exxos's Atari, Amiga & retro forum

I managed to get my TF330 to work well at 66MHz.

Karl wrote: 03 Jan 2022 10:44 I managed to get my TF330 to work well at 66MHz.

I've never seen an 030 up at 66mhz.. so a 132Mhz crystal and some cooling?

terriblefire wrote: 03 Jan 2022 10:47

Karl wrote: 03 Jan 2022 10:44 I managed to get my TF330 to work well at 66MHz.

I've never seen an 030 up at 66mhz.. so a 132Mhz crystal and some cooling?

Yes. Cooling is just passive. 5V is just 4.9V at TF330 board.
The 030 can do 75MHz with a FW I found, but I get disk errors. With this FW I get RAM issues at 75MHz.
I added one cycle delay in the ata.v to get the disk to work. I also use standard power mode... maybe it should always be that and I messed something up.

Karl wrote: 03 Jan 2022 10:54

terriblefire wrote: 03 Jan 2022 10:47

I've never seen an 030 up at 66mhz.. so a 132Mhz crystal and some cooling?

Yes. Cooling is just passive. 5V is just 4.9V at TF330 board.
The 030 can do 75MHz with a FW I found, but I get disk errors. With this FW I get RAM issues at 75MHz.
I added one cycle delay in the ata.v to get the disk to work. I also use standard power mode... maybe it should always be that and I messed something up.

You'll need to pad out ram timings or it will fail.

Also are you compiling inside ISE? if so please stop an use the make files instead. ISE screws stuff up lots.

terriblefire wrote: 03 Jan 2022 11:27

Karl wrote: 03 Jan 2022 10:54

Yes. Cooling is just passive. 5V is just 4.9V at TF330 board.
The 030 can do 75MHz with a FW I found, but I get disk errors. With this FW I get RAM issues at 75MHz.
I added one cycle delay in the ata.v to get the disk to work. I also use standard power mode... maybe it should always be that and I messed something up.

You'll need to pad out ram timings or it will fail.

Also are you compiling inside ISE? if so please stop an use the make files instead. ISE screws stuff up lots.

Yes, I used ISE for compiling a few different versions. I couldn't successfully compile using the make files, so I tested ISE.

My goal was to lower disk performance to what it was in early versions of the FW, on a recent FW. Because those works at 120 MHz - 133 MHz for me.

Yes, ISE screwed it all up. I made 5 different versions and made notes of the differences. Like standard power, low power, delay added, RAM timings loosened...
When I now try to replicate the version that works, it doesn't work.
I actually have no idea how I got it to work. The version I thought I used I had lowered disk performance down to around 3.2MB/s at 50MHz. That would keep me below 5MB/s (where I start to see problems) when overclocked.
The strange thing is that the working version gives around 5.5MB/s. That's too high for a 30% clock increase from 3.2MB/s, I think.

I'm happy I've got a working .jed though. I might spend some more time on this in the future. I will have to properly set up the toolchain first...

What was up with the make files. I use Linux + ISE 10.1. If thats installed correct then it works first time. If you have a different version just edit the Makefile.cpld to point to the correct version.

On windows i actually have 10.1 tarred up from a Linux install and unzipped inside a WSL Ubuntu container.

terriblefire wrote: 03 Jan 2022 17:25 What was up with the make files. I use Linux + ISE 10.1. If thats installed correct then it works first time. If you have a different version just edit the Makefile.cpld to point to the correct version.

On windows i actually have 10.1 tarred up from a Linux install and unzipped inside a WSL Ubuntu container.

I will try that version.

I must first get better SDRAM if I'm going to try to reach higher frequencies.
What SDRAM would you recommend? I currently use ISSI speed grade 6I.

I'm already close to 5V when using a fast FW (early versions can do 66MHz at 0.15V less voltage).
To increase the voltage separately for the CPU will be tricky. I don't want to overvolt my CD32.

Higher voltage will introduce cooling issues.
The 030 is not easy to cool. I think the die has better thermal connection to the bottom side of the package, than the top. That gold lid is not like an IHS on modern CPUs.
I've put a little copper block in the center of the CPU socket that is connected to the PCB with a thermal pad.

Then there will probably be more IDE issues...

I must thank you for making all of this possible! I really enjoy tinkering with my TF330! :D

50MHz vs. 66MHz

You dont need better SDRAM. The SDRAM you have would work at 132Mhz no problems it just needs the timings padded.

SDRAM clocks arent the issue here. Its because the SDRAM is tuned to be at max performance on the timings at 50Mhz... Better SDRAM isnt any better here. You need to RTFM the timings your ram can do and alter sdram_defines.vinc

terriblefire wrote: 04 Jan 2022 10:33 You dont need better SDRAM. The SDRAM you have would work at 132Mhz no problems it just needs the timings padded.

SDRAM clocks arent the issue here. Its because the SDRAM is tuned to be at max performance on the timings at 50Mhz... Better SDRAM isnt any better here. You need to RTFM the timings your ram can do and alter sdram_defines.vinc

I already adjusted the timings for my current 66MHz board, but which SDRAM make/version is the 'Samsung B-die', so to speak?
I want to use the best components available If I decide to try higher than 66MHz.