Faster/wider CPU...

Nagging hardware related question? Post here!
User avatar
vanpeebles
Commissario Pebbli
Posts: 2136
Joined: Sat Nov 20, 2010 7:13 pm
Location: North East UK

Re: Faster/wider CPU...

Postby vanpeebles » Tue Jan 31, 2012 3:41 pm

Brane2 please keep to resonable and friendly debate. :)


User avatar
Dave
SandySuperQDave
Posts: 2417
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Faster/wider CPU...

Postby Dave » Tue Jan 31, 2012 3:53 pm

Sorry, I didn't know a counter-argument was needed.

I'll do this because I want to, regardless of what you think ;)


Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 4:23 pm

I'm not trying to be an asshole.

I'm just offering my point of view.I took teh time last night to reread/skimp through QDOS Bible in MArcel Kilgus page one more time after so long. I took a look at SMSQ/e docs.

I still stand by my opinion. Even SMSQ/E has many problems that need to be adrressed and holes to be filled.

And everything needs to be reinvented. Just patching with bog-down available stuff is not enough.
Patching is enough to make it work somehow, but not for being competitive with mainstream solution in at least one of market niches.


On the journey of life I chose the psycho path...
Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 4:41 pm

Dave wrote:Sorry, I didn't know a counter-argument was needed.

I'll do this because I want to, regardless of what you think ;)



Sure. You are revolutionary with a vision and I'm kind of human sheep, trying to stop you.
If that's how you want to see it, fine with me.

But just FYI, you are not the only one with the ideas.

In 1989/90 or so, I was full of ideas on short on money. And even shorter on equipment. But with great vision. Ive found a kindred soul and persuaded him to byu one QL for himself. He was living in a house some 7km away from my flat. I can't count how many days and night did we loose trying to make wireless connection between us through network interface. That was long, long before anyone around us did even know what wi-fi means.

He and I routinely did hardware we needed and connected it to QL CPU, be it simple programmers or switches or whatever we might need at the moment.

So, if someone had headbutted every day against numerous shortcommings and limitations of QL, i suppose we were somewhere at the top of the list.

This is not trolling.
If you want machine that people will actually find useful and use, then you'll have to adress those issues successfully in some way.

If all you want is to feel fuzzy and warm while recreating 68000-based toy for umpteenth time, fine, but don't expect to have much audience during "QL was SO much better" lectures.


On the journey of life I chose the psycho path...
Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 5:01 pm

I forgot one of the more annoying QL's features:

because of various HW/FW/SW deficiencies and mistakes, many programs took illegal route and accesed various resources directly instead through OS, so any change on FW/HW had some negative consequences.

So, after all that talk about modular design developers ended up scared from simple changes and upgrades.

Buck has to stop at some point. Which was IMHO reached long, long time ago.
Last edited by Brane2 on Wed Feb 22, 2012 5:29 am, edited 1 time in total.


On the journey of life I chose the psycho path...
User avatar
Mr_Navigator
QL Fanatic
Posts: 782
Joined: Mon Dec 13, 2010 11:17 pm
Location: UK, Essex
Contact:

Re: Faster/wider CPU...

Postby Mr_Navigator » Tue Jan 31, 2012 5:08 pm

Apologies if I caused offense, I did say IMHO :(

I am not claiming to have all the answers, in fact I left the QL scene in the late 1980's and only returned just over a year ago so relatively still a newcomer. One of the reasons I came 'Back To The QL' was a) fond memories of programming b) I liked SuperBASIC and most important of c) It had nothing to do with work, I teach IT, ICT & Games Development.

But back to the case in point.

This is an interesting Post on another forum (spit spit) http://www.powerdeveloper.org/forums/viewtopic.php?p=14610

and even better a wiki on the minimig http://en.wikipedia.org/wiki/Minimig, what is really an interesting point is this:

Minimig started in secrecy around January 2005 as a proof of concept by Dutch electrical engineer Dennis van Weeren. He intended Minimig as the answer to the ongoing discussions within the Amiga community on implementing the Amiga custom chipset using an FPGA.


and also this which comes under the Purposes and Intent

Run Amiga specific application software to convert files to newer platforms.
Run software only available on Amiga.
Running Amiga video games.
FPGA development experience using Verilog.
Creating something for the community.
Proof of Concept.
Allows creation of new games that take advantages of the new features in Minimig (faster memory, more memory sprites, colours, etc), while maintaining full compatibility with the Amiga


Dave, do you think it might be an idea to start with something like "Purposes and Intent" and then go from there?

two more points before shutting up [Star Wars reference]

1. Farnell sell this for about £6 is this the same device as you are refering to Dave, the MC68EC000EI8 ? they have about 15 in stock btw


2. The spec for this device is 8 Mhz but I seem to remember an article somewhere saying they overclocked this quite comfortably at twice the frequency, anyone else?


-----------------------------------------------------------------------------------
QLick here for the Back 2 the QL Blog http://backtotheql.blogspot.co.uk/
Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 5:19 pm

If you go for 68EC, there is IMHO no reason not to go for 20 MHz version in smallest housing you can find.

This is AFAIK QFP. It pays to have smallest housing, since it will have smallest parasitic inductances, which is especially important on GND and Vcc lines. With good decoupling caps they will perform the best and because of the thinnest plastic they can be most effectively cooled, if needed.

That being said, there is very limited gain to be had with 68000. You can gain say 2x performance increase at the same frequency with 16-bit bus and no waitstates and multply that with say 4-5 due to clock increase. All in all, 10x original abysmal performance.

It would sound very nice in 1985, but now it's more in the "peanuts" category...


On the journey of life I chose the psycho path...
User avatar
Dave
SandySuperQDave
Posts: 2417
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Faster/wider CPU...

Postby Dave » Tue Jan 31, 2012 5:35 pm

I was looking at the 16MHz variant of that dame IC, and I have a source here in the US at $4/part, QFP. The nice thing about it is it avoids ALL the compatibility issues.


Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 5:36 pm

One more thing, if you decide to go with 68EC route:

Chips can be overclocked fenomenaly, but:

1. Their outputs have to be loaded as lightly as possible.
2. pumping-up the juice - ( powering them on the high side of allowed range, like 5.25V )can be helpful.

3. 68000 bus was never meant for such high frequencies, so although whole bus caycle time is 4T, time actually available for read or write is much shorter, IIRC 1,5T for read and 2T for write.

Which means that with CPU clk of 32MHz bus cacle time would be just 125 ns ( 8MHz) and that you needed RAM read acces time of just 32ns and write time of just 65 ns.

With that in mind, I tried to make intelligent decoder- one that would actually scan what instruction is CPU actually executing and based on that predict next accesses and start the cycle early.

I never finished it, but I suspect it should be doable...


On the journey of life I chose the psycho path...
Brane2
Trump Card
Posts: 179
Joined: Fri Dec 30, 2011 5:42 pm
Location: Ljubljana, Slovenia

Re: Faster/wider CPU...

Postby Brane2 » Tue Jan 31, 2012 5:48 pm

Two more ideas for 68EC route:

1. I was playing with it long time ago and then there was no SEC version, which is now available.
Since it is meant for 2-3V Vcc and more "modern" than EC, it could very well be that it is even more insanely overclockable. To top it off, you can try povering them higher, like 4 or even 5V. There is good chance they would take it.
There might be some issue with MOVE SR being privileged, though IIRC.

2. Instead of using DRAM for "slow RAM", very good option would be cache SRAM chips from 386/386/Pentium era motherboards. Those chips were usually of 16-128kx8 organisation and their access time was 15-25ns. And they were cheap. Of course, that would mean that you would have to make your own ULA1 within CPLD, but those are cheap and ULA1 was really simple. And since no one uses flash bit in mode 8, you could ignore it or use it for extra 8 colours. And you could easily make your own modes with that kind of budget...


On the journey of life I chose the psycho path...

Who is online

Users browsing this forum: MSN [Bot] and 6 guests