Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Motormouth on October 21, 2015, 04:03:15 AM
-
Has anyone every tried the 14 Mhz 68000 mod on their A1000, A500, or A2000?
I have a spare A2000 and A500. I was consider modding one or the other.
How is compatibility with Zorro II boards or Side Expansion slot with this Mod?
-
It's not meant to give a massive speedup IIRC, as performance is limited by other factors (chipset and bus speed, the 68k has no on-board cache). Nothing like what the doubling of speed would suggest.
Which is why 68k socket accelerators also include their own fast memory as well as using CPUs with caches.
-
It's not meant to give a massive speedup IIRC, as performance is limited by other factors (chipset and bus speed, the 68k has no on-board cache). Nothing like what the doubling of speed would suggest.
Which is why 68k socket accelerators also include their own fast memory as well as using CPUs with caches.
Agreed, This would basically be like the ICD AdSpeed without the cache.
I was only expecting like a 5-10% speed increase.
I guess to be more specific, Is the time to do the hack and the less than $20 US of parts worth it vs the hardware and software incompatibilities. Ultimately I would put a scsi controller and some fast ram on either config via Zorro II for the A2000 or Zorro II to side expansion on the A500. Examples: I here from other sources floppy timing can get screwed up.
At the end of the day I was wondering if anyone had tried it. or if CIA chips blow up
-
At the end of the day I was wondering if anyone had tried it. or if CIA chips blow up
I think you'd have more to worry about from floppy drive timing. Isn't this the hack that screws that all up?
IMHO I was interested in it some years back but the amount of work wasn't worth my time. Get a 68010 for the 2-5% boost it gives you for $5 on ebay (and be able to make use of the 'quit' key in WHDLoad), then add some fast ram for the 20-25% increase that gives you. Much better bang for your buck if you want to slightly boost a low-end machine. :)
That being said, if you try it let us know how it goes, and post some pictures! :)
-
I think you'd have more to worry about from floppy drive timing. Isn't this the hack that screws that all up?
No, that should be harmless. The average 14Mhz hack only speeds up the CPU, but none of the external chips that are not designed to run at a higher clock rate anyhow. The timing for the disk drives is derived from the timer.device, and hence from the CIA hardware timers that still run at ~700Mhz. Hence, no problems to be expected.
Otherwise, what would a floppy do with a 68060@50Mhz? (-:
-
No, that should be harmless. The average 14Mhz hack only speeds up the CPU, but none of the external chips that are not designed to run at a higher clock rate anyhow. The timing for the disk drives is derived from the timer.device, and hence from the CIA hardware timers that still run at ~700Mhz. Hence, no problems to be expected.
Otherwise, what would a floppy do with a 68060@50Mhz? (-:
Here is where I got that information from. Memory's not what it used to be, but I knew I wasn't making it up. ;)
http://aminet.net/package/docs/hard/14MhzA500
trackstep.doc
[FONT="]If you have performaed the 14Mhz modification and find that you have disk
errors, it is probably because of the track stepping speed having been
altered.
All of the disk read/write operations aren't directly controlled
by the 68000 cpu so our sppedup mod shouldn't affect this. However the
track stepping on the drives is controlled directly by the CPU.
Normally the trackdisk.device uses a value of 0BB8 as a count between it
when it should send step pulses to the drives. At high speed the time taken
for the count will diminish, in some cases to a point where the step pulses
are sent to the drive quicker than it's track to track access time which
could be around 20ms.While using an 80track 5 1/4" drive with a 5ms track
time there is no problem whatsoever, however with the internal drive on some
machines they cannot keep up with the cpu.
[/FONT]
I take no credit for the reliability of this (nor blame for any lack thereof), just copying & pasting. ;)
TL;DR. Nearly all accelerators contain separate oscillators and logic for the processors anyway, so they're independent of any system timings. The 68060@50Mhz example is irrelevant. :lol:
-
Here is where I got that information from. Memory's not what it used to be, but I knew I wasn't making it up. ;)
http://aminet.net/package/docs/hard/14MhzA500
trackstep.doc
I take no credit for the reliability of this (nor blame for any lack thereof), just copying & pasting. ;)
This information is not quite accurate. Actually, the trackdisk device
TL;DR. Nearly all accelerators contain separate oscillators and logic for the processors anyway, so they're independent of any system timings. The 68060@50Mhz example is irrelevant. :lol:
And the 14MHz hack does what? It surely doesn't clock the system components, i.e. the CIAs with twice the regular speed as this wouldn't work. There is no difference between a turbo board with its own oscillator, and a 14MHz 68K that derives its clock from the internal clock rate. In both cases, the CPU runs at a higher than nominal rate, and in both cases all other system components (necessarily) keep running at their nominal speed.
-
IIRC there were some pretty ugly 14MHz hacks back in the day.
Later I seem to remember there popping up some proper ones that synced down to 7MHz when bus access was needed.
I would go for a 68010 if already trying out 14MHz to get those few extra percent more, at least if you have fastram.
-
IIRC there were some pretty ugly 14MHz hacks back in the day.
Later I seem to remember there popping up some proper ones that synced down to 7MHz when bus access was needed.
I would go for a 68010 if already trying out 14MHz to get those few extra percent more, at least if you have fastram.
The 68K bus is asynchron anyhow, i.e. the CPU waits until the bus signals that it is available/data is available. There is no bus-clock on the 68K, quite unlike the 6800 or 6502 bus which is a synchronous bus and hence bus speed depends on the clock rate.
The only synchronous part is the E-clock by which the CIAs are driven, and the protocol around the E-clock. However, if you clock the CIAs by twice the rate, all the timing goes wrong, not only trackdisk.
-
Well, I think the first ones did indeed upset the CIAs.
Something like lift up a pin and solder it to a 14MHz clock or 28MHz with a divide by 2 chip inbetween.
-
This information is not quite accurate. Actually, the trackdisk device
Thanks for helping clear this up!
IMHO I hope OP does do this hack. I remember lots of people talking about it back in the day, but I can't remember anyone actually doing it. Good luck with it, hope you don't blow up your Miggy! :banana:
-
IIRC there were some pretty ugly 14MHz hacks back in the day.
Later I seem to remember there popping up some proper ones that synced down to 7MHz when bus access was needed.
I would go for a 68010 if already trying out 14MHz to get those few extra percent more, at least if you have fastram.
... Well, I think the first ones did indeed upset the CIAs.
Something like lift up a pin and solder it to a 14MHz clock or 28MHz with a divide by 2 chip inbetween.
The 68K bus is asynchron anyhow, i.e. the CPU waits until the bus signals that it is available/data is available. There is no bus-clock on the 68K, quite unlike the 6800 or 6502 bus which is a synchronous bus and hence bus speed depends on the clock rate.
The only synchronous part is the E-clock by which the CIAs are driven, and the protocol around the E-clock. However, if you clock the CIAs by twice the rate, all the timing goes wrong, not only trackdisk.
Thanks for everyone's input, yeah I notice the synchronous e-clock which made me nervous about CIA over clocking (I was worried about damage), I am going to look into how other accelerators run the e-clock asynchronously. If it is as easy as putting the standard 7mhz signal on the e-clock this should be fairly easy to do. If it requires a buffer, things become much more difficult.
What did the Ad-speed do?
I also liked Oldsmobile_Mike and Northway's idea of using a 68010, though I would probably try a 68000 first. One should always only change one variable at a time when experimenting.
-
One should always only change one variable at a time when experimenting.
That, sir, is not the Amiga way. :roflmao: :roflmao:
-
If you look at 68000 instruction timings:
http://oldwww.nvg.ntnu.no/amiga/MC680x0_Sections/mc68000timing.HTML (http://oldwww.nvg.ntnu.no/amiga/MC680x0_Sections/mc68000timing.HTML)
You'll see that there are plenty of instructions which will complete in the equivalent of half the number of clocks with a 14 MHz 68000. Of course, all memory accesses continue to run at 7 MHz, so on a CPU which needs to load all instructions from memory, the speed is largely limited.
The 68010 goes well with the 14 MHz speedup because it has a special loop mode where a loop which consists of two instructions (three words total) can be run without refetching those instructions, and therefore can run at the full 14 MHz.
-
That, sir, is not the Amiga way. :roflmao: :roflmao:
:rofl:
-
As an other mentioned, get a 68010. It is a direct drop i and works great! I did it to my A1000 in the past, as well as an A500 for a friend. You find them on ebay most of the time for near nothing.
-
That, sir, is not the Amiga way. :roflmao: :roflmao:
I resemble that remark :lol:
-
So far everyone seems to have missed the boat on this topic! :laugh1:
The biggest problem with this hack is not the 2x CPU speed, rather it's the 2x E clock speed.
The E clock is generated by the 68000 CPU and supplied to the CIA chips for a timing reference.
The easiest fix would appear to be to divide the E clock by 2. But there is still some inaccuracy with this approach in that the E clock is not a 50/50 duty cycle clock, rather it's a 40/60 duty cycle clock.
The second but smaller problem is that there should be one 14 MHz wait state added at cycle termination to meet the 7 MHz 68000 timing specs. But with some moderate skew on the 14 MHz clock and some extra timing margin in on-board devices you might get by OK without the extra wait state. ;)
Now, when you consider the time, effort and extra logic needed to do it right is it then really worth all the effort?
-
Now, when you consider the time, effort and extra logic needed to do it right is it then really worth all the effort?
Nope! If you're talking legacy hardware on the cheap, add an 010, AdSpeed, SupraTurbo28 or get an 030+ to use in the Fast Slot in your A2000 and be done with it I say! ;)
-
Nope! If you're talking legacy hardware on the cheap, add an 010, AdSpeed, Supra28 or get an 030+ to use in the Fast Slot in your A2000 and be done with it I say! ;)
But is it compatible with old floppy games?
-
But is it compatible with old floppy games?
There's this little thing called WHDLoad... :rolleyes:
-
Adding fast RAM will take you to 1.0 mips. If overclocking gives you a 10% speed increase. That will take you to 1.1 mips.
-
But is it compatible with old floppy games?
As mentioned, can always use WHDLoad as it "fixes" any incompatibilities, but if you're talking about the 68010 or the other 68k accelerator options, I've personally seen very few programs or games that don't like 'em. The 28mhz 68000 SupraTurbo in my A2000 has yet to balk at anything I've thrown at it. Best accelerator for the money IMO.
-
If you are into educational torture, this might be the project for you! ;)
I tried this hack back in the early 90's for an A500. I thought there was only 1 version of the hack. The most memorable point was the continuing system crashes, I don't remember getting the system to run for more than a minute or 2 at best when jumpered to 14 MHz. As I was preparing to build this, I was never able to find a 68010 that was rated for 14 MHz.
As others have suggested, a 68010 at the normal clock rate, coupled with some true FAST ram is in every respect a better option.
I agree with Save2600 that the Supra Turbo28 is far and away the best accelerator for an Amiga with 16-bit memory. I have 2 of them! Come to think of it, I think I traded my AdSpeed to him for an A2088. Was that you? :)
I was sorely disappointed with the AdSpeed, too. I didn't get even close to 2x, and while it wouldn't crash if it was alone in the system, it was incompatible with every expansion I had! ICD was infamous for selling hardware that would not play nice with stuff from other makers.
-
I was sorely disappointed with the AdSpeed, too. I didn't get even close to 2x, and while it wouldn't crash if it was alone in the system, it was incompatible with every expansion I had! ICD was infamous for selling hardware that would not play nice with stuff from other makers.
Boy did I lust after the AdSpeed in magazine adverts at the time. I don't think it was ever good for 2x speed, maybe closer to 30% because it had the onboard cache? That incompatibility though! :(
-
Boy did I lust after the AdSpeed in magazine adverts at the time. I don't think it was ever good for 2x speed, maybe closer to 30% because it had the onboard cache? That incompatibility though! :(
I also had the Adspeed. I added the optional switch so you could "switch it on the fly". On some programs the difference was hardly noticeable, but I remember some of the scenery generating programs really picked up speed and some flight simulators ( I think A-10 Tank killer, Wings, and others) increased their frame rates to a usable level.
-
Nope! If you're talking legacy hardware on the cheap, add an 010, AdSpeed, SupraTurbo28 or get an 030+ to use in the Fast Slot in your A2000 and be done with it I say! ;)
If you are into educational torture, this might be the project for you! ;)
I tried this hack back in the early 90's for an A500. I thought there was only 1 version of the hack. The most memorable point was the continuing system crashes, I don't remember getting the system to run for more than a minute or 2 at best when jumpered to 14 MHz. As I was preparing to build this, I was never able to find a 68010 that was rated for 14 MHz.
As others have suggested, a 68010 at the normal clock rate, coupled with some true FAST ram is in every respect a better option.
I agree with Save2600 that the Supra Turbo28 is far and away the best accelerator for an Amiga with 16-bit memory. I have 2 of them! Come to think of it, I think I traded my AdSpeed to him for an A2088. Was that you? :)
I was sorely disappointed with the AdSpeed, too. I didn't get even close to 2x, and while it wouldn't crash if it was alone in the system, it was incompatible with every expansion I had! ICD was infamous for selling hardware that would not play nice with stuff from other makers.
I certainly understand what you are getting at. Like many of you "back in the day" I wanted an AdSpeed or Supra Turbo28 or a mega-midget racer but could not afford them. Now, I already own several big box amigas with 030a and 040s and an A500+ with ACA500/ACA1233-40.
I don't really "need" the hack. It is more for the fun of it or the nostalgia of trying. It was really fun to hack the amiga 500, whether is was to "upgrade" to 1 meg of chip ram, making a homemade audio digitizer, a homemade VBS unit, make a Zorro II adapter for the side expansion, or parnet cables; the amiga was fun (and reasonably easy) to hack.
I looked at deeper into the aminet 14mhz hack docs the following looks the most complete:
http://aminet.net/package/docs/hard/new14acc_31
It includes all necessary timing including the e-clock.
Further it discusses the latching speed. This would avoid a wait-state.
note: the author says this is shareware. I would agree with this as it looks like he did a bit of work getting the 14mhz "hack" right. :)