Building the TerribleFire TF530 accelerator, part 6: Faster! Hotter!

With a little more experience and new information gained, there was something left to do...

Disclaimer: This is not an instruction. Use at your own risk. No responsibility taken for whatever you do. Safety first. Kids, dont try this at home. 

32MHz might not be too shabby, but then it's a weird number, and the TF530 is capable of more. It's been quite some time since I built my TF530. In the meantime Mr. Leary has released new firmware versions, improving the accelerator's compatibility with higher clock rates (among other things, probably). So the day had come to try a little tweaking.

First of all I had to test the latest firmware release. I fired up my laptop running Xilinx's CPLD programming software, connected the USB-cable to one, and TF530 5V power input to another USB port on the laptop. Surprise: The Amiga boots! It pulls enough power from the laptop's USB port to boot to Workbench! And of course the laptop (running Windows) complains about a USB device pulling way too much power! Ok, I'd better disconnect the power, remove the TF530 from the Amiga, reconnect, and try again...

Faster!

Without any changes to the TF530 I uploaded the latest software to the CPLDs. Hm, not bad! Worked flawlessly, and SysInfo shows a little speed increase of about... 3%! Not precisely earth-shattering, but a welcome improvement.

From ~5800 Dhrystones to ~6000 just by firmware update!

I had never thoroughly tested the memory chips, and the new firmware needed something to do, so I started Microbotic's excellent MBRTest-2 memory testing program, and ran a couple of tests. To my surprise there were no errors - my handling and soldering of the memory chips was fine from day zero!

No RAM errors, very good!

So far, so - very - good. I had seen some accelerator configurations where the 68030 had been overclocked by about 25%, and the FPU on my TF530 is a 40Mhz model anyway, so, yeah, why not try 40Mhz on the 33Mhz CPU?

33Mhz CPU and 40MHZ FPU - will they run fine at 40MHz?

Replacing the tiny SMD type 32Mhz oscillator with a 40Mhz one requires some fiddling if you don't have a heat-gun, but these oscillators seem to be quite tolerant to heat, and I managed to do it using a soldering-iron without any visible damage. After some cleaning it was time to switch the Amiga on, and see how she does...

~7500 Dhrystones!

Hotter!

Once again, to my surprise, this worked flawlessly! Wow, running at 40Mhz we get about 7500 Dhrystones in SysInfo! The 68030 is getting a bit hot - I'll be adding at least a small heatsink just to avoid unnecessary stress on the material.

TerribleFire 530 with CF card adaptor inside the "Amiga 530"

As a final touch to my "Amiga 530" I wanted to delete AmigaOS3.9 from the CF card, and reinstall OS3.1. With some little adjustments OS3.9 ran quite well on the machine, but it's still comparatively resource-hungry, and in some parts unnecessarily bloated, thus slow. Installation of OS3.1 - no surprise here - was quick and worked flawlessly, too. Another reset, and...

...the Amiga now boots to Workbench in about 9 seconds!

This is really good. I could try to get a 50Mhz CPU and install a 50Mhz oscillator - the 40Mhz FPU should be able to take that (the 33MHz CPU probably not so much...) - but for now I'm really perfectly satisfied with the performance. Firmware update, plus faster clock, plus operating system downgrade really make a difference. I'm assuming an added heat-sink will make the setup safe and stable in the long term.

Now will THIS be the last part of the series of articles about "Building the TerribleFire TF530 accelerator"? Who knows. Probably not! There's still so much to experiment with - we haven't even touched the SPI port yet!

Once again a huge Thank You! to Mr. Stephen Leary for developing and releasing this awesome accelerator!

Overview and back catalogue of the series:
http://amigaalive.blogspot.de/p/hardware-building-terribefire-tf530.html

Building the TerribleFire TF530 accelerator board, part 5: Testing. More testing. And... happiness.

I'm learning new things about the TF530 almost daily...

Documentation is a problem when trying to build these accelerators: there's no place where you can quickly look up how to do something. Instead you have to dig through Stephen Leary's hour-long videos on YouTube, or a single endless thread on English Amiga Board, or ask in forums. Which is not a bad thing, but it's very inconvenient if you want to make progress. That's why I created some sort of "user manual" for the TerribleFire boards: https://amihelper.blogspot.de/p/terriblefire-accelerators.html

Anyway, what's happened since part 4 of this series?

Full setup with CF-adaptor and -card

Things were actually looking pretty good. I've been using a 24Mhz oscillator to build the board, and the original firmware (from a "rev. 2 final" GitHub checkout). But I wanted to rev up the board a little... Other people are using clockrates of 36, 40, or 50 Mhz. And then there are a couple of new revisions of the firmware. And in one or the other way these have to match. I've tried a 40MHz oscillator, and two other (apart from the original) versions of the firmware. Then there's the IDE controller. Simply plugging in an IDE cable - without any device attached! - changes the A500's behaviour in combination with the 40MHz oscillator. Exchanging the 1.3 Kickstart ROM for a 3.1 one made a difference once again. Lots of testing, and lots of glitches, and more testing, and more glitches: mousepointer uncontrollable, random crashes, boot picture gone, ...

It seems that the 33Mhz 68030 doesn't like overclocking very much - but with a 32MHz oscillator (33Mhz oscillators seem to be rare), it now runs stable! I expect a few more issues - like checking RAM soldering points - but for now I declare it "working fine". It boots quickly from the 512MB CompactFlash card attached, and with the additional 2MB of RAM on the TF530 and the new 3.1 Kickstart ROM the A500 is now an up-to-date, very usable Amiga!

Yesss! I am __very__ happy with it. :-)

SysInfo speed test with the 32MHz oscillator installed

Nice to have: an LED on the CF-card-adaptor

An "in progress" conclusion

From a personal perspective, I can say that I've learned a lot. I mean: A LOT. Stephen Leary's videos give a huge amount of information, and by building the accelerator I've expanded my set of tools, improved my soldering skills, and gained some knowledge about the parts used. Overall, for me it's a huge step forward in understanding the inner workings of Amiga computers, the TF530, and electronics in general.

And of course the Amiga 500 got a huge update, it's a completely different machine now. For a very reasonable (money-wise) price.

So the whole process is a massive success for me. :-)

No outside signs of the changes on the inside... :-)

Still, there are a few things to keep in mind about the TF530 accelerator and the build process: First of all it takes an insane amount of time to build, if you don't know what you're doing beforehand. I certainly didn't. There was a lot of research involved, on all levels. Secondly, the TF530 is clearly not an industrial-grade product: it's an "in progress" project, a moving target. You have several options when it comes to CPU and oscillator, and these have some interdependence with the firmware releases, which are still going on. The physical layout of the TF530 isn't perfect, you (may) need a relocator board to fit it into your Amiga, you (may) need to use specific connectors, and still it fits only barely between the Amigas other components and casing. For example, adding a Kickstart-switch may turn out to be quite difficult now with the TF530 in place.

But all these quirks don't change a simple fact:

It's effin' great! 

Awesome!

Amazing!

Magnifique!

Thank you, Mr. Stephen Leary!

Thank you very, very much for sharing,
and lighting up the TerribleFire!

So far, so good. I currently consider this to be the last entry in my little series about building the TerribleFire 530 accelerator board. But who knows. And there's so much more to do... maybe a TF328?

Overview and back catalogue of the series:
http://amigaalive.blogspot.de/p/hardware-building-terribefire-tf530.html

Building the TerribleFire TF530 accelerator board, part 3: Parts inspection, and soldering

Let's see what we've got. And how we're gonna put it together. This one had a little surprise for me.

Disclaimer: This is not an instruction. Use at your own risk. No responsibility taken for whatever you do. Safety first. Kids, dont try this at home. 

Parts and BOM, and the shiny blue PCB.

It looks like I haven't missed much so far. Everything required to populate the board is there (except for the CPU socket connector, for which some substitute will be built), and I'm still able to identify the individual components! :-)

Parts and tools.

So, with Stephen Leary's instructional video (see below) on my computer's screen, I started soldering. May I note that not only does his video provide helpful and important information and hints, but also is a pleasant accompaniment when working on the TF530.

Soldering the first SMD chip on to the board quickly resulted in the chip being roasted, and the PCB damaged. Lol. Surprise. It turned out to be waaaaaay more difficult than I had expected.

Working with tiny SMD components poses some unexpected problems. For example if you're used to soldering hole-though parts you will most likely use way too much solder on your first attempts. Well, I did. Then, when you got used to using really very little solder, you find out that you cannot easily suck away the solder if you've made a mistake, because the amount of solder is too little for the desoldering pump to work on. Then you may learn the painful way that tiny PCB chip pins are reeaaally weak, you have to treat them really gently. And finally, when checking results of your soldering, you may find out that your grandma's magnifying glass does what it's supposed to do - to magnify - but it's scratched or unclear, and really creates more confusion than it does help.

First chip in place. Dirty, but undamaged!

But after some practicing with the damaged board and chip, things looked better. I learned to use the right amount of solder, the right amount of soldering flux, and to move the soldering iron at the right speed, and started working on a new board - with far better results.

More components.

After some hours of focused soldering and (visual) checking it looks as if my first TF530 might soon be ready for testing.

TF530r2 fully populated!

She's currently no beauty and needs some serious cleaning, but she makes her owner very proud. :-)

If you wanna try to build one, here's Stephen Leary's first "how to" video (we're still waiting for the second):

https://www.youtube.com/watch?v=xG_5oDeLdGE

Next time we need to find out if she's alive - watch out for part 4 of our series "Building the TerribleFire TF530 accelerator board"!

Overview and back catalogue of the series:
http://amigaalive.blogspot.de/p/hardware-building-terribefire-tf530.html