• Hello MLAers! We've re-enabled auto-approval for accounts. If you are still waiting on account approval, please check this thread for more information.

SuperMac Spectrum - The one that started it all

I bought this old SuperMac graphics card on eBay. I knew it was going to be an early card, given that it nearly contains only through-hole parts.

SuperMac-Spectrum.jpg

The ROM label is interesting, in that it says "SPECTRUM" but not "/8" or "III" or any additional modifier. I assume 'D6' is the ROM version. Looking at the raw ROM file reveals the version to be 1.0b6. So, a prerelease ROM? Given the mid-1987 date codes on some of the other chips, I think this is a production release.

Spectrum-D6-label.jpg

With the card installed, I was in for a bit of a treat. 1024x768 resolution in 8-bit color circa the launch of the Macintosh II. Wow! I was sure it would be only 640x480.

Although no driver is required, the SuperVideo software did not work with such an old ROM.

Older-ROM.GIF

Back to eBay. Quickly!

Someone was selling an identical board with a v1.97 ROM. (The seller said this card had weak color, but it seems to be an issue with choosing the correct sync settings. The color is beautiful with @bigmessowires Mac Sync-inator. There is some waviness which the first card doesn't have. I assume something has aged poorly in the second card.)

Spectrum-8-v1.97-label.jpg

Not sure what that copyright date is. I think 1990. I copied a '0' from the part number and made it partially transparent. Seems to fit.

Copyright-year.gif

With the newer ROM, the SuperMac software works. There are seven resolutions / frequencies which can be enabled if the correct crystals are installed.

Resolutions-1.GIF Resolutions-2.GIF Resolutions-3.GIF

Does anyone know where I could get a 30.24 MHz oscillator in a full can? Not available on Digi-Key, Mouser, or eBay. Maybe a surface mount with adapter board? But, it would need to be a 5V part. It doesn't even look like programmable oscillators are available.

The ROM is a M2764A-2FI. Ha! That's only 8KB. In fact, the earlier firmware just filled up about 5KB.

ST-M2764A.jpg

The Texas Instruments chip is blacked out on both boards. Any idea what would remove permanent marker but not the chip markings? I understand the TMS34010 was popular at the time, but that number doesn't quite fit the markings.

TMS-Chip-blacked-out.jpg

The board does have a design flaw. They put the connector too high on the slot. It barely fits a monitor connector on a Macintosh II (IIx, IIfx) model.

Fits-a-Macintosh-II.jpg

But the monitor cable connector does not have enough space on the compact Mac II series, such as a IIci. Seriously, I tried. It doesn't fit.

Cable-connector-wont-fit-in-IIci.jpg

Did Apple change the slot length specifications? That seems like such a weird thing to alter.

What's in a name?

What is the official name of this video board? The initial ROM is simply "SuperMac Spectrum". The later ROM is "SuperMac Spectrum/8".

Looking through old issues of MacWorld, the first mention of this card by name is July 1987. By the way, I didn't see an 800x600 setting in the SuperVideo software, so maybe they didn't ship that configuration.

Spectrum-RAM.GIF

But, later in the issue the card is called "Spectrum 1000/1" and "Spectrum 1000/8".

Spectrum-1000-8.GIF

In that same issue, here is a comparison of the competing cards around the time of the Macintosh II launch. The card and monitor would set you back $4500.

MacWorld-July-1987.gif

A year later, RasterOps caught up with color cards, including millions of colors.

SuperMac Spectrum resolutions.GIF

But, SuperMac also released a 24-bit color card. Hence, the Spectrum/8 name stuck to differentiate it from the Spectrum/24.

SuperMac Spectrum.GIF

As of the August 1988 MacWorld issue, Apple did not support 32-bit QuickDraw. According to technical note QD01, 32-bit QuickDraw shipped in May 1989, was built into the IIci ROM, and shipped with System 6.0.5 in 1990.

August 1988 24-bit color.GIF

Anyway, it is cool to own such an initially powerful color card, even if it pales in comparison to later generations.

ROMs attached.

- David
 

Attachments

I understand the TMS34010 was popular at the time, but that number doesn't quite fit the markings.

Very neat write-up, as always! I’m wondering if maybe it’s a TMS34061FNL. That seems to fit the markings pretty well, at least I think so…
 
A year later, RasterOps caught up with color cards, including millions of colors.
Not to highjack the thread, but that would be this thing then I guess:
RasterOps_Colorboard_104.jpg
I could never get it to actually display 24bit because apparently you need a patched version of QuickDraw and an unobtainable RasterOps driver to be able to output anything else than 8bit color.
 
I have a few very old Nubus video cards, and it seems they were designed to work with special fixed frequency monitors of the time.

Finding drivers is also an issue for anything system 6.0 and older (For the Mac II which shipped with OS 4.1 ).
 
ROMs attached.
The ROMs appear to be inverted. 1 <-> 0.

Is that a normal possibility for NuBus ROMs (i.e. can Slot Manager handle either case?) or is that a problem with the ROM reader settings?

I have a few very old Nubus video cards, and it seems they were designed to work with special fixed frequency monitors of the time.

Finding drivers is also an issue for anything system 6.0 and older (For the Mac II which shipped with OS 4.1 ).
The NuBus ROMs usually have a driver built-in. It doesn't contain extra stuff like device specific control panels though.
 

Attachments

Nice write-up as usual. @MacOSMonkey may be able to provide more insights into the board you found. It is fasincating how quickly these cards evolved in a relatively short space of time.

 
The ROMs appear to be inverted. 1 <-> 0.

Correct. This is how the bits are actually stored on the ROM chip. I used a GQ-4x4 universal chip programmer to read the EPROMs. The resulting files are useful in this state to replicate the ROMs.

When the Macintosh Slot manager communicates with the board's ROM, SuperMac likely used some sort of interface chips that inverted the signals as a side-effect. Thus, the developers flipped the bits in the ROM to compensate so the Mac sees the right values.

Attached are inverted copies of the ROM so you can disassemble and explore. But, you cannot use these files to burn EPROMs.
 

Attachments

Does it work in 256 colors?
Yes, that's the only thing it does... 1024x768 at 256 colors. There's not even a 1bit or a 4bit option.

Quoting LEM: "No drivers are required for this card, but you must have RasterOps' modified version of Color QuickDraw for 24-bit output."

None of the RasterOps driver packages I could find out there know about this card, so I'm inclined to believe that LEM is right there.
 
Attached are inverted copies of the ROM so you can disassemble and explore. But, you cannot use these files to burn EPROMs.
I added an inverting check to my SlotsParse tool which will automatically invert the data that was read from the ROM file before parsing. The output is included in the zip file attached to my last post.
 
I added an inverting check to my SlotsParse tool which will automatically invert the data that was read from the ROM file before parsing. The output is included in the zip file attached to my last post.

Very convenient. You check the header signature to see if it is valid but inverted?
 
Nice write-up as usual. @MacOSMonkey may be able to provide more insights into the board you found. It is fasincating how quickly these cards evolved in a relatively short space of time.


It’s a bummer the email notifications still aren’t working. Last year he didn’t visit often but would pop on when notified he was tagged in a post.
 
Curiously it seems the card actually supports 1016x768 and not 1024x768 in both ROMs. Close enough...I wonder why they ended up 8 pixels short.
 
Curiously it seems the card actually supports 1016x768 and not 1024x768 in both ROMs. Close enough...I wonder why they ended up 8 pixels short.

I thought I was taking crazy pills when I wrote down the resolution. I just measured the size of the screenshot I took. It is indeed 1016x768. What??
 
1016=0x3F8
1024=0x400
Maybe one of the registers is 10 bits and specifying 1024 would require 11 bits.

It's like the difference between being able to support 4095 or 4096 width. The first is 12 bits 0xFFF. The second is 13 bits 0x1000.

It could be a limitation of RAM.
1024x768 = 768 kB exactly. If that is the amount of RAM, it might need to use less than that to support hardware cursor or whatever.
 
Back
Top