Amiga Accelerator Face Off! - ACA1221ec & ACA1233n
It's safe to say that I've become quite an Amiga fanatic over the last 12 months.. I've gone from a stock Amiga 1200 sitting dormant in my loft to owning 3 A1200's, an A500, a number of different accelerators (aka expansion cards) and an Indivision A1200 scandoubler. However, I'm still trying to settle on a setup that I'm truly happy with - I've been running an Individual Computers ACA1221ec accelerator for 6 months now and while it's fine for WHDLoad and (very) light web browsing/FTP transfers, my head has constantly been turned by the availability of the ACA1233n, a (potentially) more performant accelerator with a better CPU and more memory.
Anyway, after a number of months of trawling eBay, amiBay and the Amiga Facebook group, I eventually found a second hand ACA1233n at a decent price which means I can now compare the two and make a decision on which to keep.
This blog presents a number of 'real world' performance tests that I have performed with the cards to assess the differences - I'm particularly interested in performance metrics which directly affect the user experience rather than SysInfo or AIBB benchmarks (MIPS and Drhystones etc.).
The ACA1221ec accelerator is a 'budget' expansion providing a Motorola 68ec020 CPU and 16MB of memory (although only 9MB can be used as fast memory). In addition, the card includes two clock ports, one for expansions like Real Time Clock (RTC) modules and another for the RapidRoad USB controller.
It is readily available at retailers like amigakit.com for around £104 GBP excluding shipping and while that may sound expensive for a 'budget' expansion, when compared to other options such as SRAM cards or 20-25 year old '020 based accelerators I believe it is good value, especially when you consider the extra features, new components and a warranty.
The ACA1233n accelerator is a more upmarket expansion providing a Motorola 68030 CPU clocked at 40MHz and 128MB of memory. In addition, it includes a single clock port for RTC or RapidRoad expansions.
The card retails for around £208 GBP excluding shipping and seems to be quite a popular card given that it is often out of stock at retailers like amigakit.com. That said, stock levels may be affected by the availability of 680x0 CPUs rather than Individual Computers ability to mass manufacture PCBs. I'll be honest, I originally found the price a little hard to swallow, but when you consider the number of hours of R&D that goes into these niche products, along with logistics and general service, it really isn't that expensive at all - these accelerators are enthusiast products at the end of the day...
3. Archiving - measures how long it takes to compress a sample directory containing files of various sizes using Directory Opus. The WHDLoad version of Pinball Dreams will be used. Dopus compression output can be seen below.
4. Web Page Rendering - measures how long it takes to render a web page including all images using the AWeb web browser. The following image shows the web page once fully rendered.
Anyway, after a number of months of trawling eBay, amiBay and the Amiga Facebook group, I eventually found a second hand ACA1233n at a decent price which means I can now compare the two and make a decision on which to keep.
This blog presents a number of 'real world' performance tests that I have performed with the cards to assess the differences - I'm particularly interested in performance metrics which directly affect the user experience rather than SysInfo or AIBB benchmarks (MIPS and Drhystones etc.).
The Accelerators
ACA1221ec
 |
| ACA1221ec @ 28MHz (default) |
It is readily available at retailers like amigakit.com for around £104 GBP excluding shipping and while that may sound expensive for a 'budget' expansion, when compared to other options such as SRAM cards or 20-25 year old '020 based accelerators I believe it is good value, especially when you consider the extra features, new components and a warranty.
ACA1233n
 |
| ACA1233n @ 40MHz |
The card retails for around £208 GBP excluding shipping and seems to be quite a popular card given that it is often out of stock at retailers like amigakit.com. That said, stock levels may be affected by the availability of 680x0 CPUs rather than Individual Computers ability to mass manufacture PCBs. I'll be honest, I originally found the price a little hard to swallow, but when you consider the number of hours of R&D that goes into these niche products, along with logistics and general service, it really isn't that expensive at all - these accelerators are enthusiast products at the end of the day...
Hardware/Software Summary
The performance tests will be executed on the following hardware/software:
- Commodore Amiga 1200 rev. 2B (E123C/E125C timing fixes applied)
- Indivision AGA 1200 MK2cr Scandoubler (Output: 640x512 Laced)
- Kingston 4GB Compact Flash Card
- Linksys WPC11 v3 PCMCIA WPA Wireless Card
- Classic Workbench ADVSP v28
- MiamiDX TCP Stack
- AWeb Web Browser
 |
| Under the hood: 1) WPA Card 2) CF Card 3) Scandoubler 4) ACA1233n |
Performance Test Cases
The following performance tests have been chosen based upon common tasks one might undertake on an Amiga 1200 with an accelerator card fitted. In addition, each test case has a direct impact on the general user experience which is (of course) the primary objective of this exercise.
1. System Boot - measures how long it takes to boot into Classic Workbench from the moment the power supply is turned on to the desktop being presented
1. System Boot - measures how long it takes to boot into Classic Workbench from the moment the power supply is turned on to the desktop being presented
2. Icon Render - measures how long it takes to draw 26 icons in the system partition (C, Prefs, S etc). The following image shows all icons once fully rendered.
 |
| Icon scope highlighted in red. |
3. Archiving - measures how long it takes to compress a sample directory containing files of various sizes using Directory Opus. The WHDLoad version of Pinball Dreams will be used. Dopus compression output can be seen below.
 |
| Dopus compression output highlighted in red. |
4. Web Page Rendering - measures how long it takes to render a web page including all images using the AWeb web browser. The following image shows the web page once fully rendered.
 |
| Browsing like its 1995! Images highlighted in red. |
5. Image Rendering - measures how long it takes to render a 739K, 800x600 resolution JPEG image. Further tests were conducted with larger images but due to the limited graphics memory on the Amiga 1200 most failed to render at all. The following image shows an almost fully rendered image.
Test Results
The following chart presents the results of the performance tests - all tests are measured in seconds*.
You may have noticed that there is a metric missing for the web render test using a stock configuration. To get a stock CPU configuration with enough fast memory to run the tests, I have to disable the ACA1233n's CPU leaving me with the built in 68020 @ 14MHz and 8MB of fast RAM. Unfortunately I'm unable to run my A1200 with the wireless card inserted AND the CPU of the ACA1233n disabled, the system just will not boot and therefore I can not run the test. Should I find a solution, I will update the chart as appropriate.
You may have noticed that there is a metric missing for the web render test using a stock configuration. To get a stock CPU configuration with enough fast memory to run the tests, I have to disable the ACA1233n's CPU leaving me with the built in 68020 @ 14MHz and 8MB of fast RAM. Unfortunately I'm unable to run my A1200 with the wireless card inserted AND the CPU of the ACA1233n disabled, the system just will not boot and therefore I can not run the test. Should I find a solution, I will update the chart as appropriate.
Conclusions
As you can see from the test results, the ACA1233n is the clear winner in most of the performance tests (as one might expect given the CPU specifications and price). The differences are particularly notable in the archiving and and image/web page rendering tests, but certainly not so different in the system boot or or icon rendering tests. So, which one would I recommend (and which will I keep)?
In my opinion, it all comes down to what you want to do with your Amiga 1200. If you only want to run WHDLoad games from Workbench or use a few productivity applications, then the ACA1221ec will provide a (more than) adequate user experience at a fantastic price point. Even extending the use case further, if you think you may venture into the world of networking via the PCMCIA interface, the card is more than capable of running Telnet or FTP client software for those file transfers for example. Furthermore, if you love to tinker then you may be able to overclock the card to 42MHz provided you add the necessary cooling - there are no guarantees though.
On the other hand, if you are a 'more serious' user of the Amiga 1200, and would like to open (or refresh/redraw) many windows, switch between applications, view images or you're just impatient at the desktop, then the ACA1233n is going to be the bare minimum for a pleasant experience in my opinion. There will still be a little delay while rendering icons and opening applications (this isn't a 68060 afterall!), but the whole system just feels much smoother when compared to the ACA1221ec. However, I must be honest and say that a slight drawback with the 1233n is the price vs the performance, the gains just aren't that great over the 1221ec in my opinion.
So in summary, given the performance differences between the two, and the fact that I find myself in the 'more serious' user category outlined above, I will be keeping the ACA1233n and moving on the ACA1221ec at a later date. I must say though, they are both fantastic accelerators and either will serve you well!
** Update 29/08/17 **
After much debate, I decided to run the test cases against the ACA1221ec while overclocked to 42MHz - the updated results can be seen in the chart above. While the numbers look very promising, and they raise the question of whether one should pay an additional £100 GBP for the ACA1233n, I'm sticking with my original conclusion because I experienced system instability at this clock speed. While running the image render test the A1200 locked up part way though - the CPU was super hot! A low profile heatsink may solve this issue, but space is limited in the trapdoor spot and overclocking may reduce the lifetime of the card anyway so I'm sticking with the default of 28MHz.
In my opinion, it all comes down to what you want to do with your Amiga 1200. If you only want to run WHDLoad games from Workbench or use a few productivity applications, then the ACA1221ec will provide a (more than) adequate user experience at a fantastic price point. Even extending the use case further, if you think you may venture into the world of networking via the PCMCIA interface, the card is more than capable of running Telnet or FTP client software for those file transfers for example. Furthermore, if you love to tinker then you may be able to overclock the card to 42MHz provided you add the necessary cooling - there are no guarantees though.
On the other hand, if you are a 'more serious' user of the Amiga 1200, and would like to open (or refresh/redraw) many windows, switch between applications, view images or you're just impatient at the desktop, then the ACA1233n is going to be the bare minimum for a pleasant experience in my opinion. There will still be a little delay while rendering icons and opening applications (this isn't a 68060 afterall!), but the whole system just feels much smoother when compared to the ACA1221ec. However, I must be honest and say that a slight drawback with the 1233n is the price vs the performance, the gains just aren't that great over the 1221ec in my opinion.
So in summary, given the performance differences between the two, and the fact that I find myself in the 'more serious' user category outlined above, I will be keeping the ACA1233n and moving on the ACA1221ec at a later date. I must say though, they are both fantastic accelerators and either will serve you well!
** Update 29/08/17 **
After much debate, I decided to run the test cases against the ACA1221ec while overclocked to 42MHz - the updated results can be seen in the chart above. While the numbers look very promising, and they raise the question of whether one should pay an additional £100 GBP for the ACA1233n, I'm sticking with my original conclusion because I experienced system instability at this clock speed. While running the image render test the A1200 locked up part way though - the CPU was super hot! A low profile heatsink may solve this issue, but space is limited in the trapdoor spot and overclocking may reduce the lifetime of the card anyway so I'm sticking with the default of 28MHz.
Foot Notes & References
* My preferred unit of measure would normally be the millisecond, but since these tests were measured with a stopwatch I would estimate that these results could be out by as much as 500ms based upon my own reaction times.
ACA1221ec - Individual Computers
ACA1221ec - Wiki
ACA1233n - Wiki
ACA1221ec - Individual Computers
ACA1221ec - Wiki
ACA1233n - Wiki
Nice comparison! I recently acquired an ACA1221ec and noticed today that it overheats at the default 28MHz clock speed after a while especially if I run very processor-intensive software. For instance, five minutes of Alien Breed 3D II made the computer lock up and the board became very hot. Are you using any cooling methods, or does your card/system stay stable at 28MHz without any measures?
ReplyDeleteTo minimize the heat , get yourself a low profile heatsink , keep the trapdoor open and even check if you could install a mini fan in your case. Some games do require a 030 cpu but if like me you dont care to not play those few games , get the cheaper card. I dont think it is necessary to overclock. The extra ram and the cpu at 28 mhz will be more than ok for most games and whdload.
ReplyDelete