Terrible Fire Accelerators

[alenppc]

Ok, so I _finally_ managed to get an A500... they are getting quite rare these days.

My cards booted up at first attempt! Right now I have a 25 MHz version with original firmware and 36 Mhz version on beta firmware (all rev 2s) all booting up and IDE seems to be working (not tested very thoroughly yet). Overall I didn't do a lot of testing cause it was late last night, so I am gonna test more this weekend, but so far so good.

Using the CPU FASTROM command speeds the system up a bit by copying the ROM into the blazing fast static ram.

This basically confirms that the card is incompatible with the A1000, or its boot roms most probably. Not sure if I should enter this as a bug into Github, what do you think plasmab?

[plasmab]

I've absolutely no idea TBH. I have an A1000. Need to test


Sent from my iPhone using Tapatalk

[michaelz]

Nearly done with two tf520's (waiting for xilinx programmer so I can do the last step) so looking for parts for the tf530. But what rev tf530 should I go for? I think I've got an 68030@25Mhz and an 68882 @25Mhz. Is Rev 3 recommended?

[plasmab]

Rev3 has better power supplies. That's really the only difference


Sent from my iPhone using Tapatalk

[Eriond]

A question to those with in-depth processor knowledge:
Reading the '040 manual @ page 11-3, it mentions a BCLK cycle time of min. 25ns. If I understand the bus operation properly, it takes two BCLK to complete a read or write operation (page 7-12), thus meaning that any memory with 50ns access time (or better) is enough to ensure zero wait states. Am I correct?

[whiteb]

Originally Posted by michaelz

Nearly done with two tf520's (waiting for xilinx programmer so I can do the last step) so looking for parts for the tf530. But what rev tf530 should I go for? I think I've got an 68030@25Mhz and an 68882 @25Mhz. Is Rev 3 recommended?

Well yes Rev 3 is recommended, but if you can wait a bit longer, he is working on a 4MB version (TF534), using different Ram (and different CPLD) if I have been following properly.

But if the 2MB does it for you, then yes the 2MB REV 3 will work just fine.

[plasmab]

Originally Posted by Eriond

A question to those with in-depth processor knowledge:
Reading the '040 manual @ page 11-3, it mentions a BCLK cycle time of min. 25ns. If I understand the bus operation properly, it takes two BCLK to complete a read or write operation (page 7-12), thus meaning that any memory with 50ns access time (or better) is enough to ensure zero wait states. Am I correct?

Unless its performing a cache fill in burst mode. So you'll get waits then.

[alenppc]

I did a few more tests on the 3 fully built boards and I have some issues, so I was wondering if anybody had any insight of what might be causing this.

1) Original firmware at 24 or 25 Mhz

When data cache is active, I get mild graphics corruption in some tools (AIBB for example, and a few (very few) 'sparkles' toward the bottom of the screen in Frontier. See first attached pic for an example. These problems disappear when data cache is turned off.

Also this does NOT happen with the BETA firmware, but because the beta firmware won't run at 25 MHz, I can't use it with 25 MHz CPUs of which I have a few.

2) Both ORIGINAL and BETA (36 mhz) firmwares

I can't write to the CF card without causing data corruption and invalidating the drive. I can reproduce this 100% of the time with 2 CF different cards, and all 3 TF530s I built.

The system will BOOT and read the CF card fine but as soon as I write something, it screws the filesystem. I though this was a MaxTransfer issue, but it isn't, it's already set correctly. I am using a CF adapter connected to an IDE cable (not plugged directly into the board).

3) Helpful tidbit: using CPU FASTROM command doubles the IDE drive throughput, getting it close to 2 mb/sec (reading, according to sysinfo)

[plasmab]

Originally Posted by alenppc

I did a few more tests on the 3 fully built boards and I have some issues, so I was wondering if anybody had any insight of what might be causing this.

1) Original firmware at 24 or 25 Mhz

When data cache is active, I get mild graphics corruption in some tools (AIBB for example, and a few (very few) 'sparkles' toward the bottom of the screen in Frontier. See first attached pic for an example. These problems disappear when data cache is turned off.

Also this does NOT happen with the BETA firmware, but because the beta firmware won't run at 25 MHz, I can't use it with 25 MHz CPUs of which I have a few.

2) Both ORIGINAL and BETA (36 mhz) firmwares

I can't write to the CF card without causing data corruption and invalidating the drive. I can reproduce this 100% of the time with 2 CF different cards, and all 3 TF530s I built.

The system will BOOT and read the CF card fine but as soon as I write something, it screws the filesystem. I though this was a MaxTransfer issue, but it isn't, it's already set correctly. I am using a CF adapter connected to an IDE cable (not plugged directly into the board).

3) Helpful tidbit: using CPU FASTROM command doubles the IDE drive throughput, getting it close to 2 mb/sec (reading, according to sysinfo)

1. Many bugs fixed in the original firmware

2. Very odd. I copied NEMAC IV and it worked fine. I wonder if its CF card specific. 36Mhz seems to be a good frequency to start with. Can you let me know which CF cards you're using.

3. I am only getting 1.5 mb/sec. Unsure why but that may be the difference.

I am using 0x1ffe for the max transfer.

The font corruption issue on AAIB .. havent seen that since the original firmware. Are you on Rev2 or Rev3

[whiteb]

Sounds like you have reproduced Steve's original sparkles problem, of which there was an issue in the firmware that he later fixed (It is in one of his later videos).

I know Steve is working on a script that spits out 25, 33 and 50 Mhz automatically. Its not a problem for me, as I have a 40Mhz chip, but thanks for the tip about CPU FASTROM, which CPU are you using because I was under the impression that FASTROM was only available to full Release chips (with MMU), by copying the ROM to Fast Ram.

Does not help me because I have an EC, (No MMU or FPU).

[EzdineG]

@whiteb

What revision TF530 are you using? Is it completely stable at 40 MHz?

I have now built two TF530 Rev 3 units, and get random memory errors (never the same area twice) on both.

My 68030s are both 33mhz, and my only "workable" oscillator is 40mhz, so I don't have an ideal test scenario yet. I don't know if what I'm seeing is firmware/CPU OC/2x bad soldered boards so it would be good to hear someone's got a working Rev 3.

Thanks

[plasmab]

Originally Posted by EzdineG

@whiteb

What revision TF530 are you using? Is it completely stable at 40 MHz?

I have now built two TF530 Rev 3 units, and get random memory errors (never the same area twice) on both.

My 68030s are both 33mhz, and my only "workable" oscillator is 40mhz, so I don't have an ideal test scenario yet. I don't know if what I'm seeing is firmware/CPU OC/2x bad soldered boards so it would be good to hear someone's got a working Rev 3.

Thanks

I found that the 33Mhz CPUs are borderline at 40Mhz if the room is warm.

[whiteb]

Originally Posted by EzdineG

@whiteb

What revision TF530 are you using? Is it completely stable at 40 MHz?

To be Honest I have not built one yet, I was holding out for the TF534, but I am thinking of doing a Rev 3 run anyway, I have 3x A500's and an A1000 to play with and also I want to get an Indivision ECS.

[edit] if I have a Rev 3, then a Rev 4 as well I can just move the CPU between them (Unless I get another stupidly cheap 68030 chip again that was not from Chinakong).

[plasmab]

Originally Posted by whiteb

To be Honest I have not built one yet, I was holding out for the TF534, but I am thinking of doing a Rev 3 run anyway, I have 3x A500's and an A1000 to play with and also I want to get an Indivision ECS.

[edit] if I have a Rev 3, then a Rev 4 as well I can just move the CPU between them (Unless I get another stupidly cheap 68030 chip again that was not from Chinakong).

Not sure that it matters but i decided to reset the revisions on the TF534... there are enough changes that i'm likely to make a mistake.

[EzdineG]

Originally Posted by plasmab

I found that the 33Mhz CPUs are borderline at 40Mhz if the room is warm.

I've a stack of five, and one of the two units is socketed. Unfortunately they all behave the same. I'm starting to doubt my ability to solder without magnification. If a third goes together, might have to pick one up first.

Yea yea, "don't try this without a camera." Fuck it man, YOLO.

[alenppc]

Originally Posted by plasmab

The font corruption issue on AAIB .. havent seen that since the original firmware. Are you on Rev2 or Rev3

Yes it is the original firmware on rev 2 @25 MHz -- ok so basically it's just a minor bug, nothing I did wrong there.

So the only real issue that remains is the weird IDE behaviour - I can't understand what causes it.

@whiteb I have a whole bunch of full 68030s rated at 25 MHz - found them at $10 each

[EzdineG]

Originally Posted by alenppc

So the only real issue that remains is the weird IDE behaviour - I can't understand what causes it.

Did you follow the earlier exchange about the resistance value for R1? What did you end up with here?

[whiteb]

Originally Posted by alenppc

Yes it is the original firmware on rev 2 @25 MHz -- ok so basically it's just a minor bug, nothing I did wrong there.

@whiteb I have a whole bunch of full 68030s rated at 25 MHz - found them at $10 each

FFFFFFFFFFF, where from ? Not Ebay.

[whiteb]

Originally Posted by plasmab

Not sure that it matters but i decided to reset the revisions on the TF534... there are enough changes that i'm likely to make a mistake.

Oh well, Well so the TF530 will stay at REV 3, and the TF534 back to Rev 1+

[alenppc]

Originally Posted by EzdineG

Did you follow the earlier exchange about the resistance value for R1? What did you end up with here?

Tried both 47k and 10k, no difference. Just dug up one of the antique hard drives, let's see if it's any different.

@plasmab I was using transcend 512MB and Sandisk 4 GB cards