Question about ACA1233n-40 availability

Quote from ljmarent

Are these products planned to be restocked ?

I'm working on the ACA1234, which will hopefully go into production in late April. The ACA1233n is impossible to make for the old/advertised prices, as both CPUs and CPLDs prices have gone up considerably. The new ACA1234 design makes sure that the price increase is outweighed by new useful features. 400 dollars Canadian will cut it, though not for the fastest one. 128M is standard for these new cards.

Quote from ljmarent

I see there is work being done on a 68040 card, but I'm afraid that for me, an 040 or 060 will break a lot of old software compatibility.

Many people think like that, which is why we continue to make 68020 and 68030 cards. The ACA1221lc is surprisingly fast for a 68020!

Quote from ljmarent

PS, Sorry for all the nose prints on your web-shop window.

Hi Jens, great to see the ACA1234 coming along! Will it be feature-wise the same as the ACA1233n? I am especially curious about the possibility to add a FPU. I am aware that there is not much point to adding an FPU to a 030, and probably not recommended, but still, curious... thanks!

Quote from VosCoBoss

Hi Jens, great to see the ACA1234 coming along!

Like sooo many things, production has been delayed. Partly because some ships were stuck in the Suez canal, partly because circuit board production was delayed, and of course the ever-recurring DHL-messup: Transportation takes 2 days, but passing customs takes a whopping 18 days.

Some chips are even not available until next year, so I had to switch to alternatives and make changes to the PCB before I could go into mass-production. So it's not going to be April, but more likely late May or even June.

Quote from VosCoBoss

Will it be feature-wise the same as the ACA1233n?

No. I had to drop one feature, and that's the "memory card" function for the A1200. In other words: The card can be switched off, but if you did switch it off, then it's completely off and you have no additional memory for the 68ec020 CPU.

Further, speed changes are not possible during normal operation any more. If you want to change the clock speed, changes only take effect after a reset. Possible speeds are 25MHz, 33MHz, 40MHz and 50MHz. I may add a "limp mode" if power supply fluctuations are detected - still need to see how well I can detect any power supply weaknesses. Further, maximum speed is subject to licensing - if you buy a 25MHz card, you won't be able to run it any faster unless you buy a license for the higher speed memory/bus controller.

There's one new key feature, but not all parts are completely developed yet on the software side, so I won't announce them yet.

Quote from VosCoBoss

I am especially curious about the possibility to add a FPU.

The card has an empty FPU footprint for a PGA FPU (68881 or 68882). However, same rule as with all my other 68030 accelerators: If you solder on the card, your warranty is gone. I've seen way too many fake FPUs, and my own tests show that using a 68881 FPU is pretty certain to make the card crash sooner or later.

Clock option for FPU is a bit easier, as you don't need an oscillator. You can choose clock by soldering an SMD resistor in the right place to get:

12,5MHz

25MHz

50MHz

CPU clock (whatever you're running at)

CPU clock divided by two

Sounds great Jens, thank you for the update! If no other features are dropped I am assuming it will still have an IDE speeder which is nice. I personally don't care about the "memory card function or changing clockspeed on the fly. I like how there's no longer a crystal needed for the FPU. Regarding voiding warranty when installing an FPU, I hadn't expected otherwise and that's totally fine by me

Quote from VosCoBoss

I am assuming it will still have an IDE speeder

confirmed.

Would be nice to see a ACA1234 for people to seek a great 68030 card. Have a fast ACA1233n-55 and a FPU on it I don’t care. Lot of FPU 68882 50MHz are fake and on a ACA1233n-55 it’s not the best way to do it because the warranty is gone. Have two Blizzard 1230 IV’s with FPU’s 68882 50MHz on it and don’t use it (maybe some Demos). Don’t see the benefit to use a FPU on a 68030 card.

Should be great if the ACA1234 have a 68030 50MHz CPU such as on the ACA1233n-55. Was to late to buy a second ACA1233n-55 for my fifth AMIGA 1200 project. Use the ACA1233n-55 on the ACA500Plus V2 which run great. Must use the ACA1221 on the ACA500Plus V1 for some WHDLoad Games which don’t like the ACA1233n-55.

For me I would like to buy the ACA1260 if that’s possible this year. Hope that the 68060 Rev6 CPU is on it. The Warp 1260 is only for Beta tester, the TF1260 come not out and the Vampire 1200 V2 is not the real deal and the waiting list is to long. And a Blizzard 1260 is very rare to buy.

Hope that the ACA1234 have better compatibility for WHDLoad Games and Demos than the ACA1233n-55. There is always room for improvement and my two Blizzard 1230 IV’s runs now great but for how long. Maybe the ACA1234 have better IDE Speeder on it and a new card design than the ACA1233n-55.

As for the IDE Speeder... is there also a way to speedup the PCMCIA port?

Quote from djbase

As for the IDE Speeder... is there also a way to speedup the PCMCIA port?

In theory you could shorten the access to Gayle, but since it generates the timing for the PCMCIA port by it's own pace, it just crashes if you deviate from a 68020 access. The trick of the IDE speeder is to make Gayle not "see" the IDE access at all. I've had the idea of speeding up the IDE for a *very* long time and actually tried the first steps on the very first prototype of the ACA1230 (that was back in 2010!), but gave up on it after seeing that Gayle crashes whenever you do anything different from a 68020-access.

Quote from JohanSSID

Hope that the ACA1234 have better compatibility for WHDLoad Games and Demos than the ACA1233n-55.

To be honest, I would not know what the hardware could do differently to be "more WHDload compatible" - isn't WHDload geared towards making games compatible with faster CPUs? Further, since the ACA1233n-55 has the possibility to switch back to "68ec020 with fastmem", you actually have more compatibility than the ACA1234 will offer, as that will still let you run WHDload. The ACA1234 will only have the possibility to switch off completely, so the 68ec020 does not see any extra memory. This will most likely limit you to very few games with WHDload, as you only have chip ram.

Quote from JohanSSID

Should be great if the ACA1234 have a 68030 50MHz CPU

There will be versions with a genuine 50MHz CPU, of course.

Quote from JohanSSID

Maybe the ACA1234 have better IDE Speeder

The IDE speeder is essentially the same, as it uses the same approach. I will most likely reduce the maximum speed, but if you have any trouble with the ACA1233n IDE speeder, then you always need to look at the whole chain that's involved: Activating the IDE speeder requires higher performance from your mass-storage device, so if something is not working right, you can't blame the IDE speeder alone. It might as well be your CF-card/harddrive/IDE adapter that can't provide the high speed that the IDE speeder requries. We have lots of customers being very happy with an SSD and performance beyond the 10MB/s mark.

Quote from Jens

To be honest, I would not know what the hardware could do differently to be "more WHDload compatible" - isn't WHDload geared towards making games compatible with faster CPUs? Further, since the ACA1233n-55 has the possibility to switch back to "68ec020 with fastmem", you actually have more compatibility than the ACA1234 will offer, as that will still let you run WHDload. The ACA1234 will only have the possibility to switch off completely, so the 68ec020 does not see any extra memory. This will most likely limit you to very few games with WHDload, as you only have chip ram.

Must see if WHDLoad making updates for Games for faster CPU’s. It’s more that the sound is faster such on The Secret of Monkey Island. Did do with ACATool lower the CPU speed at it stay the same problem. Did also switch back to 68ec020 with Fastmem but the ACA500Plus V2 Boot CF 3.1.4.1. don’t boot up it freeze up. Work with AOS 3.1 to switch back to 68ec020 with Fastmem but the issue with the sound is the same. With the Blizzard 1230 IV every WHDLoad Games and Demos works perfect also on the ACA500Plus V2 and V1. Don’t see any issue with the Blizzard must be the ACA1233n-55 is to fast for a 68030. Had a ACA1232-33MHz and the ACA1233n-40 and it works great with my favorite WHDLoad Games and Demos. Most of the WHDLoad Games and Demos works with the ACA1233n-55 but have the same problems as the ACA1233 55MHz 2015 Limited Edition which I use now for 5 weeks. Has swap it for the Blizzard 1240 which has CPU heat problems and I don’t like the 68040 CPU. Would nice if I had do a upgrade with a 68060 Rev6 CPU but there are so many fakes Rev6 CPU’s out there.

Quote from Jens

There will be versions with a genuine 50MHz CPU, of course.

That it’s great that the ACA1234 still use a genuine 68030 50MHz CPU. I’m a big fan of the 68030 50MHz CPU. Would be great to see the difference of the ACA1234 with the ACA1233-55 and the ACA1233n-55.

Quote from Jens

The IDE speeder is essentially the same, as it uses the same approach. I will most likely reduce the maximum speed, but if you have any trouble with the ACA1233n IDE speeder, then you always need to look at the whole chain that's involved: Activating the IDE speeder requires higher performance from your mass-storage device, so if something is not working right, you can't blame the IDE speeder alone. It might as well be your CF-card/harddrive/IDE adapter that can't provide the high speed that the IDE speeder requries. We have lots of customers being very happy with an SSD and performance beyond the 10MB/s mark.

Do also performance test with the 4GB MicroDrive which is faster on AOS 3.1 than a CF card. Use it now with the Blizzard 1230 IV. Latter this week I test it with the ACA1233n-55 maybe the IDE Speeder is faster. Maybe AOS 3.1.4.1 don’t like the IDE Speeder with SanDisk CF cards which I use. Know of the SSD performance 10MB/s mark.

Quote from JohanSSID

Boot CF 3.1.4.1. don’t boot up it freeze up.

OS3.1.4 activates the MMU library, which did cause a conflict with an older version of ACAtool. The latest version fixes that - please see if that will resolve this particular problem.

Quote from JohanSSID

Most of the WHDLoad Games and Demos works with the ACA1233n-55 but have the same problems as the ACA1233 55MHz 2015 Limited Edition which I use now for 5 weeks.

These two cards have one thing in common, and that's optimizing chipram access: Both even stretch the CPU clock by half a cycle under certain conditions in order to gain more speed when accessing chip ram.

I have gone even further on the ACA1234, as the 55MHz cards did not take every chance to gain the last bit of speed when accessing chip ram: While the 55MHz cards have "only" optimized at the end of the cycle, the ACA1234 also optimizes starting the cycle: It allows a "late start" in order to not miss a whole 14MHz cycle - not always, but under certain conditions, that is. This does not change "full speed sequential" access, as that can never exceed 7.1MB/sec, but it speeds up accesses that happen pretty rately, but still take up considerable time. The result is considerable improvements on the "writepixel" test of AIBB.

I never considered these optimizations to be an issue - after all, the 68ec020 CPU is the fastest, as it runs in sync with the chipset. The 68030 can never exceed the 68ec020 chip ram access speed, so "being slower" (aka "not optimizing") is more likely to cause an issue, and these issues even have a name in the scene: They are called "Turrican slowdowns", and can often only be overcome by activating CPU caches.

Now that you mention it, we may make these optimizations switchable in some kind of expert menu. I'd switch them "on" by default (for reasons explained above), but give you some toy-settings to switch them off. Got to see if that breaks other things in the CPLD code, as I had to make a few of these settings indicators/conditions for other parts of the logic that are frequency-sensitive. Note that this is the first time that I am offering a 68030 accelerator with an SD-Ram controller that can run fully sync with the CPU at different speeds - the ACA1233n reverts back to async RAM access when the CPU is not running at 55MHz.

Darn - I thought I'm done with feature creeping on the ACA1234

Quote from JohanSSID

Maybe AOS 3.1.4.1 don’t like the IDE Speeder with SanDisk CF cards which I use.

Since you need ACAtool to activate the IDE speeder, please use the latest version from the Wiki, which fixes issues with OS3.1.4 and it's standard activation of Thor's MMU libs.

Quote from JohanSSID

Would be great to see the difference of the ACA1234 with the ACA1233-55 and the ACA1233n-55.

Here's an AIBB Writepixel screenshot - the 55MHz card is most likely faster, as Writepixel probably spends much more time executing OS code rather than waiting for chip ram access to happen. Still, with your portfolio of ACA123x accelerators, it would be interesting to see how the older cards hold up against this. My own prototypes of the ACA1232-50 and ACA1233-55 are dead (seems like quick-turn PCBs never survive longer than a year or two), so I don't have a quick way to run this test on the older cards.

This test computer is an A1200, PAL, Kick3.1. Activate MapROM and both iCache and dCache, but give it 68000 (not 68020) code, so the value is comparable against "smaller" accelerators. Also, AIBB and other benchmark programs are sensitive to cache alignment: You get different results on different launches of the program. You can try to open windows or fill RAM disk between program launches to force a newly-loaded program to a different location.

Quote from Jens

OS3.1.4 activates the MMU library, which did cause a conflict with an older version of ACAtool. The latest version fixes that - please see if that will resolve this particular problem.

Thanks Jens it fixed it ACAtool v2.5.1.

Quote from Jens

Note that this is the first time that I am offering a 68030 accelerator with an SD-Ram controller that can run fully sync with the CPU at different speeds - the ACA1233n reverts back to async RAM access when the CPU is not running at 55MHz.

Darn - I thought I'm done with feature creeping on the ACA1234

That would be fantastic on the ACA1234. 👍👍👍

Quote from Jens

Many people think like that, which is why we continue to make 68020 and 68030 cards. The ACA1221lc is surprisingly fast for a 68020!

Oh yeah! I purchased it when I got tired of waiting for the 030 accelerators to be restocked, and didn't regret it. Running off my ACA500plus at the default 26MHz, and SysInfo shows it to be faster than a stock A3000. Almost 100% compatible with 68000 stuff (an odd demo may have problems, and some of those can be made to work by fiddling with menu options). It can be overclocked to 40MHz, but I see no need to do it.