RiscPC breaks 1.5GHz barrier - RISC OS News, Software and Information

RiscPC breaks 1.5GHz barrier

Published: 12th May 2004, 13:40:30GMT Source: drobe.co.uk
By Ian Hawkins

Page 1 of 1

Calm down dear, this is a case mod

You may have read my previous article in which I took a standard 200MHz StrongARM and overclocked it to a stable 305MHz with nothing more than 2 peltier effect chips and a couple of large heatsinks. Well the butchery has not stopped I'm afraid, and in this article I take a standard ARM610 RiscPC1 and transform it into a power spouting, heat pumping, 1.67GHz workhorse2.

Disclaimer: This article assumes you're competent with metal based practical work and any modifications carried out on your kit is done entirely at your own risk.

Sweet innocence
Here we have an ARM610 based RiscPC. Once the flagship of Acorn's range of computers, it has lived for over a decade and has more than proven that the design of computer from motherboard to case has been through thick and thin, and the whole concept works, or worked, extremely well. Indeed people still, alledgedly, use these as their daily machine - which isn't bad going and as a minority platform, it's quite useful when all the email worms, virii and other problems with today's computers simply don't happen.


Acorn RiscPC 600		2 or more slices = expandability

So to recap, in today's harsh and fast moving, world the RiscPC is now:

Simple
Slow
No frills
Slow
Has a very expandable case design
Did I mention that it was slow?

The ARM610 is slow. So is an ARM7500. Useless when compared to today's computers. A StrongARM may have stood a small chance, but basically they're all way, way too old. But PCs, now they're going somewhere and so is emulation. So, the first thing we do is open it up and rip out all the guts inside, making way for our modifications.

RiscPC base, void of electronic bits

Planning the comeback
Now we have a nice empty case ready for new, faster, better, innards. After a quick look around to see what I have available, I return with the following:

1x 40GB ATA-66 IDE Disk
1x 52x IDE CDROM drive
1x Micro-ATX form Motherboard
2x 256M PC-133 SD-RAM
1x Athlon XP 2000 Processor
1x Flex ATX power supply (180W)

We now have to make sure that everything is going to fit before I commit myself by drilling lots of holes and getting busy with a fretsaw. The worst thing is having a 'swiss cheese' case, or something that looks like a 3 year old tried to sit on top of to make things fit.

So first we take the base part of the RiscPC case, and lay our components out on it. As you see from the photo below, everything fits quite snugly and I am even able to fit the harddrive into the original mounts. The PSU is the same height as the RiscPC PSU, but considerably shorter in length, and width. If you look closely you can see a right-angle PCI connector adapter. This will come in handy later on in the article

Once everything is nicely arranged in the bottom of the case we mark the positions so I can later drill holes. A fine tipped pencil is ideal for this.

Laying components out

Careful now
Now everything on the base unit is marked out we can finally get busy with the power drill and hacksaw. I waste no time at all in chopping large chunks out of the case for the ATX bracket, and drilling holes for the motherboard mounts.

On most RiscPC base units the insides are coated with a rough conductive type of paint. This was intended to reduce the electromagnetic inteference that the RiscPC generated - to my knowledge it only just scraped through the tests. Because of this you have to treat any swarf generated by cutting as if it was made of metal (ie: as if you were drilling holes in a metal case) and removing all traces of it before putting your new CiscPC guts in, otherwise you could be in for a shock. No, really.

After carefully milling parts of the case out you should end up with something like the image below:

Hacking the base for the ATX bracket

Good, it still fits
Once all the drilling is out of the way then it's time to start assembling the base. I had an old ATX case lying about, so I pinched the motherboard mounts from it and carefully attached them to the holes drilled in the case. One of the best things about the RiscPC case is that it is made of plastic. This means it is really easy to cut bits out, or cut huge chunks out of.

The base unit with everything mounted so far

Sliced, diced
Once the base unit is more-or-less out of the way with it's time for the first slice. RiscPCs can be upgraded very easily by a simple stacking system comprised of 'slices'. Each slice has room for 1 x 5.25" and 1 x 3.5" as well as expansion card space at the back.

All we need to do for the first slice is remove a small chunk at the back for the ATX metal backet, and chop a small section (pictured below) out from one of the plastic brackets in the middle of the case to allow the RAM to snugly fit under it. Again, if the case is coated make sure you remove all swarf after cutting.

Thanks to the ideal placement of the podule bay, it makes it really easy for us to ditch heat from the CPU out of the case. As you can see from the picture below the simplest solution is to literally just blow the air out the back of the case (through the now vacant podule slot). The blower used is the same high-volume one as is used inside the webserver drobe.co.uk resides on.

One of the plastic brackets needs adjusting.

Too tall, sir
Full height PCI cards will not fit into the case if stood upright, there is just no room in the RiscPC case if you are only going to use 1 slice. If you're going to use 2 slices then go for it. You're only going to be limited lengthways by the 5.25" drive bay that the CDROM drive is normally housed in. For this exercise we are going to be using half-height PCI cards, but will use a right-angled PCI connector adapter to make sure we have no height problems.

Full height PCI cards are a no-no.

Assembled
Here you can see everything assembled. Everything fits quite nicely, even the graphics card has room to spare. As you can see the power supply is quite short, and has no power switch at the front unlike the original RiscPC power supply. The original power switch can be used as long as you find a suitable replacement for it to push against. I used a normal ATX power switch from an old spare case lying about, and with a little help from some superglue, fixed it in place at the front of the case so I could keep it looking 'normal'.

The CiscPC with new innards.

First stages of the dark side
Here's the rear of the case with everything plugged in and working. If you look on the LCD screen to the right you can see the box is busy running the ever wonderful Windows XP. It could equally be running Linux just as well.

We are gates of Borg. You will be assmutilated.

Still to be done
As you can see, the back of the RiscPC still leaves a lot to be desired with the large gaping holes, and the case also looks a tad 'tatty'. The holes will be solved with a simple bit of wire mesh, to allow air to still be directed out of the case, to keep things from falling inside the case and to make things look nice.

The case will also be degreased and receive a coat of paint. I'm thinking possibly a nice orange colour would go down a storm.

So, what have we ended up with? Lessons learnt
Well, I now have a faster-than-StrongARM RiscPC, with the aid of VirtualAcorn, and a PC that can run the latest applications, games and spyware from the darker side of the OS world. Who needs all this native computer nonsense when it's dead simple to build your own for a little cash, readily available components and a little time, around 7 hours.

The RiscPC case is stupidly easy to modify to take an ATX style motherboard, and it took no time at all to do it. The resulting CiscPC is quite quiet in operation (the blower is very quiet as opposed to a conventional fan) and also looks cool. Which is what counts.

Damage done
I had to buy a couple of bits (that were not donated) to this little project, and here's a rough breakdown:

Component	Bought from	Price
Flex ATX PSU	Maplin	about 45ukp
ATI Radeon 9000SE PCI	PC World	49ukp
Pot of superglue	Halfords	3.99ukp

TOTAL		~ 97.99ukp

1: Well, the case survived, nothing much else did.
2: I call this creative license.

Links
Back to the front page

This article has been linked to, or is available in the following formats:

[Printable] [Digg this] [Blog search]

Phlamethrower (-1.0) 12/5/04 3:47PM	You evil, evil man
nunfetishist (-1.1) 12/5/04 3:59PM	Why call it "CiscPC" ?
Rimmer (+1.0) 12/5/04 4:01PM	nunfetish: Risc comes from the fact that the ARM is a RISC (Reduced Instruction Set Chip) chip PC's use CISC chips (complex instruction set Chips) At least, that's what I think they mean...
fylfot (+1.0) 12/5/04 4:05PM	"The ARM610 is slow. So is an ARM7500". Ooh, I just love the politics.
fylfot (+1.0) 12/5/04 4:08PM	"I'm thinking possibly a nice orange colour". OK, that as well. Nothing gets by the fot!
simo (+1.0) 12/5/04 4:50PM	Now we just need to get a 3.4GHz PC in there and we can say the RPC600 outperforms an Iyonix! Good article.
j5m1th (+1.0) 12/5/04 4:58PM	No problem doing that, just spend a bit more cash... Now there's an idea for what to do with my dead RiscPC
Jaco (+1.0) 12/5/04 6:19PM	Of course they're called CISC when they execute complex instruction sets. That's what the word means. Even if the micro code core of a x86 is RISC the complete thing can (and must) still be called CISC.
imj (+2.0) 12/5/04 7:19PM	Assuming you mean, running VirtualRPC, a 3.4Ghz PC does not outperform an Iyonix. I have basic P4, 3.4Ghz machine (512K), 800Mhz FSB DDR RAM and a fast nVIDIA graphics card, built for a few hundred. Tests of raw emulated CPU show it's about 455Mhz effective speed. Disc IO is ludicrously fast, of course. So maybe one of the "Extreme Edition" 3.8Ghz P4's with 2MB cache would get us toward Iyonix speed at a guess. It's bloody noisy though!
AMS (-0.9) 12/5/04 7:34PM	*In reply to Ian:* The Extreme Edition is massively overpriced (IMHO) and also has cache ECC turned off (not a good state of affairs - but as it was targetted at "Gamesters" in the PC market Intel probably felt nothing too critical would be effected should things go a bit "tragic"). It would probably be more prudent to wait for an appropriately fast PC CPU (rather than a "forced"* arrangement like the Extreme) to arrive. Extreme would still be slower (but closer to*) the Iyonix in performance terms - - - but probably just as expensive [Dabs price the 3.2GHz Extreme at a stunning 648.99 gbp and that's just for the processor !!!!]
Dougal (+2.0) 12/5/04 8:26PM	nunfetishist: Please, let it go. It was just a joke calling it the CiscPC, it was funny, and attempt to make us laugh. Last time I checked jokes didn't need to be literally true. The external interface to the pentium whatever is CISC, even though we know that modern day x86 processors do run RISC cores. The case of Crusoe is an unusual one - do you call something by its deep internals or by its external specification? An interesting question, but hardly on topic for this conversation. If you're going to be a picky sod (and clearly you are), the name RiscPC was a misnomer too, as the ARM processor isn't truely a RISC processor. It's mostly a RISC processor, but like all real work RISC processors it sacrafices principles at times for the sake of performance (e.g., multiple word data transfers). The CiscPC was a good pun at the end of a nice article. Why you have to try and spoil things I don't know.
Sawadee (+1.0) 12/5/04 9:55PM	Amusing upgrade idea for my older model machines. What about Adjust Select and 32 bit???
nunfetishist (-1.0) 12/5/04 11:56PM	*In reply to Dougal:* Because it's an unnessasary smear that just spreads false information. That's why. (Yes, the ARM clearly isn't truely RISC, because of the existance of the MOV and CLZ instructions, and even LDM/STM are debatable.)
md0u80c9 (+1.0) 13/5/04 2:10AM	Have you no shame but to savage a poor, unARMed RiscPC :o(???
hzn 13/5/04 8:20AM	Well, seems to me that case modding hit us here On the other hand putting a PC in the RiscPC case does give you the chance to fool or impress the odd user. I hope you closed that big hole in the back and so on so that the disARMed RiscPC does comply to the odd standard for electro magnectic distoption (or whatever that's called in the UK).
AMS 13/5/04 8:48AM	Oddly I am not even sure this is a case of case modding (most of what's done involves chuking existing electronics out and putting new electronics in - the case largely seems to remain the same )
Spriteman 13/5/04 9:28AM	I liked the CiscPC pun :laugh: -- Spriteman
hzn 13/5/04 10:31AM	To AMS Doesn't this depend on the point of vier: Take a PC and put it into a RiscPC case ... :angel:
mavhc 13/5/04 1:24PM	Why does having more than the minimum number of instructions disqualify a CPU from having a Reduced Instruction Set?
AMS 13/5/04 1:44PM	It doesn't. The ARM,6502 and M68000 are not worlds apart in the number of instructions they use. It's got to do with the complexity of the instructions and number of addressing modes. The 6502 and M68000 have lots of addressing modes. The 68000 also has quite complex (almost high level) instruction types (so is definately a CISC). The ARM has fewer addressing modes, most commands are relatively simple (so would IMHO qualify as a RISC), it does (however) have a few instructions that are a bit more complex than a RISC should have - things like multiple store/load (STM/LDM). As most of its instructions fall into the RISC camp - I'd suggest (tentatively) its a RISC, but not a typical one
flypig 13/5/04 1:56PM	The minimal number of instructions a computer can have (technically) is just one, so resumably a RISC CPU has to have more than the minimum: http://en.wikipedia.org/wiki/OISC
2 comment(s) are below your moderation threshold. Login to view them.