Amiga 3000 030/16/25 Amiga 1200 020/14..

redblade · 28 April 2020, 11:22

Hi.

Can one of you hardware guys please explain to me why the Amiga 3000 was released at 030@16mhz or 25mhz and the A500/600/1200 was clocked at the TV frequency? I don't quite understand it especially as the 68ec020FG16 in the A1200 was down synced to 14mhz compared to the A3000 kept at 16mhz.

Thanks

sneeker · 28 April 2020, 13:11

Knowing commodore, probably just down to cost, the A3000 had a flickerfixer etc as standard being a high end machine, where the wedge amigas are more likely used on a tv therefore cheaper clock it at the tv frequency? Snd they would sell a lot more wedges then big box, so every cent helped that they could save.

roondar · 28 April 2020, 13:37

It was probably cost based, yeah. The A1200 CPU runs synchronous to the internal chip memory bus. That might simply be a cheaper design than clocking the CPU asynchronously and using more complicated bus arbitration. Might even have saved them a crystal.

Do note, I'm not a hardware guy - the above is simply a guess

Kin Hell · 28 April 2020, 14:24

Quote:

Originally Posted by redblade

Hi.

Can one of you hardware guys please explain to me why the Amiga 3000 was released at 030@16mhz or 25mhz and the A500/600/1200 was clocked at the TV frequency? I don't quite understand it especially as the 68ec020FG16 in the A1200 was down synced to 14mhz compared to the A3000 kept at 16mhz.

Thanks

Soak Testing!

This happens to all Fabrications in the Miro-computing world. If they don't run at the rated speed, they are speed binned accordingly. Save's Waste & perpetuates a cheaper alternative/solution.

A1200 uses an MC80020, where as the A3000 uses an MC68030. Different Fabrications & also rated @ different speeds. Don't also forget, Ceramic MC68020 & 030 CPU's were always rated at Higher Clock Speeds than the cheaper QFP variants. '030's only ever went as far as 33Mhz in QFP style.

https://en.wikipedia.org/wiki/Motorola_68030

Jope · 28 April 2020, 16:17

Quote:

Originally Posted by roondar

It was probably cost based, yeah. The A1200 CPU runs synchronous to the internal chip memory bus. That might simply be a cheaper design than clocking the CPU asynchronously and using more complicated bus arbitration. Might even have saved them a crystal.

Basically this. The only Amigas that didn't have all clocks derived from a single crystal were the 3000s and the 4000s. :-)

Kin Hell · 29 April 2020, 11:15

Quote:

Originally Posted by Jope

Basically this. The only Amigas that didn't have all clocks derived from a single crystal were the 3000s and the 4000s. :-)

That Sir, is an absolutely fantastic Analogy!

Lord Aga · 29 April 2020, 12:21

So you're saying they had a Karen running around, deciding where to use the power of crystals?

redblade · 29 April 2020, 22:34

Thanks, so that explains why my Amiga 1200 Accelerators the Viper 030, and Blizzard 1220/68EC020FG25 run at 28mhz.

Do the NTSC/PAL frequencies have much influence on the other screen modes like the EURO72, MULTISCAN and do you get faster write speed to the CHIP bus if you send the screen mode to NTSC while you are copying/manipulating data before you display it?

(On a sad note my Viper 030 memory 8mb memory map is to the PCMCIA memory space and not the other 32bit memory space

So you get what you pay for

Thanks

Jope · 30 April 2020, 09:03

Quote:

Originally Posted by redblade

Thanks, so that explains why my Amiga 1200 Accelerators the Viper 030, and Blizzard 1220/68EC020FG25 run at 28mhz.

Nothing is stopping people making async accelerators for the A1200, but yeah, those two cheaped out and are clocking them off the motherboard.

Quote:

Do the NTSC/PAL frequencies have much influence on the other screen modes like the EURO72, MULTISCAN and do you get faster write speed to the CHIP bus if you send the screen mode to NTSC while you are copying/manipulating data before you display it?

You will get more DMA slots available in NTSC versus a higher refresh rate mode, but swapping screen modes before displaying would not be feasible. A CRT chassis might fail with all the resolution switching and a TFT would probably not sync to anything at all because they take so long to settle when the resolution changes.

Quote:

(On a sad note my Viper 030 memory 8mb memory map is to the PCMCIA memory space and not the other 32bit memory space

So you get what you pay for

Indeed. For some use cases these are totally fine, but unfortunately you can't have it all with these cards.

TroyWilkins · 02 May 2020, 07:18

My understanding is that it's not JUST the cost savings involved with using the one clock signal and multiplying and dividing to get all the other clock signals required.

I could be wrong here, but I believe that's just one of the reasons.

To run a CPU at a speed that isn't exactly a multiple of the speed of things like the chip ram bus, you also need to work out how to get that CPU to work with the bus speeds that it's no longer in sync with. That requires more complicated circuitry. So it's not JUST the cost of a seperate clock crystal for the CPU in addition to the clock crystal for the display and custom chips, but also extra things like buffers.

In a machine where they were trying to minimise cost, such as the A500 and A1200, this would have added far too much extra cost for something that would have seen very little practical benefit. Sure, the 68ec020 in the A1200 may have been capable of running at, lets say 15MHz or even 16MHz, but if the machine only came with the 2MB Chip RAM, that extra speed would not have been usable, as it would have had it's ram bus speed still clocked at the standard 14 MHz - so literally anything it had to do with anything external to the CPU would have had to slow to 14MHz anyway, and then there is the issue of Agnus/Alice taking control of the bus (which is why adding fast RAM speeds up the performance of a machine, as the CPU then has RAM it can access without needing to wait for Agnus/Alice to finish with it), so there wouldn't have been any benefit, but extra costs.

In machines such as the A3000, which was designed to have not just a 16MHz, but also a 25MHz CPU, and where it was sold as a more premium product, the extra complexity and cost was justified by the significant performance boost, and these machines were also designed so that future, faster CPUs could be added.

Lord Aga · 02 May 2020, 10:43

Quote:

Originally Posted by TroyWilkins

Sure, the 68ec020 in the A1200 may have been capable of running at, lets say 15MHz or even 16MHz,

Well, A1200 CPUs ARE, in fact, 16MHz rated CPUs. So they are actually underclocked.

TroyWilkins · 02 May 2020, 11:25

Quote:

Originally Posted by Lord Aga

Well, A1200 CPUs ARE, in fact, 16MHz rated CPUs. So they are actually underclocked.

Yes, but as I've tried to explain above, actually running it at that speed with only Chip RAM would not have given any advantage and would have cost more as well.

Lord Aga · 02 May 2020, 11:35

True dat. Maybe they could have gone with a triple clock and reach 21MHz with a small passive cooler.

Adding a FAST RAM card would give that a tremendous boost

TroyWilkins · 02 May 2020, 12:22

True, but it's easy for us to forget how expensive RAM was back in those days. Even 1MB of true Fast RAM would have made a HUGE difference to the 1200, but it would have also made a huge difference to the price as well. Even adding the extra circuits to add that fast RAM would have caused the price to go up when it came to retail, and lets not forget the rapidly dropping prices of 386 PCs and the like at the time, and there was also competition (however weak or strong it may have been) from Atari with their ST line...

I don't know if 21MHz would have been possible, not only on that CPU, but also with the math involved from the 7MHz base clock, as I could be wrong, but I think it was only possible to do simple multiply or divide on the clock rates.

roondar · 02 May 2020, 14:05

The thing that always made me wonder was the low chip ram bandwidth to begin with. RAM was capable of running much faster by 1992 and it still somewhat baffles me that Commodore didn't make use of that when designing AGA (well, they kind of did with the display fetching but the RAM was still only 140ns, while 60ns was already commonly available).

Kin Hell · 02 May 2020, 18:22

Another weird anomaly....

An A1200 was actually quicker at accessing Fast Ram than an A4000.

redblade · 03 May 2020, 01:05

Quote:

Originally Posted by Kin Hell

Another weird anomaly....

An A1200 was actually quicker at accessing Fast Ram than an A4000.

Really ?? So as Roondar mentioned about 60/140ns RAM, it is slower to access that on the A4000? That sounds crazy

TroyWilkins · 03 May 2020, 01:18

Yes, the fast RAM on an A4000 is slower to access than fast RAM on an A1200. My understanding is that it's down to the design of the A3640 in the 4000/040. The fast RAM access is faster in the 4000/030 with the A3630, but is still limited by the main board in the 4000. That's why so many 3rd party CPU cards for the 3000/4000 had fast RAM on the card along with the CPU and had much faster fast RAM access as a result.

AmigaHope · 03 May 2020, 04:06

Quote:

Originally Posted by TroyWilkins

Yes, the fast RAM on an A4000 is slower to access than fast RAM on an A1200. My understanding is that it's down to the design of the A3640 in the 4000/040. The fast RAM access is faster in the 4000/030 with the A3630, but is still limited by the main board in the 4000. That's why so many 3rd party CPU cards for the 3000/4000 had fast RAM on the card along with the CPU and had much faster fast RAM access as a result.

I'm pretty sure the fastest A4000 motherboard fast RAM was with the Cost Reduced model that had the 68030 on the motherboard.

TroyWilkins · 03 May 2020, 04:07

Quote:

Originally Posted by AmigaHope

I'm pretty sure the fastest A4000 motherboard fast RAM was with the Cost Reduced model that had the 68030 on the motherboard.

Yes, that wouldn't surprise me at all

28 April 2020, 11:22	#1
redblade Zone Friend Join Date: Mar 2004 Location: Middle Earth Age: 40 Posts: 2,130	Amiga 3000 030/16/25 Amiga 1200 020/14.. Hi. Can one of you hardware guys please explain to me why the Amiga 3000 was released at 030@16mhz or 25mhz and the A500/600/1200 was clocked at the TV frequency? I don't quite understand it especially as the 68ec020FG16 in the A1200 was down synced to 14mhz compared to the A3000 kept at 16mhz. Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Amiga 1200 020 + fastram vs blitter	sandruzzo	Coders. Asm / Hardware	5	02 April 2016 13:03
FS:Amiga 1200 with 030 setup.	Megamix	MarketPlace	1	26 March 2015 19:02
020 030 040?	Claw22000	support.Hardware	9	30 April 2011 06:43
020/030 swapable?	Nikolaj_sofus	support.Hardware	6	25 November 2010 14:56
FS: Amiga 1200 with 030, 32mb, SDFF	danhans	MarketPlace	7	11 August 2009 23:15

28 April 2020, 13:11	#2
sneeker Registered User Join Date: Mar 2008 Location: Manchester! Posts: 308	Knowing commodore, probably just down to cost, the A3000 had a flickerfixer etc as standard being a high end machine, where the wedge amigas are more likely used on a tv therefore cheaper clock it at the tv frequency? Snd they would sell a lot more wedges then big box, so every cent helped that they could save.

28 April 2020, 13:37	#3
roondar Registered User Join Date: Jul 2015 Location: The Netherlands Posts: 3,437	It was probably cost based, yeah. The A1200 CPU runs synchronous to the internal chip memory bus. That might simply be a cheaper design than clocking the CPU asynchronously and using more complicated bus arbitration. Might even have saved them a crystal. Do note, I'm not a hardware guy - the above is simply a guess

29 April 2020, 12:21	#7
Lord Aga MI clan prevails Join Date: Jul 2010 Location: Belgrade, Serbia Posts: 1,443	So you're saying they had a Karen running around, deciding where to use the power of crystals?

29 April 2020, 22:34	#8
redblade Zone Friend Join Date: Mar 2004 Location: Middle Earth Age: 40 Posts: 2,130	Thanks, so that explains why my Amiga 1200 Accelerators the Viper 030, and Blizzard 1220/68EC020FG25 run at 28mhz. Do the NTSC/PAL frequencies have much influence on the other screen modes like the EURO72, MULTISCAN and do you get faster write speed to the CHIP bus if you send the screen mode to NTSC while you are copying/manipulating data before you display it? (On a sad note my Viper 030 memory 8mb memory map is to the PCMCIA memory space and not the other 32bit memory space So you get what you pay for Thanks

02 May 2020, 07:18	#10
TroyWilkins Registered User Join Date: Jan 2015 Location: Melbourne, Australia Posts: 548	My understanding is that it's not JUST the cost savings involved with using the one clock signal and multiplying and dividing to get all the other clock signals required. I could be wrong here, but I believe that's just one of the reasons. To run a CPU at a speed that isn't exactly a multiple of the speed of things like the chip ram bus, you also need to work out how to get that CPU to work with the bus speeds that it's no longer in sync with. That requires more complicated circuitry. So it's not JUST the cost of a seperate clock crystal for the CPU in addition to the clock crystal for the display and custom chips, but also extra things like buffers. In a machine where they were trying to minimise cost, such as the A500 and A1200, this would have added far too much extra cost for something that would have seen very little practical benefit. Sure, the 68ec020 in the A1200 may have been capable of running at, lets say 15MHz or even 16MHz, but if the machine only came with the 2MB Chip RAM, that extra speed would not have been usable, as it would have had it's ram bus speed still clocked at the standard 14 MHz - so literally anything it had to do with anything external to the CPU would have had to slow to 14MHz anyway, and then there is the issue of Agnus/Alice taking control of the bus (which is why adding fast RAM speeds up the performance of a machine, as the CPU then has RAM it can access without needing to wait for Agnus/Alice to finish with it), so there wouldn't have been any benefit, but extra costs. In machines such as the A3000, which was designed to have not just a 16MHz, but also a 25MHz CPU, and where it was sold as a more premium product, the extra complexity and cost was justified by the significant performance boost, and these machines were also designed so that future, faster CPUs could be added.

02 May 2020, 11:35	#13
Lord Aga MI clan prevails Join Date: Jul 2010 Location: Belgrade, Serbia Posts: 1,443	True dat. Maybe they could have gone with a triple clock and reach 21MHz with a small passive cooler. Adding a FAST RAM card would give that a tremendous boost

02 May 2020, 12:22	#14
TroyWilkins Registered User Join Date: Jan 2015 Location: Melbourne, Australia Posts: 548	True, but it's easy for us to forget how expensive RAM was back in those days. Even 1MB of true Fast RAM would have made a HUGE difference to the 1200, but it would have also made a huge difference to the price as well. Even adding the extra circuits to add that fast RAM would have caused the price to go up when it came to retail, and lets not forget the rapidly dropping prices of 386 PCs and the like at the time, and there was also competition (however weak or strong it may have been) from Atari with their ST line... I don't know if 21MHz would have been possible, not only on that CPU, but also with the math involved from the 7MHz base clock, as I could be wrong, but I think it was only possible to do simple multiply or divide on the clock rates.

02 May 2020, 14:05	#15
roondar Registered User Join Date: Jul 2015 Location: The Netherlands Posts: 3,437	The thing that always made me wonder was the low chip ram bandwidth to begin with. RAM was capable of running much faster by 1992 and it still somewhat baffles me that Commodore didn't make use of that when designing AGA (well, they kind of did with the display fetching but the RAM was still only 140ns, while 60ns was already commonly available).

02 May 2020, 18:22	#16
Kin Hell 0ld0r Git Join Date: Mar 2009 Location: Cornwall, UK Posts: 1,665	Another weird anomaly.... An A1200 was actually quicker at accessing Fast Ram than an A4000.

03 May 2020, 01:18	#18
TroyWilkins Registered User Join Date: Jan 2015 Location: Melbourne, Australia Posts: 548	Yes, the fast RAM on an A4000 is slower to access than fast RAM on an A1200. My understanding is that it's down to the design of the A3640 in the 4000/040. The fast RAM access is faster in the 4000/030 with the A3630, but is still limited by the main board in the 4000. That's why so many 3rd party CPU cards for the 3000/4000 had fast RAM on the card along with the CPU and had much faster fast RAM access as a result.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)