Macintosh IIcx - Won't Power Down, then Won't Power Up, and SuperMac ColorCard is bad?

[David Cook]

I recently purchased a Macintosh IIcx where the seller indicated it wouldn't power up. I feared a battery bomb -- but I was really just after the SuperMac ColorCard with Accelerator daughterboard installed in the IIcx.



I disassembled the IIcx and was pleasantly surprised to discover a really clean board. No bombs. No capacitor leakage. I dutifully removed all of the old capacitors, ultrasonically washed the motherboard, and blew out all of the dust from the power supply and floppy drive.

I have so many old "Macintosh II Video Card"s (aka "Toby") that I always use one of them on newly acquired computers until I'm sure the voltages and computer are good. So, instead of reinstalling the cards that came with the IIcx purchase, I just installed a Toby.

The computer powered up perfectly. What did the seller mean it wouldn't power up? Did they even try?

MacTest Pro confirmed everything was great. But, after choosing Shutdown, the IIcx displayed the "It is now safe to switch off your Macintosh" dialog instead of actually powering down. Holding down the rear power button to force a power off also did nothing. This is weird because the IIcx motherboard is in perfect shape.

Swapping for a known-good power supply did not fix the problem.

The Mac IIcx is easier to diagnose power problems because the power on/off circuitry is not underneath the power supply -- unlike the Mac IIci. With a little probing, I determined the issue was in this portion (image below) of the power-off circuitry. Pin 4 of UL2 is supposed to slowly rise to 5V after power up, to act as a delay before it will honor a power down. Otherwise, pressing the power button would immediately toggle between turning on and turning off. On this IIcx, the Pin 4 signal is always low.




After powering off (by pulling the power cord), I checked the traces, measured R26, and measured C10. Capacitor C10 is shorted. But this is a brand new capacitor rated at 5x the voltage. Do you see the short in the image below?



If you don't see the short, it's because it is covered up by the capacitor. Upon removing the capacitor, I discovered something really weird. You know how sometimes a part of the leg of an old capacitor is still stuck to the pad when decapping? I am usually excellent at removing all remnants. This time, I must have missed it. The old capacitor leg remained and bridged the pads of C10 when the new capacitor was installed. I could not have made that happen if I tried.

Upon reassembling the IIcx, everything worked perfectly, including powering down.

SUPERMAC COLORCARD 24

Time to swap in the SuperMac ColorCard! *** NOW THE IIcx WON'T POWER ON ***

Uh, oh. The seller was kinda right. The IIcx won't power up with the ColorCard installed. The power supply is producing the telltale click, click, click when it detects a short.

The multimeter indicates a short on the 12V rail of the card. Fast forward: capacitor C78 is bad and is shorting the 12V rail. But why? The answer is in the image below. (No stray metal pieces this time.)



Take a look at the capacitor rating.

SuperMac installed 10 V rated capacitors in 12 V situations! Tantalums should be twice the rating -- they absolutely must not be underrated.

Replacing both capacitors with properly rated replacements fixed the card and the allowed the IIcx to power up. Thank you Astec power supply for preventing a fire.

If you have a SuperMac ColorCard check for this factory error!

My card only has v1.0 of the ROM (attached). If you have a newer ROM, would you kindly upload it?

Aside: It looks like SuperMac left MacsBug symbols in the code in the ROM. I see them following RTS opcodes (4E 75).



PERFORMANCE

The Graphics Accelerator ColorCard/24 daughterboard works, which made me happy. I was surprised SuperVideo 2.7.5 software works with the v1.0 ROM.



However, the performance is stinky. With 256 colors (the Norton Utilities default), the card is slower than the stock Apple card, except for drawing a picture when acceleration is on. This shocks me. Why would any card be slower than an unaccelerated vanilla card?



As would be reasonably excepted, the performance is even slower with millions of colors. Except, scrolling with acceleration is faster than in 256 colors. Acceleration in millions helps in all cases except CopyBits. I guess Norton copies from local memory to the card, whereas scrolling is entirely within video memory.

In summary, for 256 colors with this ROM and Control Panel, the SuperMac ColorCard 24 performs poorly and is barely accelerated. I suspect this was targeted primarly at millions of colors usage.

- David

 

[Arbee]

It's not uncommon for declroms to have MacsBug symbols. Even the ROM on the big A/ROSE-based Ethernet card has them (for the on-card 68000 routines, which is nice).

 

[David Cook]

It's not uncommon for declroms to have MacsBug symbols. Even the ROM on the big A/ROSE-based Ethernet card has them (for the on-card 68000 routines, which is nice).

That's cool. I should look at my other ROMs.

 

[MacOSMonkey]

I have been away for a while, but just logged back in and noticed this thread.

I know there was a v1.2 or v1.3 ROM for the ColorCard/24 from around the 1990 timeframe. I have it on a board somewhere.

It is definitely worth upgrading and fixed a number of issues:
-- config on the IIci and IIfx - black screen issues/round robin problems
-- added support for Thousands of colors on 13"
-- added PAL support - but not broadcast standard - may or may not work
-- all the gamma-related stuff - not implemented correctly
-- general accelerator performance improvements and bug fixes
-- added better Macsbug support - 1-bit mode swapping issues -- especially coming 24-bit -- either it didn't work or there were long delays, artifacts and screwed up gamma
-- various fixes for A/UX - losing video sync and maybe a _Debugger trap ($A9FF) in Video Reset - not sure
-- fixed Adobe issues -- FreeHand
-- fixed acceleration artifacts with screensavers - there was a lot of line garbage in programs like Pyro that made extensive use of line drawing for the fireworks

If there are any issues with SuperVideo, the contemporary release would have been v2.0.7 or 2.0.8..

The presence of debug symbols isn't so surprising, but the $A9FF in Video Reset, if present, was because the original firmware engineer didn't believe that Video Control csCode=0 (Init/VideoReset) was ever called and may have left a little present in the 1.0 ROM. But -- guess who called it -- A/UX 2.0! So, every board that had that adorable little debug trap hung the machine at A/UX startup. If you launch A/UX 2.0 and it hangs...that is the most likely reason. As I may have mentioned in another thread, for those of you who wish to detect A/UX, I think the low-memory global word is at $b200 and the flag is the $400 hex bit. Drop into A/UX and look at that location -- or ferret it out in the low memory globals. It may also be in Cards and Drivers. I forget.

In terms of performance, there isn't much of any 8-bit acceleration - mostly software speed-ups with lines, perhaps and a few other tricks. The acceleration is for 24-bit and covers most on-screen transfers/modes/regions, etc. It should be faster than other cards of the same era in 24-bit. The daughtercards ("Rider") for Calvin and Easy (Spec/24 III) were the first generation of the SuperMac video acceleration (GAL-based, as we have discussed at length in other threads). For it's day, it was great, but was eventually replaced by a Xilinx-based solution in the original PDQ and then with a custom ASIC - "Squid," or SQ01. See other threads on this issue.

I will see if I can find the ROM for you.

 

[MacOSMonkey]

Edit above: "especially coming 24-bit" -- should be "especially concerning 24-bit" ... and the point about the gamma support was that it was not implemented correctly in v1.0.

And it looks like v1.3 did ship in 1990. So say the wildly overpriced eBay gods. I sent one for David Cook to try out on his board. If it works, then he can post the files here.

 

[MacOSMonkey]

After some digging, I found my board - and luckily, it is v1.3! For some reason, I thought it was v1.2. Attached are the files from dump of my production ColorCard/24 v1.3 ROM -- including:
- binary file
- inverted binary file
- motorola srec file

The ROM size for ColorCard/24 is: 27C256

On Windows, srec_cat is a very useful tool for dealing with ROMs.
See: SRecord Tools Download

 

[jmacz]

I have been away for a while, but just logged back in and noticed this thread.

Welcome back!

 

[David Cook]

@MacOSMonkey definitely welcome back!

You set me up for success by suggesting using a 27C256 and providing the .bin file. It worked perfectly, the first time. The SuperVideo software recognized and displayed the version as 1.3.

That was the good news. The bad news is the performance is unchanged.

1. I tried the same test conditions with the same test results.
2. I switched to SuperVideo 2.07 with Accelerator 1.0.1. Same results.
3. I even tried 256 colors without any software. Same results. This suggests no acceleration in 256 colors.

From this, I conclude the updated versions of firmware were bug fixes, rather than performance enhancements.

- David

 

[MacOSMonkey]

Great! Glad those files are good - thanks for confirming! Nice to see that you could upgrade your card!

Regarding SuperVideo, I think that the ColorCard/24 v1.3 firmware update included the SuperMac VDI ("Video Driver Interface") spec - a patchable, generalized driver that could, for example, include newer accelerator or other driver replacement code embeded in SuperVideo or other dedicated INITs. The main focus of
VDI was to add additional upgrade flexibility and SuperVideo support.

SuperVideo 2.0.7 did not include any replacement code. However, even if a newer SuperVideo, like v2.75, contains a resource with new accelerator code with bug fixes or other changes - I don't know - it would still come down to the capabilities of the daughtercard/accelerator -- it does translations at the hardware level that are not patchable. It does what it does in hardware as fast as it can do it.

If the board you have is stock 1.0, then it could be from the initial production run of the ColorCard/24 -- so it may have an early version of the PAL-based accelerator hardware and code -- possibly before the switch to (faster, cheaper) GALs. There were accelerator code changes to support the newer GAL devices, but those changes were also backward compatible with the PAL-based accelerator daughterboards.

Otherwise, there is nothing special about SuperVideo and the accelerator. SuperVideo just makes driver calls that set 2 different flags in memory -- one for enabling the accelerator code -- the "Rider Enable Bit" and one for enabling the Capslock Disable functionality, where the accelerator code looks at the KeyMap low-memory global and bypasses operation if the Capslock key is down.

As I mentioned above, there is no 8-bit hardware acceleration, but there may be some accelerator code software optimizations for 8-bit that might help -- like picture drawing -- maybe some things that would impact lines and certain pattern modes - not sure.. Also, v1.3 supports thousands of colors, at least in 13" mode, but 16-bit/Thousands mode is unaccelerated. It was mostly to increase software compatability. I thought it also added at least 8-bit support for the SuperMac 16" monitor at 832x624 (55 Mhz), but I guess not. Hmm.

ColorCard/24 memory consists of 12 ZIP VRAMs that should be 4x256, or 1.5Mb of VRAM (vs. 3Mb on Spectrum/24 Series III). Therefore, the ColorCard/24 is theoretically capable of supporting resolutions of up to 1024x768 in 8-bit mode (and 19" monitors...but didn't, for marketing reasons vs. Spec/24 Series III). For example, 1024x768 at 16-bit would be exactly 1.5Mb with no room for a origin CopyBits BFEXTU buffer, so not possible, but 8-bit 1024x768 should be there as a Virtual Desktop on a 13" monitor - check SuperVideo.

Also, 16-bit/Thousands mode should be possible up to 832x624 (including smaller PAL at 768x576), whether natively or as a Virtual Desktop. At 16-bit 832x624, max RAM would be just under 1Mb (1,014K). In terms of what resolutions and/or VDTs should work, the board technically supports 640x480 (13", NTSC), 768x576 (PAL), 800x600 (maybe 16" or VDT), 832x624 (maybe 16" or VDT), and definitely as VDTs: 800x600 (up to 8-bit), 1024x768 (up to 8-bit), 1024x1536 (up to 4-bit), 2048x1536 (up to 2-bit), 4096x1536 (up to 1-bit).

The larger ColorCard/24 VDTs typically use 768K of VRAM, so it would not be possible to go up to the next highest bit depth (double) without running into VRAM boundary issues. I will have to use the board to see what other things I might recall.

I have a Toby card (black/metal IC package) and what I think is a ColorCard/24 built as v1.3 board (later rev), so maybe that would be a good comparison with your data. When I have some time, I will look at the board and we can compare parts and figures. I'm curious if it has the same cap problem as yours did. I haven't used it in quite a while, and the Toby board I found (black/metal IC package) is somewhat dusty, but should work. I might also run it on a IIci against motherboard video -- that was really the killer app, since it didn't have to deal with NuBus transfers.

Regarding your benchmarks, they generally show that ColorCard/24 in 24-bit accelerated mode is around 2x faster than Toby, extrapolating linearly to estimated 24-bit operations from your 8-bit timing (or giving Toby a baseline value of 0.33 for 3x the pixel data vs. 1.0 at 8-bit). The only things that will be accelerated in hardware will be on-screen transfers (source and destination on the card, especially including CopyBits and Region operations of all types). As you observed, CopyBits operations from main memory to the card are not accelerated, because the source is not the card and it requires a NuBus transfer. At release, it was at the top of the performance curve.

In terms of software speed-ups, you an see the impact on picture drawing in 8-bit mode with the "accelerator" enabled (not hardware) vs. the 8-bit Toby board. QuickDraw had certain inefficiencies that allowed for better software optimization.

So, that's it for now.

@David Cook - If yours is still set up, let me know about any supported 16" modes. And, I will have to set up an look at my cards when I have time. If there's no 16" mode support, then it's because of Spectrum/24 Series III, as above. ColorCard/24 was designated at small screen only, albeit with 8-bit PAL support -- perhaps since nobody in the USA used it.