Misinformation surrounding aca1233n

General hardware issues and troubleshooting etc

Moderators: terriblefire, Terriblefire Moderator

Steve
Posts: 2570
Joined: Fri Sep 15, 2017 11:49 am

Misinformation surrounding aca1233n

Post by Steve »

Hi all
I hope this isn't too much of a contravertial topic (I am blissfully ignorant of whether this is an approproate subject or not) but I'm hoping to clear up details surrounding modification of the aca1233n. Specifically replacing it's CPU of a different clock speed. (Of course I am aware it voids any warranty but I believe it is in the owners right to do whatever with hardware he owns, as long as it isn't copying it for resale etc)

On Amibay a person was selling a modified board with 50mhz CPU and fpu. You see, there are two Xtals, one regular one for the FPU (if added) and a SMD Xtal. According to the Amibay thread, this guy swapped the 80mhz Xtal for a 100mhz which clocked his 40mhz board to 50mhz.
Thread:
http://www.amibay.com/showthread.php?99 ... t=aca1233n
Quotes:
"I replaced the SMD oscillator with a 100MHz one.. "

"I have DONE this to more than one board (In the past) and all work perfectly after CPU change (Cheap plastic CPU and NO MMU yuck).. But to do it, you need to use a solder pot to get the old CPU out in one piece (Solder temp setting must not be too high or the board will get burned)..... In the end a very nice board with a better processor fitted (In a socket now), AND it's a 50MHz processor with MMU (Not the 40MHz one without MMU)..."
But then on the icomp.de forum a customer asks if he is able to do this mod. He is told no as the Xtal is 80mhz for both the 26 and 40 MHz varient (he is initially unaware that his board is 40mhz but gets corrected by his board ID info) and the clock speeds are set via the CPLD. Which would require an expensive tuning by icomp.

Quotes:
Question from customer on icomp.de forum:
I have ACA1233N 26mhz and I want to put faster CPU 40mhz or 50mhz. If I takeout this 68030 26mhz and I solder new 68030 40mhz or 50mhz, Ofcourse I swap crystal as well, will it work?
Answer provided:
"No need to swap the Xtal; both the 26MHz and the 40MHz card use an 80MHz crystal. The CPU swap won't get you far; it will run at the exact same speed.""You can buy an updated CPLD for the difference in price of the two cards. That means: You ship the card to us with the new CPU on, we do the re-programming of the CPLD to 40MHz for 80,- EUR incl. VAT"
https://forum.icomp.de/index.php?thread ... -or-50mhz/

So from processing this information it seems:
If the 26 and 40 use the same Xtal and the clock divider is set in the CPLD, then the 50mhz varient (since the information was ommited by the icomp rep) probably does use 100mhz, changing the divider speed. But I am guessing that on a stock 50mhz board the CPLD would be updated to have its divider values updated for reference purposes in the ACATool app (where different divisions can be selected, for under clocking purposes only)

So I surmise that even though he was told no on the official forums, changing the Xtal to 100mhz would overclock any processor on the board, regardless of CPLD. This is assuming that britlord on the Amibay forums is telling the truth.

What do others think? Am I allowed to ask these questions, I am surprised that this isn't being asked anywhere on the internet in any Amiga forums, almost like big brother is watching...
Steve
Posts: 2570
Joined: Fri Sep 15, 2017 11:49 am

Re: Misinformation surrounding aca1233n

Post by Steve »

With 35 views and zero replies I'm starting to think I'm secretly being put on some hit list.... Ok time to take the meds... Please don't hurt me Amigans. :shock:
User avatar
PaulJ
Posts: 1568
Joined: Sun Apr 08, 2018 1:14 am
Location: USA

Re: Misinformation surrounding aca1233n

Post by PaulJ »

Keep posting Steve.. might of responded but I am not familiar with the machine although it it has a moto processor I should be. Maybe a picture is order. :)
User avatar
JezC
Posts: 2083
Joined: Mon Aug 28, 2017 11:44 pm

Re: Misinformation surrounding aca1233n

Post by JezC »

Also interested in this kind of topic but don't know enough to comment !
Steve
Posts: 2570
Joined: Fri Sep 15, 2017 11:49 am

Re: Misinformation surrounding aca1233n

Post by Steve »

PaulJ wrote: Thu Sep 17, 2020 12:04 am Keep posting Steve.. might of responded but I am not familiar with the machine although it it has a moto processor I should be. Maybe a picture is order. :)
Ah sorry Paul, it's not actually a machine, it's an accelerator card for the Amiga 1200. I just don't fully understand why icomp.de don't want people to know how to do this. I mean - they say that they got annoyed supporting people who were overclocking their previous board. Well surely the answer is to just say no support if you modify it? Either way, I personally have one of these boards and wouldn't mind working it out (had it since before Terriblefire joined the forums here)
terriblefire
Moderator Team
Moderator Team
Posts: 5368
Joined: Mon Aug 28, 2017 10:56 pm
Location: Glasgow, UK

Re: Misinformation surrounding aca1233n

Post by terriblefire »

i just simply dont know about this architecture for the card to comment.
———
"It is not necessarily a supply voltage at no load, but the amount of current it can provide when touched that
indicates how much hurting you shall receive."
User avatar
GadgetUK164
Posts: 430
Joined: Fri Jan 04, 2019 2:26 pm

Re: Misinformation surrounding aca1233n

Post by GadgetUK164 »

My take on this - division is done in the CPLD by the sound of things. By changing the XTAL you are automatically increasing the speed of any clocks (not just the CPU clock) provided by that CPLD. Therefore you are getting the expected result if you calculate the main crystal speed from what was already on there, and what CPU clock speed it provided, you can work it out % wise how much increase you will get by putting a faster XTAL on there.

The only problem I see with this kind of mod (and I did a similar mod on my ACA1220) is the RAM needs to be fast enough, and the CPLD too. The tuning is going to refer to other domains probably - ie. when increasing the CPU clock by brute forcing it at the top level with the XTAL is going to mean that any clock divided from that master clock are perhaps faster than they should be. If it works, great! I would be checking temperature of the CPLD and RAM though. Things like wait states on some RAM might start to be required too.

I don't know exactly how these boards work, but I would guess RAM timing is affected at the very least.
My YouTube Channel - www.youtube.com/GadgetUK164
juz400
Posts: 36
Joined: Wed Sep 11, 2019 8:18 pm

Re: Misinformation surrounding aca1233n

Post by juz400 »

Changing the clock crystal has generally worked since they were on motherboards, I got some really funny results swapping out the crystals on 286 motherboards in mid 90`s. What else the clock is tied into is quite often the failure point, HDD controllers, Amiga chipset, you get the idea.

Was a regular overclocker in the past but seeing new users to overclocking forums who had only just found out about the `overclocking revelation`
spending £650+ on a graphics card, frying it in 3 days trying to beat so and so`s records who new what they were doing and then expecting EVGA or whoever to replace the card, then doing it again with the replacement. /Facepalm

I would say change it if you want but DO expect failure and DONT expect any warranty!
Steve
Posts: 2570
Joined: Fri Sep 15, 2017 11:49 am

Re: Misinformation surrounding aca1233n

Post by Steve »

GadgetUK164 wrote: Thu Sep 17, 2020 12:51 pm My take on this - division is done in the CPLD by the sound of things. By changing the XTAL you are automatically increasing the speed of any clocks (not just the CPU clock) provided by that CPLD. Therefore you are getting the expected result if you calculate the main crystal speed from what was already on there, and what CPU clock speed it provided, you can work it out % wise how much increase you will get by putting a faster XTAL on there.

The only problem I see with this kind of mod (and I did a similar mod on my ACA1220) is the RAM needs to be fast enough, and the CPLD too. The tuning is going to refer to other domains probably - ie. when increasing the CPU clock by brute forcing it at the top level with the XTAL is going to mean that any clock divided from that master clock are perhaps faster than they should be. If it works, great! I would be checking temperature of the CPLD and RAM though. Things like wait states on some RAM might start to be required too.

I don't know exactly how these boards work, but I would guess RAM timing is affected at the very least.
Yeah I agree with your analysis. Do you think that they would go to all the trouble of sourcing slower RAM just for the boards that are going to receive slower CPUs though? In my mind they probably produced all the boards (no matter what CPU they were going to receive) and then just programmed the CPLD accordingly. Since they would have needed to source the CPUs and test them later. I will check out the specs of the ram on my unit and check what speed it supports... also as far as I know, it is SD RAM, which generally is minimum 100mhz-133mhz anyway. Also the CPLD should be the same part surely too? Just with different divider details (I assume purely for display, not function).
User avatar
GadgetUK164
Posts: 430
Joined: Fri Jan 04, 2019 2:26 pm

Re: Misinformation surrounding aca1233n

Post by GadgetUK164 »

Steve wrote: Thu Sep 17, 2020 12:54 pm
GadgetUK164 wrote: Thu Sep 17, 2020 12:51 pm My take on this - division is done in the CPLD by the sound of things. By changing the XTAL you are automatically increasing the speed of any clocks (not just the CPU clock) provided by that CPLD. Therefore you are getting the expected result if you calculate the main crystal speed from what was already on there, and what CPU clock speed it provided, you can work it out % wise how much increase you will get by putting a faster XTAL on there.

The only problem I see with this kind of mod (and I did a similar mod on my ACA1220) is the RAM needs to be fast enough, and the CPLD too. The tuning is going to refer to other domains probably - ie. when increasing the CPU clock by brute forcing it at the top level with the XTAL is going to mean that any clock divided from that master clock are perhaps faster than they should be. If it works, great! I would be checking temperature of the CPLD and RAM though. Things like wait states on some RAM might start to be required too.

I don't know exactly how these boards work, but I would guess RAM timing is affected at the very least.
Yeah I agree with your analysis. Do you think that they would go to all the trouble of sourcing slower RAM just for the boards that are going to receive slower CPUs though? In my mind they probably produced all the boards (no matter what CPU they were going to receive) and then just programmed the CPLD accordingly. Since they would have needed to source the CPUs and test them later. I will check out the specs of the ram on my unit and check what speed it supports... also as far as I know, it is SD RAM, which generally is minimum 100mhz-133mhz anyway. Also the CPLD should be the same part surely too? Just with different divider details (I assume purely for display, not function).
If there are any photos of a slower board vs a faster board I guess we can check! One thing I do know (and I've not tested it yet) - I think if I replaced the RAM on my ACA1220, I could get to 40Mhz or beyond. The clock division thing is the problem on the ACA1220 as the RAM is not fast enough.

My guess on thee ACA1233 is they went with faster RAM, or the divider is different.

Yes, I agree its probably the same CPLD and the same RAM - its probably timing related to that RAM that gets altered, and perhaps its clock too. However, if it "just works" then maybe they don't do any changes at all, perhaps they just dont want people modding them. It could also be the case that the RAM gets pushed up to its maximum performance and could be borderline depending on how much of an "overclock" is going on.

If I ever get an 030 or 060 board for my 1200 I will have a go at replacing the RAM on my ACA1220 with something that might work with 40 or 50Mhz. I have a 33Mhz 020 on there, but I bet if the RAM supported more than 100Mhz I might get 40 or 50Mhz out of that 020. Assuming timing hasn't been messed up for other clock domains etc.
My YouTube Channel - www.youtube.com/GadgetUK164
Post Reply

Return to “AMIGA HARDWARE HELP”