Before I Break My RISC OS Machine..

Howdy Howdy everyone. I just fired up my RISC OS machine (PandaBoard ES) that has been sitting for quite some time. I am currently writing this post from it! I attached an SSD via USB → SATA adaptor and everyone here helped me configure RISC OS to boot from the SSD. I’ve moved the files from the SD card over to the SSD, it’s booting up just fine including !Boot. I should be able to remove all the files from the SD card correct? The PandaBoard just needs the files on the FAT partition to initialize the board right? If I am understanding it correctly anyway..

I should be able to remove all the files from the SD card correct?

The only disadvantage is that you would thus destroy a known working backup. Also you would (if you deleted the whole of the filecore partition) leave a very strangely formatted FAT SD card with most of the card protected from use (by the filecore partition in the partition table, which overlaps with the FAT partition) but not used.

OK. I will make a backup of the files to a USB drive. What I was getting errors from the internal SD card probably looking for stuff in !Boot on the wrong drive. The SSD is much faster and has double the space of the SD card so it’s a welcome addition. What could go wrong? :)Thanks for the response. I am hoping to use this machine more frequently.

I am surprised that a Pandaboard is still a viable machine. My Pandaboard overheated and died despite fairly infrequent use. RISC OS will gradually fade away if people keep using obsolete machines based on development boards.

The Raspberry Pi offers a cheap solution.
The ARMX6 kit (or mini.m) a higher performance but essentially cheap (not much different from Pandaboard) solution that is still being actively developed and sold.
The Titanium a high performance machine purpose-built for RISC OS but more expensive.
The ARMbook a portable solution.

I am assuming that the Iyonix has died out by now and see no point in supporting it beyond RISC OS 5.16 when things stopped working because of an unwise and ill thought through C library change. The Beagleboard died off a while ago. IGEPv5 still hangs on but I haven’t seen an update for years. Virtual RISC PC Adjust still provides a very useful, fast and convenient emulator.

I have Raspberry Pi 1,2,3 and a 2x BBxM’s and 2 or 3 Pandaboard ES machines. I started out on the RPi 1 and discovered RISC OS that way. Eventually I want a Titanium. I’m on 5.22 right now. Everything is working, but would like a better browser and some more processing power for it. As long as I don’t do much heavy web use this Pandaboard is great, especially with the SSD. Can’t imagine the Titanium with an SSD!

The RISC OS machine I use the most at the moment is an AI Touchbook. Quite fun. I also have a bunch of Panda Boards and Raspberry pi’s, but I use them less.

In order to bring RISC OS to new markets, it is important to show that it is possible to continue to support a platform for a long time.

I am assuming that the Iyonix has died out by now and see no point in supporting it beyond RISC OS 5.16 when things stopped working because of an unwise and ill thought through C library change.

So I better take this 5.27 (14-May-2020) softload off my Iyonix then and get rid of the flashed 5.24 as well as it clearly doesn’t work :-)

I use the older machines for testing and I still think that support is required for them and why else would ROOL also make 5.27 available for RPCEmu and real RiscPC’s?

I am surprised that a Pandaboard is still a viable machine.

The CPU on a PandaBoard ES is faster than on an i.MX6 (at least in theory because of the higher clock speed, I think they are very very close in benchmarks, not sure why).

My Pandaboard overheated and died despite fairly infrequent use.

IIRC, you overclocked it to 1.5 GHz before it died. Both of mine work very reliable, and during the time when they were the fastest native RISC OS machine money could buy, got used quite heavily.

RISC OS will gradually fade away if people keep using obsolete machines based on development boards.

PandaBoard ES is far from obsolete. Even the original ARMini/BeagleBoard-xM is fine for most RISC OS stuff and also far from obsolete. However, there are still a lot of Risc PCs in operation, which…well, if they work, keep on using them.

The Raspberry Pi offers a cheap solution.

However, if you need ARMv7 compatibility, it is also quite a bit slower than the PandaBoard ES. It is difficult to obtain the correct RPi 2 hardware revision, and the RPi1/Zero is significantly slower than the PandaBoard ES.

The ARMX6 kit (or mini.m) a higher performance but essentially cheap (not much different from Pandaboard) solution that is still being actively developed and sold.

If you already have a PandaBoard (or ARMiniX), it is very cheap. 0 EUR, which converts to 0 UKP. Hard to beat. The i.MX6 platform is of course faster for the one S-ATA I/O device and has faster networking (faster than Gigabit-over-USB2.0 though? Never tried that…), but CPU is slower (although close).

The Titanium a high performance machine purpose-built for RISC OS but more expensive.

And it suffers from its video capabilities being non-4K, so you need special solutions like side-by-side-dualhead, which is incompatible with my KVM-based setup and many monitors.

I am assuming that the Iyonix has died out by now

Two of mine are still going strong.

and see no point in supporting it beyond RISC OS 5.16 when things stopped working because of an unwise and ill thought through C library change.

I don’t have the slightest idea what you are talking about. The only real problem on the IYONIX was the not-entirely-correctly-working RTC some years ago on a certain year change, which needed a true OS upgrade to fix it properly. Apart from that, softloading new RISC OS versions is not a problem at all.

Compared to e.g. the various RPis, it has vastly faster I/O, both disc and networking. It has higher video resolution than a Titanium. And its CPU is ARMv5 compatible, which might be valuable (see Alignment Exception On/Off “solution” for other machines). Depending on your use case, it just does not make sense to replace it.

The Beagleboard died off a while ago.

Mine are still working fine, including the BIK (an ARMini in disguise). As long as your screen can handle 24Hz@FHD, they are still fine machines.

IGEPv5 still hangs on but I haven’t seen an update for years.

New ROMs are built daily. Mine suffers from being a true development board not even having a case…

My overall point is: many of the older machines still work fine and there is little reason to replace them if you don’t need the extra power (may it be CPU, disc I/O, video capabilities or network speed). And getting something that improves significantly on the PandaBoard ES in every respect – CPU, video, disc I/O, networking – is impossible.

IIRC, you overclocked it to 1.5 GHz before it died.

It was switching between 1200MHz and 1500MHz as workload changed. It was not running at 1500MHz continuously.

It seems a pity that we do not have a single platform that is better, in every department, than its competitors. This (apart from Titanium) is the result of us having to use boards developed by third parties for other purposes than RISC OS.

It seems a pity that we do not have a single platform that is better, in every department, than its competitors. This (apart from Titanium) is the result of us having to use boards developed by third parties for other purposes than RISC OS.

The advantage is that we get economy of scale,

why else would ROOL also make 5.27 available for RPCEmu and real RiscPC’s?

Easy answer – we have no emulation solution that is something more recent than a RiscPC.

I think they are very very close in benchmarks, not sure why

Different core technology. Of the two Pi2 versions, the original ARMv7 version is a Cortex-A7 and the later ARMv8 version is a Cortex-A53. Both are clocked at 900MHz. The A53 is quoted (on Linux that can use all four cores) as being about 20% faster.
At the same clock speed.

So differences in clock speed can be offset by differences in core technology. A machine that’s supposed to be slower might be able to beat a machine that’s supposed to be faster.

I’ve not run any benchmarks, but I perceive my Pi2 (A7 core @ 900MHz) to be faster in operation than the Beagle xM (A8 core @ 1GHz).

Even the original ARMini/BeagleBoard-xM is fine for most RISC OS stuff

This. I find that in my use of machines, I need quick bursts of speed (converting an image, rendering a web page, compiling something), which the rest – like 99.9% of the time – the machine is doing absolutely nothing. A BBC Micro could keep up with a person typing stuff, it’s hardly a struggle for a RISC OS machine of any generation. ;-)

However, there are still a lot of Risc PCs in operation, which…well, if they work, keep on using them.

This I would disagree with. Unless it is required for some specific 26 bit software that won’t work under Aemulor, a basic Pi solution (even the Pi1) will run rings around any RiscPC (and greatly improved video capabilities so you can use a big monitor with all the colours).
Yes, it costs. You might be able to scavange bits or recycle from an old PC. Or maybe not. Either way, I’ve calculated that leaving my Pi on all the time costs me “about a tenner a year”. Let’s be generous to EDF and say it’s fifteen a year (the prices always go up, not down, somebody’s gotta pay for all those millions of smart meters…).
How much time with a RiscPC does that get you?

I’ll give you a hint, my PC draws ~100W at 230VAC; if left on 24/7, that would amount to ~€228 a year. Rather more than a tenner.

I don’t have the slightest idea what you are talking about.

I’m guessing he’s sore about the change to the broken fscanf function to work correctly, which tripped up some software that made assumptions.
More details here.

As for the “unwise and ill thought out”, you’ll notice that the message that I linked to notes that the bug was fixed in the ROLtd library a long time before it was fixed in RISC OS 5. So really, the software being broken should have been sorted out an eternity ago.

It seems a pity that we do not have a single platform that is better, in every department, than its competitors.

A rather amusing comment given we’re using an OS that isn’t capable of fully utilising the hardware that we do have. ;-)

What is your working definition of “better in every department”?

For me, cost is a factor. And I think for the ballpark of fifty quid/euros/shekels/goats ¹, a quad core Pi with 1GB RAM, a clock speed in the order of 1GB, built in networking, USB hub, and WiFi/Bluetooth on some models… even if it does have its limitations, it’s insanely cheap for what it offers and is a perfectly workable solution for using RISC OS.

Yes, the chip originally came out of some sort of media player. That’s why the graphics don’t suck. ;-)

¹ Yes, I’m aware that it’s closer to 220 shekels or about a third of a goat…

A rather amusing comment given we’re using an OS that isn’t capable of fully utilising the hardware that we do have. ;-)

Which just makes the problem easier as superfluously good hardware can be ignored.

What is your working definition of “better in every department”?

The Titanium with 4k video for example would meet this definition. Yes a bit expensive (if it existed) but it would not suffer from the criticism ‘it doesn’t do this quite as well as this other machine’.

The Raspberry Pi is then the alternative if cheapness is the important consideration.

Different core technology.

I compared i.MX6 with Pandaboard ES, both use Cortex-A9. So we can rule that out. Maybe cache sizes? I have lost track of all the different variants.

I’ve not run any benchmarks, but I perceive my Pi2 (A7 core @ 900MHz) to be faster in operation than the Beagle xM (A8 core @ 1GHz).

Unlikely, the Cortex-A8 should be a bit faster than the Cortex-A7 at the same clock speed.

Unlikely, the Cortex-A8 should be a bit faster than the Cortex-A7 at the same clock speed.

[Source]

So while it might seem as if the A7 is an A8-Lite, it is built with the benefit of eight years of advances in processor design. And, ultimately, I would imagine what the processor feels like depends both on what you’re actually using it for, and what it is attached to.
For example, the Beagle xM appears to have DDR SDRAM clocked at 200MHz. The Pi clocks memory at 450MHz by default, so it’s faster than the Beagle. I just know if it’s a little faster or a lot faster as I can’t find a source to say if the memory on a Pi2 is DDR or not.

Maybe cache sizes? I have lost track of all the different variants.

Yeah, which of the many iMX6 cores is it?
https://en.m.wikipedia.org/wiki/I.MX#i.MX_6_series

The Panda ES is an OMAP4460, with a shared 1MB L2 cache, but perhaps of more interest is that the OMAP appears to have two LPDDR memory channels. It might increase apparent throughput by simply being able to access memory more quickly?