Mayhem Accelerator - Prototype Complete [Sam Coup

Quazar · June 2006

I am now pleased to announce that on the 8th June the prototype of the Mayhem Accelerator was finally completed.

With work on the accelerator starting in January 2004, then first unveiled as a work in progress with an initial (and dodgy!) 12MHz speed at the ORSAM show in November 2004. With work on it spanning two and a half years on and off this has been my biggest Sam Coup? project to date over the 11 years that i?ve been working on the Sam.

There's been plenty of hiccups along the way, including the latest just a couple of weeks ago when I blew the prototype, so for quickness I rebuilt another full prototype from scratch and carried on where I had left off as I really wanted to get the hardware finished!

What does the Mayhem Accelerator offer? Speed! It?s what the Sam Coup? has been crying out for and the Mayhem Accelerator speeds up the Sam to 10MHz, 16MHz and 20MHz speeds, with uncontended memory reading for a further boost. The final design also allows a 6MHz ?legacy? speed to be selected as well, to run at stock speed without having to unplug the interface.

Using ?BogoMIPS? (by Frode Tennebo) and ?Fractal Explorer? (by Dr Andy Wright and Simon N. Goodwin) as two test programs to give some benchmarks of the final design here are the results:

Fractal Explorer (generating the default Mandlebrot image)

Normal Sam: 21.30 seconds
Mayhem @ 10MHz: 10.56 seconds (202% speedup)
Mayhem @ 16MHz: 7.20 seconds (296% speedup)
Mayhem @ 20MHz: 6.06 seconds (352% speedup)

BogoMIPS

Normal Sam: 0.360 BogoMIPS
Mayhem @ 10MHz: 0.748 BogoMIPS (208% speedup)
Mayhem @ 16MHz: 1.208 BogoMIPS (336% speedup)
Mayhem @ 20MHz: 1.513 BogoMIPS (420% speedup)

The final stage now is to complete the designs for the PCB for the production version and build one up to go through testing, then the Accelerator will be ready! All going well it will be available in 8 weeks time at the start of August.

More information will be in the forthcoming issue of Sam Revival (issue 15 - out later this month), along with pricing and preorder information.

For previous news and updates please see:

ORSAM 2004 Show report - Sam Revival issue 11
Creating Mayhem article - Sam Revival issue 12
Quazar News - Sam Revival issue 14
www.samcoupe.com website - for news and videos of the
Mayhem Accelerator in action.

All the best,

Colin

Quazar · December 2006

After a somewhat abrupt break from things which lastest a few months, i've endured a hard slog of about 30-40 hours over the last few days to get back on track in the process of turning prototype to finished article and completed the final PCB designs for the Mayhem Accelerator, with 1-off quantities ordered bright and early this morning for testing, before going ahead and getting the first full batch made up.

Also on the horizon this month is issue 16 of Sam Revival magazine, with both issue 17 of SR and the first built up batch of Mayhem Accelerators following very early in the new year!

Colin.

DEATH · December 2006

Mayhem Accelerator - ain't that something like Hawkwind's Argone Accumulator? :D

Quazar · December 2006

I?ve received now the 1-off quantities of the PCBs for the Mayhem Accerlator, which i?m busy soldering up and testing! (Photo up on my website at www.samcoupe.com)

I should be finished all the testing sometime next week - and then i?ll be going ahead and getting a full batch of PCBs made up and getting the preordered units out early in the new year.

Colin.

CKay · December 2006

one word... Duhude :D

haven`t checked ya` page in a while... the new SAM startup screen looks amazing, have you had to mess on with the ROM as well (might be a stupid question, I don`t know..)

Did you get my last mail Colin? There was a couple of things I was interested to know...

anyway, your hard work is appreciated ;)

Quazar · December 2006

Yeah, i've been tinkering. The Mayhem ignores whatever ROM is physically in your Sam as it has it's own as it needs to be able to read it at a fast speed! Most of which contains the standard ROM3 (with permission from Andy Wright) as well as my own code for the a few things such as the new startup screen etc to show what speed the Mayhem is running at.

Erm.... let me go through archived emails... but to be on the safe side chuck me a PM here and i'll get back to you later tonight...

Chris P · December 2006

Mmmmm, methinks an accurate clone of the Williams classic 1982 blaster "Robotron 2084" might be possible with a 20mhz Z80 to push things along!! ;-)

Quazar · December 2006

Chris P wrote:

Mmmmm, methinks an accurate clone of the Williams classic 1982 blaster "Robotron 2084" might be possible with a 20mhz Z80 to push things along!! ;-)

Hmmm, nice idea Chris :)

Or... could the 6809 be emulated... there's a few arcade games that use the same hardware isn't there, I see Defender does...

Chris P · December 2006

Quazar wrote:

Hmmm, nice idea Chris :)

Or... could the 6809 be emulated... there's a few arcade games that use the same hardware isn't there, I see Defender does...

Emulation of the 6809 would be possible - but it's such a different beast to the Z80 (2 stacks, big-endian, hardware mult, Etc.) - that even a 20mhz Z80 probably wouldn't leave much room to emulate the other parts of the system. Remember, the Williams games used separate hardware for sound, with another 6809(ish) CPU on board. :(

The biggest hurdle would be the later William's games, Robotron included, use of blitter chips - a pair of them, devoted to block-copying large chunks of on-screen data - these were cunningly called "Williams Special Chips". These beasts could shunt just under a million bytes-per-second (complete with masking) and independent of the main CPU.

The early Williams games, Defender, Stargate (Defender II), simply had a 6809 and a nibble-per-pixel bitmap - no blitters. They still had separate sound hardware though.

A 20mhz Z80 might struggle a little emulating 2x6809's and a couple of high-speed blitters all running in parallel! :wink: Mind you, it might be an interesting exercise though - if I had the time... :roll:

Perhaps a clone - written from scratch - would be a possibility. Looking at "Wave 9" in Robotron... That's an awful lot of data for a single CPU to shunt on-screen!!! :wink: Oh, and it would have to have the twin joystick control method - the keyboard simply wouldn't be an option! :wink: :wink:

Mmmmm, if only I had a good few months of uninterrupted coding time... :( :(

Quazar · December 2006

Chris P wrote:

Emulation of the 6809 would be possible - but it's such a different beast to the Z80 (2 stacks, big-endian, hardware mult, Etc.) - that even a 20mhz Z80 probably wouldn't leave much room to emulate the other parts of the system. Remember, the Williams games used separate hardware for sound, with another 6809(ish) CPU on board. :(

The biggest hurdle would be the later William's games, Robotron included, use of blitter chips - a pair of them, devoted to block-copying large chunks of on-screen data - these were cunningly called "Williams Special Chips". These beasts could shunt just under a million bytes-per-second (complete with masking) and independent of the main CPU.

Blimey! Didn't realise the 6809 was such a complicated beast, let alone thinking of the custom chips in the arcade!

Chris P · December 2006

Quazar wrote:

Blimey! Didn't realise the 6809 was such a complicated beast, let alone thinking of the custom chips in the arcade!

Yes, it's quite a different beast. You can do things in 6809 in around half to a quarter the time taken in Z80. A 1-mhz 6809 is roughly equivalent to a 4-mhz Z80 - given the way the 6809 breaks down its instruction execution process. The following was taken from a 6809 document, and being a hardware guru you will understand:

"The 6800 and 6809, like the 6502 series, used a single clock cycle to generate the timing for four internal execution stages by using the rising and falling edges of the base cycle (not just rising edges), and another clock 90 degrees out of phase (giving two rising and two falling edges per cycle) - this allowed instructions to execute in one external 'cycle' rather than four for most CPUs..."

Sort of makes me wish MGT had equipped the SAM with a 4-mhz 6809... Oh, and a blitter or two! :) :) :)

Quazar · December 2006

aowen wrote:

It has occurred to me, because AA mentioned it in another thread about cartridges for the TS2068, that an timer capable of generating an interrupt could be a useful addition to the mayhem board.

The Sam already has line interrupts, which you can set to trigger at any of the 192 scan lines for the displayable area - and most commonly used for things like changing the screen mode or palette registers etc.

I had originally planned to have an extra interrupt generator, to aid sample playback and such instead of using the line interrupts and timing in the border area for sample timing with the likes of the Quazar Surround soundcard, but I decided early on not to incorporate it as I really don't want to have any hardware extras that means the software becomes 'Mayhem Only'. The concept being that it means any new software will run normally on the Sam with or without the Mayhem, even if it does take advantage of the extra speed when it detects it.

(And thankfully the hardware design phase is now finally over with the prototype completed a few months ago and the final PCBs to be tested over the next few days .... a few years of tinkering in total (on and off) and a very hefty sting in the wallet! :) )

CKay · December 2006

Quazar wrote:

Yeah, i've been tinkering. The Mayhem ignores whatever ROM is physically in your Sam as it has it's own as it needs to be able to read it at a fast speed! Most of which contains the standard ROM3 (with permission from Andy Wright) as well as my own code for the a few things such as the new startup screen etc to show what speed the Mayhem is running at.

Erm.... let me go through archived emails... but to be on the safe side chuck me a PM here and i'll get back to you later tonight...

Thanks, forget what it was I was wanting to know :-P, so I`ll check and PM :)

Quazar · December 2006

aowen wrote:

I sympathise. I spent at least £300 prototyping the Spectrum SE, and eventually it was unviable to put into production. At least it's emulated though.

Thanks, it can be quite hard to really explain just what goes into hardware development as you'll know yourself. When I started with the Quazar Surround soundcard for the Sam Coupé way back in 1995, it cost somewhere in the region of £500 from my own pocket to get it up to the stage where the first batch of units were shipping.

With something like the Mayhem it's many times that, with a fair whack been spent out getting in new test equipment (new oscilloscope, several IC programmers - at least with equipment it can be used for other projects but it's still a hefty purchase at the time!)

And also what people don't realise is the time .... I just wish I got paid by the hour when im working on my own stuff!

Mayhem Accelerator - Prototype Complete [Sam Coup

Comments