A sneak peek at the ACA1240/1260 cooling system

Quote from PeteB

Hello, I'm guessing the 1260 will like my old aca1230 still preserve PCMCIA compatibility as I run a SCSI CDrom through it and have the ide speedup of the new ACA's?

Yes, the design is 100% PCMCIA-friendly for most operational modes. At least one of the not-yet-published modes of operation will not be PCMCIA-compatible, but I guess you won't mind

Quote from PeteB

Also would you have any idea when the pre-orders can go in for the boards, if they'll ship before 01/01/21 and a ball park figure for the cost? Given the UK has made the genius decision to leave the Eu if it's sent after then I'll get hit with an increased import duty. Thank you

I Have no idea when we can start pre-orders. While development is still going on and I can't even announce all the features yet, I really don't want to jump a few steps ahead. It's always been company policy to not take any money unless all suppliers and service companies involved have confirmed their delivery dates.

Price-wise, I'm still aiming at "under 500,- EUR" for the 68040/25 version.

No news here, sorry.

Nope, definitely not dead. Just postponed, as there are so many other projects in the pipe. I tried to make nice updates for existing projects these past months, as you have hopefully seen on updates to Indivision ECS V2, Indivision AGA MK3 and Chameleon. This is because I'm expecting a huge amount of work still ahead for the ACA1240/1260, and we need lots of free resources for that. As a result, it's a good idea to close as many bug reports for the existing customer base as possible. Anything else would exceed the resources of iComp.

Quote from Sucoy

Can the high-speed port be used for other things besides rtg? Can the users put other boards of their own for what we want if there is any sdk?

It's directly on the memory bus, but it changes protocol when the specific memory area is accessed. Also, I have my doubts that you'll find the connectors, or even want to make a circuit board for a 0.8mm-pitch high-speed connector. So no, I currently don't see this as an "open connector", as it would be too easy to cause damage or instability - not only with a design error on the 2nd-source-board, but even cause instability with too-high signal load.

There's indeed one more port that has not been verified yet. Work on the ACA1240/1260 has been paused since march 2020 if I remember right. Got to get that back on the desk when I've paddled free of the administrative nightmare that our lawmakers gave us...

The ACA1240/1260 is back on our desk - after almost one year! Judging the performance is a different beast, though. Using Sysinfo is not an option, as it turns out that the Dhrystone-speed test is running completely in cache, so it scales almost linear with the CPU clock. Memory performance (which has a huge influence on everyday-performance) has almost no influence on te number: A 68060 will always get around 750 Dhrystones per MHz for Sysinfo.

So what we need are first-hand performance measurements in form of AIBB modules (AIBB launched in 68000 mode, that is). However, these should be accompanied with the exact CPU frequency and only when using Thor's 68040/68060 libraries, as these are the ones we'll ship with the product. There are AIBB modules on Aminet, but they were not made under the controlled conditions that we require for a fair comparison.

If anyone has access to a Warp1260, that would be great. This card is probably the only one that can compare to the ACA1240/1260 in terms of memory performance, where most (if not all) 1990s designs will suffer from slow SIMM speeds of the days.

Then there is Sysspeed, which also works with "modules". If possible, please also run tests with this tool, so we have more comparison material.

For memory performance, "Bustest" is still state of the art. Only Fastmem and Chipmem performance are interesting. Kickstart ROM performance will mostly be the same as RAM, as anyone will activate MapROM for best performance. Once again, for proper comparison we'll need the exact CPU frequency. Most 040/060 accelerators are probably overclocked these days, and the wildest frequencies may be used - sometimes 66.666MHz, sometimes 66.0MHz - and sometimes totally odd frequencies because "an oscillator was found somewhere". To properly normalize results, the exact frequency is key data.

Please post into this thread - thanks in advance!

Quote from Higgy

Is AIBB 6.7 ok?

Sure, but again, the exact test conditions should be recorded. In case of this AIBB version, it must be started in 68000 mode, otherwise it won't start on a 68060 processor.

Quote from Higgy

I run the benchmark and can save a 'module' file. And you would like this file rather than the summary displayed at the end?

Exactly - the file contains all results, much more detailed than the summary at the end.

Quote from Higgy

but I do have a modern 060 card

The only other modern 060 card I know of is the GB060, which was built largely with my help, including insight into the Apollo CPLD source codes. Georg Braun even figured out why the Apollo design fails in an OCS/ECS Amiga, and he followed pretty much every suggestion I made, including choice of the PLL clock multiplier, which gives you the choice of several standard frequencies - your description sounds very much like this.

Quote from Aardvark82

The decisive factor for my purchase would be compatibility with Mediator busboard. Has ACA1240/1260 been tested in such hardware configs?

A decicive factor for me to sell these cards would be a stable system. With Mediators mostly powered by outdated PSUs with improper grounding, I am not really interested in making the card "compatible", as it will only result in support hell. The quantity will be limited by CPUs anyway, so I prefer selling to customers with a less questionable setup. There are other 68060 projects - please consider those, as we won't be able to support Mediator setups - not only due to lack of hardware to test with, but also due to the more-than-likely outcome that all the work is in vain, and a stable system cannot be built with a Mediator and the funny case/PSU that comes with it.

Quote from VosCoBoss

Hoping for at least fast IDE onboard as I had mixed results with the IDE speeder on the ACA1233n/55.

One of the next things to test is a similar kind of IDE speeder on the ACA1240/1260. Since the ACA1233n-55 has rather limited logic, only two speeds were implemented (PIO mode 3 and 4), but the ACA1240/1260 can implement much finer granularity in terms of timing, so we can push/relax timing much better. Still, this may require a buffered adapter or high-quality CF card to work reliably. I've had great results with a buffered IDE adapter and a physical IDE drive (old IBM 16G drive), as that was MUCH faster than any CF card or DOM that I could find (>10MB/s).

On-board IDE is not planned. I don't want contraptions with a wobbly CF card adapter on such an expensive card.