Fun things to do with an MC68EC020....

Nagging hardware related question? Post here!
User avatar
Peter
QL Wafer Drive
Posts: 1953
Joined: Sat Jan 22, 2011 8:47 am

Re: Fun things to do with an MC68EC020....

Post by Peter »

tofro wrote:The 68SEC000 20MHz version can BTW easily be overclocked to 50MHz - That would be close to achieving the 68020's speed, with a CPU that is much easier to obtain.
Very interesting remark :D

I was often sceptical when 68K overclockers in the Atari/Amiga scene said that a system runs stable, because it depends a lot on the code the CPU executes. E.g. I could overclock the Q60 to 100 MHz and under SMSQ/E everything looked fine. Different story to compile a kernel or GCC from source under Linux.

But in this case I would indeed expect no problem. I just vaguely remember there was a die shrink of the 68SEC000 end of the nineties, and you should use the newer part.

Unfortunately the bus sizing of the 68SEC000 is static.


Nasta
Gold Card
Posts: 443
Joined: Sun Feb 12, 2012 2:02 am
Location: Zapresic, Croatia

Re: Fun things to do with an MC68EC020....

Post by Nasta »

tofro wrote:I really don't see things becoming much easier when using an 8-bit bus only. Only slower ;)
The 68SEC000 20MHz version can BTW easily be overclocked to 50MHz - That would be close to achieving the 68020's speed, with a CPU that is much easier to obtain.
Tobias
Well, that has been counted on :P
Even at 20MHz with an 8-bit bus it should work around GC speed, if a decent memory controller is implemented. After that, upping the clock has diminishing returns because there is no cache. But it would be a nice improvement on the BB QL.
It is true that a very similar design of the memory controller and decoder can be used for a 32-bit incarnation of the hardware, but there is a fair bit more routing to do on a PCB.

@Peter,
There is a clue that it is overclockable on the outset, as it's a 3-5V part and is specced up to full speed at 3V. So... what happens at 5V? :)


User avatar
Dave
SandySuperQDave
Posts: 2765
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Fun things to do with an MC68EC020....

Post by Dave »

tofro wrote:I really don't see things becoming much easier when using an 8-bit bus only. Only slower ;)

The 68SEC000 20MHz version can BTW easily be overclocked to 50MHz - That would be close to achieving the 68020's speed, with a CPU that is much easier to obtain.

Tobias
I can get the 68EC020 in bulk for $3.33, Freescale '12+ manufacture. The 68SEC000FU20 is $2.55. The FU batch has very poor overclockability. The AF and AL batches I have access to are around $7 -- more than the 020, but more overclockable. The CPU is such a small proportion of the cost of the whole system that cpu component price isn't really a consideration.


User avatar
Peter
QL Wafer Drive
Posts: 1953
Joined: Sat Jan 22, 2011 8:47 am

Re: Fun things to do with an MC68EC020....

Post by Peter »

Nasta wrote:Even at 20MHz with an 8-bit bus it should work around GC speed, if a decent memory controller is implemented.
At first sight this looks wrong to me. I have not actually measured timing on the GC, but it has 80 ns RAS access time DRAM chips, so I would guess DRAM cycle time around 150 ns, without using page mode. If we assume the real cycle time 30% slower, the GC would be under 200 ns cycle time for 16 Bit.

Because of the 8 Bit bus, the 68SEC000 would need 100 ns cycle time, which is two 20 MHz clock cycles, to break even. But it has minimum 4 clock cycles!
Nasta wrote:After that, upping the clock has diminishing returns because there is no cache.
I would say: After that, upping the clock starts to make real sense!

Dave wants to use SRAM, so he can achieve the 100 ns cycle times needed for 40 MHz without waitstates. Even 80 ns for 50 MHz if he spends more for speedgrade.
Indeed I would find some speed more important for a machine like this, than over 2 or 4 MB RAM size.
Nasta wrote:@Peter, There is a clue that it is overclockable on the outset, as it's a 3-5V part and is specced up to full speed at 3V. So... what happens at 5V? :)
Hehe. Good hint.


User avatar
tofro
Font of All Knowledge
Posts: 2685
Joined: Sun Feb 13, 2011 10:53 pm
Location: SW Germany

Re: Fun things to do with an MC68EC020....

Post by tofro »

From a software point-of-view:

I'd guess many QL users will struggle to even make proper use of a Trump Card's full 786k of memory.

Typical QL programs that need and will make use of more than about 1M of memory are nearly exclusively PE programs. PE programs also need good performance, especially to and fro the screen memory, and they need proper mass storage to get the stuff loaded from.

The classic QL tends to be started and run in "sessions" - 1 program, one boot disk specialised for a specific task, and re-booted to another boot scenario when done. Many QL users have never gone beyond this point. Once you have beyond 2M and mass storage, you tend to set up more general purpose sessions like in modern PCs. Below 2M, "General purpose" is difficult because of memory size.

My opinion: Finding the right balance between CPU speed, memory performance and size and proper mass storage might be important, but difficult. A QL close (up to ~twice) to the original's speed doesn't need more than ~2M of memory - beyond that, it becomes boring to "fill" the memory. Beyond 4M, you should have at least 4-10 times original QL CPU performance. And for SMSQ/E to feel happy, you should have at least 4M.

In my opinion, having a large amount of memory and a relatively slow CPU is just as useless as the other way round - Both needs to be improved in parallel.

Tobias


ʎɐqǝ ɯoɹɟ ǝq oʇ ƃuᴉoƃ ʇou sᴉ pɹɐoqʎǝʞ ʇxǝu ʎɯ 'ɹɐǝp ɥO
User avatar
Peter
QL Wafer Drive
Posts: 1953
Joined: Sat Jan 22, 2011 8:47 am

Re: Fun things to do with an MC68EC020....

Post by Peter »

tofro wrote:And for SMSQ/E to feel happy, you should have at least 4M.
(Only) at this point I tend to disagree, because this system has no own graphics hardware. It is not so much usage of SMSQ/E that would call for 4 MB, but usage of highres+highcolor programs, which are ruled out here.

I even find SMSQ/E the best choice for a slow 2 MB system, because of the much faster Basic.

With RAM of 1 MB, admittedly, the length of the OS itself becomes quite relevant.


User avatar
tofro
Font of All Knowledge
Posts: 2685
Joined: Sun Feb 13, 2011 10:53 pm
Location: SW Germany

Re: Fun things to do with an MC68EC020....

Post by tofro »

Peter wrote: I even find SMSQ/E the best choice for a slow 2 MB system, because of the much faster Basic.

With RAM of 1 MB, admittedly, the length of the OS itself becomes quite relevant.
With SMSQ/E in ROM - 2M might be perfect. With SMSQ/E taking RAM - that would be not much of an improvement over a Trump card or clone (Agreed, the Basic is much better) to justify development of new HW. Adding some M of memory actually isn't sooo much work on top given that you add a CPU and most probably a relatively complex CPLD to adapt to the QL bus.

Tobias


ʎɐqǝ ɯoɹɟ ǝq oʇ ƃuᴉoƃ ʇou sᴉ pɹɐoqʎǝʞ ʇxǝu ʎɯ 'ɹɐǝp ɥO
User avatar
Peter
QL Wafer Drive
Posts: 1953
Joined: Sat Jan 22, 2011 8:47 am

Re: Fun things to do with an MC68EC020....

Post by Peter »

Still I was considering 2 MB, because at the access times needed for overclocking, the additional SRAM for 4 MB instead of 2 MB costs about 9 € extra. Not so adequate with the low CPU price.

The overclocked 68SEC000 allows Gold Card speed with 8 bit bus, so that could at least overcome GoldCards being expensive, not made anymore, battery life, etc.

Wether development of new processor HW for QL ist justified at all, is of course a different question ;)


User avatar
Dave
SandySuperQDave
Posts: 2765
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Fun things to do with an MC68EC020....

Post by Dave »

Peter wrote:Dave wants to use SRAM, so he can achieve the 100 ns cycle times needed for 40 MHz without waitstates.
Yes... but :)

I wanted to use SRAM because of the lack of refresh cycles, the speed and the simplicity of design. The catch is the low density of SRAM means this would require a large number of expensive devices, each being a load on the bus.

We've identified some good DRAM chips with 50ns timings that will have similar performance with a much smaller bus footprint, in exchange for designing a not that much more complex memory decoder/handler to handle the refresh cycles. There's also a self refresh version of the IC for the same price though I don't know if that would be an advantage.

Two or four 50ns DRAMs is sure to beat 32 70ns SRAMs plus their buffer ICs. And the cost of 16MB DRAM is $5. SRAM is $96.

So my desire for SRAM is dead on every count.


User avatar
Dave
SandySuperQDave
Posts: 2765
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Fun things to do with an MC68EC020....

Post by Dave »

tofro wrote:From a software point-of-view:

(snip)

Tobias
Thank you Tobias. This is the most enlightening and informative post and has given me some clarity and focus.


Post Reply