Max no of male sheep with Emulator

My Risc PC eventually grew to 128MB, with a 64MB SIMM and 64MB on the Kinetic. I was never that pleased with the performance of the slightly faster Kinetic memory, and I found myself spilling over in to the normal RAM a lot when using Photodesk. But the main issues was the I/O was hopeless without DMA available, so in the end I went back to a 233MHz StrongARM and got another 64MB SIMM, and was able to use DMA again on fast IDE card.

Am convinced Raik is talking about Kinetic.

No. I’m talking about a RiscPC 600 with StrongGuard and StrongArm with RISC OS 4.39. With ArmSwitcher it was also working.
But it take same times before I find the right EDO RAMs for this. I have try any variations before I find stable working.
As I study I work in a computer recycling company and collect all PS/2 RAMs I found ;-)

Back in the 90s when we were using Risc PCs in industry, we were buying lots of different RAM chips and found most of the EDO was unreliable so stuck it in PCs instead, as who would notice.

Were the PCs themselves actually more tolerant of out-of-spec EDO? How tolerant RiscPCs were varied, I’m pretty sure – the same EDO that didn’t work in one RiscPC sometimes worked perfectly well in a different one. (We didn’t have any PCs to try them in.)

Some interesting things have turned up – is there such a thing as 256 EDO ram modules?

What are the x2 ram adapters called? If they are cheap enough many ram modules lying around could be used in RPc etc.

Several threads on stardot over the years regarding Risc PC / A7000 (+) memory.

Poor bus driving makes system intolerant of high-chip-count SIMMs, apparently worse with StrongARM. And not all SIMMs flogged as 5V (which RiscPC requires) are, many are 3V3.

Only A7000+ using ARM7500FE should use EDO RAM in EDO mode. I quite like “Interestingly, while the 7500FE can support a mixture of FPM and EDO banks, it looks like RO 3.71 doesn’t attempt to do any detection and just forces EDO mode. So if your FPM SIMMs actually are FPM SIMMs then they’re running out of spec.” [Edit: Interesting Q is what RISC OS 4 does with RAM setup]

is there such a thing as 256 EDO ram modules?

Google says yes, even as 72 pin modules… however it seems that the typical way to get 256MB was using two 128MB SIMMs. Plenty of those on offer.

Amusingly, entering “256MB EDO SIMM” into Amazon provides a long list of 256GB SD cards. Well, I guess they got the number right, if nothing else…

How tolerant RiscPCs were varied, I’m pretty sure

I dimly recall comments way back when about the StrongARM having a weaker bus driver so memory that worked well with the ARM610 or 710 processor cards would fail with a StrongARM. But since my RiscPC only ever had a 610 and later a 710, I can’t comment.

Eeh, if only we had the RISC OS 4.39 kernel source to examine. :-)

Were the PCs themselves actually more tolerant of out-of-spec EDO? How tolerant RiscPCs were varied, I’m pretty sure

The PCs could handle just about any EDO SIMM*, it was only out of spec for the Risc PC’s which were very fussy about their memory. If buying personally it was always best to use a RISC OS dealer, and getting something known to work.

*Windows would crash regardless of whether the memory was good or bad.

I remember there was a lot of timing problems. Things are get better after I send my board to UK but I not remember the details.
All things are fine with using the hardware StronGuard.
The ArmSwitcher also fix any timing problems.

Are we discussing these details in the RPCEmu forum to lobby for an option “behave unstable/crash randomly like a Risc PC with incompatible PS2 SIMMs” for true Risc PC emulation?

Looks that way. And I would love to know where that extra address line from ARM to IOMD and also from IOMD to the SIMMs has appeared from. Wouldn’t have needed IOMD2 to support 512MB on Phoebe then, would they? ;-) Smells like 4.39 miscounting somehow.

I wonder if these were the adapters used-
FutureWare SimmXtenders. These allow you to double your 72-pin simm count in a machine. Each extender converts 2 72-pin simm memory modules (FPM, EDO) into 1 72-pin simm slot.

It’s important that they be a matched pair because they are mirror images of each other. That way they can be installed back to back in adjacent simm slots.

Back to the start – if the RPC could be made to handle 512Mb ram – perhaps the Emulators could be made to handle 512Mb? :-)

The RiscPC could not use > 256MB of RAM (excl. VRAM). Just look at the physical memory map (it’s in the TRM) – only the bottom 29 address bits go to IOMD. Half of this space is given over to RAM, the rest to VRAM, I/O and ROM. The RAM space is divided into four: two slots that may each contain two banks, max 64MB per bank.

Certainly you could use SIMM doublers to stuff more RAM into the system but it may not be visible to IOMD and thence to the CPU. I suspect the extenders are for cases where you only have single bank SIMMs to hand and systems that can support double bank SIMMs. They can’t magically increase the underlying capabilities of a system. And the RISC OS 5 IOMD HAL knows that the limit is 256MB (for non-Kinetic).

Emulators with better CPU support and more RAM would be great – see other threads – but they are not RPCEmu.

Raik fitted 2 of these extenders and the OS reported 512Mb how did it figure this out?

Wasn’t room for the Vram card so he went back to 3 128 modules.

I wonder if later mother boards were different from the TRMs.

There was a Arc prog that tested memory – I wonder if that would shed some light.

Raik filled the 256Mb ram disc with large sprite files – didn’t say that got corrupted.

The SIMM doublers were from when larger capacity SIMMs were prohibitively expensive, so allowed the use of two smaller SIMMS. By the time we got close to maxing out the machines memory, you started hitting tolerance issues with such devices, so had to use proper high capacity SIMMs.

I haven’t really tried 2x 256MB RAM. Without VRAM it wasn’t usable for me. The RiscPC ran stable for a very long time, only the Connect32 sometimes caused problems and a RISC OS before 4.39 was necessary for the reconfiguration. I had two revisions of the Risc PC board. The one with 90° RAM sockets didn’t work, the one with 45° sockets worked as described. With a lot of time I can recreate the situation. I still have all the parts here, but I took some things out again because I no longer had space in the hobby room and wanted to have a smaller Risc PC again (the old case was a PC tower). Everything is currently in the attic of my garden house. I’m not sure but I think I read in the German RISC OS manual that the Risc PC can use up to 512MB of RAM. I could take a new look.

My RiscPC eventually had 32MB installed.

I think my first one came with 17MB (16MB SIMM plus 1MB VRAM). I upgraded the VRAM first, then eventually managed to get hold of a couple of 64MB SIMMs, so had 130MB (128+2).

Never ran out of RAM on that.

The second machine, bought as a CJE Developer Special, if I recall came with a single 32MB SIMM installed and 2MB VRAM. I had a ‘spare’ 32MB SIMM from the first RPC after an interim upgrade. So that one ended up with 66MB (64+2).

Writing this on a Windows 11 machine bought brand new in April this year for £230 that came with 4GB RAM and I’ve added a 16GB DIMM – so 20GB total. Looking to swap the 4GB for another 16GB giving me 32GB of RAM. The RPC couldn’t even handle a hard disk that big, never mind RAM. How times have changed!

Raik – interesting stuff – please don’t bother hunting down hardware bits for me – but if you should come find the German RPc manual :-)

Most likely finding it when looking for something else :-)