Accelerating an Accelerator - Total Systems Gemini 020 to 030 for Macintosh SE

[Ubik]

I was blessed to find a Total Systems Gemini 020 for Macintosh SE (Also called the Quesse PDS accelerator for SE). These accelerators, which came out around 1988, were designed to be upgradable to the 030 and had excellent performance, besting the SE/30 and IIx in real world tests. See MacUser August 1990, page 86.
Yesterday I upgraded mine from a 68020 CPU / 68881 FPU to a 68030 CPU / 68882 FPU (25 Mhz parts). I have not yet replaced the 32 Mhz crystal oscillator, so the board is still operating at 16 Mhz. I'll update this when I try to swap in a 50 Mhz crystal oscillator, hopefully allowing the board to run at the full 25 Mhz to meet the CPU and FPU ratings. Total Systems had a 25 Mhz 030 version of the board (as referenced in the MacUser 1990 article).

Unlike many 030 accelerators for the SE, including the nice modern MacEffects Performer, the Gemini 030 at 16 Mhz gives true SE/30 performance due largely to the use of on board 70NS 4MB RAM instead of motherboard RAM.

All of these tests were done with the board installed on a Mac SE FDHD, 4 MB, BlueSCSI V2 running System 6.0.8, Gemstart 2.2 control panel, and Speedometer 3.

Performance increase after the upgrade (Speedometer 3): (Mac SE is 1.0)​

Total Systems Gemini 030 16Mhz vs. SE/30​

Total Systems Gemini 030 16Mhz vs. MacEffects Performer PDS 68030 16Mhz​

 

[JDW]

Very interesting how it was designed for two types of CPUs. Neat!

 

[JeffC]

Interesting, I have the same board, but it looks like mine is not set up to run an '030 as-is. It is missing some ICs and the '030 socket compared to yours, though the PCB has provisions for them. Previously I assumed this is because mine is an early board, only having Quesse branding compared to other versions with other vendor branding, and that the other vendors that rebranded these boards sold them with an '030. However I see yours also only has Quesse 1988 branding and does have the necessary parts to ride the '030.

My board is a little quirky. I have tried many different drivers, including GemStart and Extreme Systems. Most times (maybe always, I don't remember for sure) on the first boot the processor is recognized as an '030, after a restart it is properly identfieid as an '020. When I bought the board it did not have an FPU installed. When I tried to boot with the board installed I got some sort of error related to the FPU, this was a couple years ago and I don't remember the exact error. I installed a 68882 I had around and the error went away. The board seems to function fine and be stable after the reboot so the CPU is properly identified.

My board has a couple bodges, I assume they are original, leading me to believe mine may be a very early version before all the bugs were worked out.

 

[nekogahora280]

I also have a Gemini II installed in my Mac Plus or SE. I replaced the 20 MHz crystal oscillator with a 33 MHz one as is and swapped the RAM for 70ns modules. I haven't touched the memory wait settings on the board, leaving them as configured by the previous owner. It's said that 4MB SIMMs can be used, but when I tried swapping to 4MB SIMMs, they weren't recognized. I think the SE is more stable, so I'm aiming for 50 MHz, but while it boots, it isn't stable.

 

[Ubik]

Interesting, I have the same board, but it looks like mine is not set up to run an '030 as-is. It is missing some ICs and the '030 socket compared to yours, though the PCB has provisions for them. Previously I assumed this is because mine is an early board, only having Quesse branding compared to other versions with other vendor branding, and that the other vendors that rebranded these boards sold them with an '030. However I see yours also only has Quesse 1988 branding and does have the necessary parts to ride the '030.

My board is a little quirky. I have tried many different drivers, including GemStart and Extreme Systems. Most times (maybe always, I don't remember for sure) on the first boot the processor is recognized as an '030, after a restart it is properly identfieid as an '020. When I bought the board it did not have an FPU installed. When I tried to boot with the board installed I got some sort of error related to the FPU, this was a couple years ago and I don't remember the exact error. I installed a 68882 I had around and the error went away. The board seems to function fine and be stable after the reboot so the CPU is properly identified.

My board has a couple bodges, I assume they are original, leading me to believe mine may be a very early version before all the bugs were worked out.

Hey Jeff, after I got the board I poured over your post on 68KMLA! Amazing design. The amount of engineering effort and investment was significant. It was that rare time when upgrading the SE, Plus and 512 was a the main focus of the burgeoning accelerator industry. I’d love to know the history of Quesse.

 

[Ubik]

I also have a Gemini II installed in my Mac Plus or SE. I replaced the 20 MHz crystal oscillator with a 33 MHz one as is and swapped the RAM for 70ns modules. I haven't touched the memory wait settings on the board, leaving them as configured by the previous owner. It's said that 4MB SIMMs can be used, but when I tried swapping to 4MB SIMMs, they weren't recognized. I think the SE is more stable, so I'm aiming for 50 MHz, but while it boots, it isn't stable.

Very cool! Fascinating to see the evolution of the board. That was about the time I purchased my SE for college. I recall when I went shopping for accelerators that the Geminis went for a premium. I guess now I know why.

 

[Ubik]

A sad update. I tried to update the 32Mhz crystal with first a 50mhz one and then a 40Mhz one with no luck. The SE would not boot.
I went back to the 32Mhz crystal and all was good. (I should have stopped there).
Then I got an idea to try to remove the on board RAM and try the faster crystal, Not only did this not help, when I replaced the RAM, and went back to the known-good 32Mhz crystal, the board will not boot and the SE has a checkerboard pattern with the board.
Here's the debugging I've done so far,

- [x] Ensure RAM SIMMS and Stock Crystal are Seated Correctly and connectors are clean - Still Checkerboard
- [x] Test Stock unaccelerated SE- OK (Also works with my MacEffects Performer PDS 68030 without issues)
- [x] Test Accelerator Board RAM SIMMS on Stock unaccelerated SE- OK
- [x] Reflow Solder on board SIMM and PDS Sockets - Still Checkerboard
- [x] Try to swap in the old 020 and 68881 - Still Checkerboard

Ideas are appreciated but this might be the end of this experiment.

 

[nekogahora280]

A sad update. I tried to update the 32Mhz crystal with first a 50mhz one and then a 40Mhz one with no luck. The SE would not boot.
I went back to the 32Mhz crystal and all was good. (I should have stopped there).
Then I got an idea to try to remove the on board RAM and try the faster crystal, Not only did this not help, when I replaced the RAM, and went back to the known-good 32Mhz crystal, the board will not boot and the SE has a checkerboard pattern with the board.
Here's the debugging I've done so far,

- [x] Ensure RAM SIMMS and Stock Crystal are Seated Correctly and connectors are clean - Still Checkerboard
- [x] Test Stock unaccelerated SE- OK (Also works with my MacEffects Performer PDS 68030 without issues)
- [x] Test Accelerator Board RAM SIMMS on Stock unaccelerated SE- OK
- [x] Reflow Solder on board SIMM and PDS Sockets - Still Checkerboard
- [x] Try to swap in the old 020 and 68881 - Still Checkerboard

Ideas are appreciated but this might be the end of this experiment.

Like you, I’ve also damaged a board’s condition by installing a high-clock crystal oscillator. Fortunately, after taking a few days to cool off and retest, it worked. I don’t think it’s likely to break easily. I believe 32MHz is probably the maximum tolerance for the component configuration of your board. I’m not entirely sure how Total Systems categorized their products back then, but it seems they selected boards with higher tolerance from their production batch and fitted them with crystal oscillators. When I bought mine from a Japanese dealer in the ‘90s, the dealer ordered the top-speed 50MHz, but they contacted me saying 50MHz was hard to come by, so I ended up with a 40MHz one.

 

[nekogahora280]

Like you, I’ve also damaged a board’s condition by installing a high-clock crystal oscillator. Fortunately, after taking a few days to cool off and retest, it worked. I don’t think it’s likely to break easily. I believe 32MHz is probably the maximum tolerance for the component configuration of your board. I’m not entirely sure how Total Systems categorized their products back then, but it seems they selected boards with higher tolerance from their production batch and fitted them with crystal oscillators. When I bought mine from a Japanese dealer in the ‘90s, the dealer ordered the top-speed 50MHz, but they contacted me saying 50MHz was hard to come by, so I ended up with a 40MHz one.

The board for sale had its 68030 and 68882 removed from an IIfx and transplanted, and the 1MB SIMM was sourced with a speed of 70ns or faster.

 

[nekogahora280]

The board for sale had its 68030 and 68882 removed from an IIfx and transplanted, and the 1MB SIMM was sourced with a speed of 70ns or faster.

It’s sold out... Could it be that you were the one who helped me? If so, thank you very much!

 

[Ubik]

It’s sold out... Could it be that you were the one who helped me? If so, thank you very much!

Yes @nekogahora280 I purchased your Gemini II. I will be honored to take good care of it, and I will post my benchmarks and photos In the meantime, I hope the Gemini 020/030 comes back to life somehow. I may try this board in my spare SE.

 

[nekogahora280]

Yes @nekogahora280 I purchased your Gemini II. I will be honored to take good care of it, and I will post my benchmarks and photos In the meantime, I hope the Gemini 020/030 comes back to life somehow. I may try this board in my spare SE.

Thank you! I'll start packing it right away. I'm looking forward to you making great use of it and sharing the results with everyone.