how they're managed. (They don't use formalized "slot spaces", etc. As an interesting aside, notice how the control registers for the YANCC Nubus controller is mapped here.)

1) I didn't start the tangent, the question about NuBus not having the bandwidth to support a 100Mb NIC came up and I responded. I was lamenting Apple's laming the 700 to compromised Mac status. Cory's "idle" sarcastic responses to that were insulting, the follow on was not so bad, even with the snark about "Fake NuBus," Those would be PDS and PseudoSlot interfaces I was talking about. Those are handled in the same manner as cards beyond the NuBus Wall by the Slot Manager. All occupy one of the six supported interrupts/Slot IDs.

2) Actually you started the productive part of the tangent.

Yep, I lost the shiny nickel bet on AAUI being the NuBus Slot hijacking culprit! As you determined, that was video being mapping to that space. Still dumb.

I'm spinning up a proper thread for the results here and elsewhere, just had to make it clear you've unjustly accused me! []

 

[Cory5412]

Going back to unproductively complaining about how many slots a midrange machine from 1991 (advertised almost always with a 13-inch monitor, worth noting): I was advised to look at a Quadra 700 motherboard. The board is totally full.

It's not for any addressing reasons, just all the hardware they added in building the Quadra filled the board and that's almost certainly why it's got two instead of three slots.

I imagine for the release of the 650/800, they were able to lay it out more efficiently or coalesce a few components into fewer, bigger chips.

 

[omidimo]

Techni-chat aside, it is worth noting that the Quadra 700 was on sale for two years, so clearly the limitations of two NuBus slots did not stop customers from buying it. If someone wanted expansion they would have gone with the Q950 and that monster was on sale for over three years. 

 

[Unknown_K]

Techni-chat aside, it is worth noting that the Quadra 700 was on sale for two years, so clearly the limitations of two NuBus slots did not stop customers from buying it. If someone wanted expansion they would have gone with the Q950 and that monster was on sale for over three years. 

Q950 was $7200 , Q700 was $5700. Neither were cheap.

 

[Cory5412]

Unfortunately, for whatever reason, prices simply Were Not Posted in MacWorld at the time, otherwise it would be interesting to go to magazines from the time and see what the catalog/magazine sellers were charging for, say, the IIci and the IIfx at the time. It would make for a better comparison than what follows.

If memory serves, the stack just before the launch of the 700 and 900 had the machines they "replaced" selling for around 1000-2000 more a pop. The IIci started in 1/0 config at $6300 and the IIfx started in 4/0 config for a bit under $8700.

The IIci included onboard video that slowed the system down and could only do 512x384, 640x480, and 640x870. The IIfx did not include video at all for that price(1). I think that the Wiki-quoted prices for the 700/900 are for basic configurations with some RAM and a hard disk. ALso note that the 700/900 have onboard that is extremely good for 1991. They can do 1152x870@256 in stock configuration, and thousands with the VRAM upgrade. They can do lower resolutions at 24-bit. This is pretty close to what the Apple 8•24 and 8•24GC can do, except generally the Quadra graphics can do it faster, although how much that mattered would have depended on the work you were doing.

My guess here is that in addition to not having room for it at the time, they figured that IIci and IIfx owners were probably adding video cards and Ethernet cards close ot universally, and if the onboard options were as good as the top-of-the-line offerings from that time, they'd save you a slot and you'd be up a slot. (a IIci/IIcx with video and ethernet installed has one slot available, a II/IIx/IIfx with video and ethernet installed has four slots available, whereas the Quadras start with the then-best-in-class versions of those things and, as such, have more available expansion for other things.)

The reasoning doesn't necessarily hold up if you keep the machine long enough to need, say, to be able to do 24-bit color at 1152x870 (the first Apple graphics solution to do that was the 4-meg configuration of the Power Mac 7100/8100 HPV card), but whether that matters depends a lot on what else you're doing.

Anyway, from a 1991 perspective, the 700 and 900 are really a great deal if you didn't already have a late Mac II and were shopping around for a top of the line Mac.

(1) I'm going to qualify this by saying I did that research a few years ago and didn't source it extremely well in that article. I wrote "GFX" and I don't know if i just forgot or if they popped in like a Toby for that price or what.

 

[Gorgonops]

As you determined, that was video being mapping to that space. Still dumb.

No, not dumb, when you consider how poor the alternatives were. Remember, the design of the Quadras still had to work around that whole issue of 24 bit mode compatibility. If you recall, in 24 bit mode the machine is chopped up so it has:

8MB of RAM space

1MB for ROM

6MB of "Nubus" space

1MB for I/O

The previous Mac II series machines (IIci/si) that had onboard video borrowed main memory for the roughly 300k of RAM they needed for video refresh, so they didn't need a dedicated spot in the memory map for that. The Quadras, on the other hand, had *much larger* framebuffers using dedicated memory, IE, more akin to "real" video cards, so where exactly could that have gone in 24 bit mode other than somewhere in the NuBus area? I suppose they could have chopped the tail off the RAM area so a Quadra in 24 bit mode only allowed 7MB of RAM to be usable, but that's not exactly a great solution.

(Also of course note that the SE/30, the closest thing to a Mac II with dedicated VRAM built-in, also uses a psuedoslot address for it. In that case it doesn't cost it a slot because, obviously, it doesn't have any. Why they did this I don't know. They must have had some reason to want to do that instead of shoehorning it into the 1MB I/O space, which otherwise would have been possible given it only had 64k's worth. My wild guess is treating it like a NuBus Psuedoslot simplified some aspects of driver development since, hey, they already wrote drivers for real NuBus cards.)

Besides, how is the customer losing anything here? All the six-slot Macs needed a video card, so they were also effectively five-slot Macs. A standard-at-the-time 8-slot PC rarely had more than 5 slots available; you lost one for a video card, and either one or two more for disk controllers and I/O, both of which were built into the Mac. Heck, if we *really* want to compare apples-to-oranges you'll have to stuff a sound card into the PC before the baselines are even. (And, if you're looking to plug in a scanner or something you'll be sucking up another card for that, since you *probably* have IDE or ESDI for your main storage controller.) When you factor in that Quadra 700 had Ethernet built in as well it actually stacks up pretty evenly against an 8 slot PC, so... no. I don't think Apple "evilly" starved their machines of slots/slash/the lack of said slots had any significant impact on their market penetration.

In the Mac II/Quadra era Macs didn't sell that well because the perception was that they cost two arms and a leg compared to a perfectly okay office PC, not because they were gimped.

 

[omidimo]

The trick with pricing is to look at the back of the magazine at the grey marketers, their pricing was the closest thing to street pricing. Here is a one from January 1992:



One of the things that Jobs did upon his return was to kill off this avenue of sales as it was eating into Apple's bottom line in the US. 

 

[1400man]

These numbers look far more reasonable. I always suspected that there was a large gulf between MSRP and street price. This was true for almost every other computer maker at the time. 

 

[Cory5412]

Here is a one from January 1992

Thank you!

It's actually super interesting, I had always basically presumed that those were more or less all illetigimate or (and, I know this one says as much to the contrary) that these were generally used machines, in the style of, later on, ShreveSystems and MacGuide.

Around 1997 or so, all of the "regular" catalog resellers suddenly started listing their prices in the Mac magazines, that must have been because of the change Jobs made.

Worth noting that at these prices, you could easily make the reverse case of what I did above (admittedly: using a bunch of numbers from ~1990, taken from reviews and announcements a few different places), that at $3333 for a IIci vs. $4444 for a Q700, it's reasonable to just drop a video card of your liking into the IIci and get going with that configuration. Although, none of that class of machines was "for you" per se if you were specifically optimizing for price.

 

[Trash80toHP_Mini]

Going back to unproductively complaining about how many slots a midrange machine from 1991  .  .  .

Play nice, you're being rude again. Devnote doesn't say if sound out was stereo on the Q700, lets hope so. The channels were tied together on the way in, hopefully not on the way out. IMO, Quadra 700 would be an egregious example of slotlessness at its price point, we can agree to disagree whether this made it a compromised Mac.

But what I said was that all Macs were slot starved from day one as compared to PCs and that market penetration suffered badly for it. Point of fact, the lowest end PC had the option of Sound blaster Pro and 8-bit Stereo out (with a side of game port) from 1990 on. Had anything like it been available (PAS16 price points don't count) to a Mac user it would have filled the single slot allotment of any LC consumer level Mac until 1993. The LC475/Q605 were released that year with stereo out at long last  .  .  .  or half the Slots in your mid-range Q700 of 1991.

Consumers spoke with their wallets. They opted for a PC running Windows 3.0 with slots for a Sound Blaster, an internal modem (in my case a FAXmodem) and any old bargain basement NIC if they so wished. Stereo on consumer Macs was three years late to that plate. But that was 8-bit sound, Sound Blaster had gone 16-bit the year before.

Whatever, Windows for WorkGroups wielding IT retreads from the Big Iron world weren't allowing Macs like the Q700-840AV onto their networks across great swaths of the business world by 1993 anyway. Deeper darker days for the Mac were already on the horizon.

Back OnT?

The list vs. street computer pricing is interesting, but how did 10Mb and 10/100Mb NIC prices compare across platforms I wonder?

What was the first Mac with built in 10/100Mb networking? Gigabit was way late.

I'd be interested in seeing the throughput jump for the same card when hobbled by a 10bT network and then turned loose in its element. The 10/100 card installed in a Mac for the present in readiness for the inevitable network upgrade was a major selling point. Measuring the bang for the buck that the infrastructure upgrade provided for the investment would be neat.

 

[omidimo]

The grey marketers were selling new units meant for other markets in the US, they were identical to the US units except for power cables, manuals, and CDs. Generally the grey marketers supplied the US variants in addition. The big change in 1997 was the Apple (online) Store, so the catalog companies compensated with bundles. Shreve fit a great niche like Sun Remarketing had, as they sold used units cheap, but many parts too. 

I really wish I had saved some MacZone, MacWarehouse, or MacMall catalogs from the mid to late 90s. Quite a few white box products are only identifiable thru the old catalogs.