A German user here ???

[Arachide]

Hello,

I need some confirmation about a german document about the Hades.

Here is the original text:

*******
Hardware

Allgemein

Der Hades wird normalerweise mit der Bustaktfrequenz von 32 MHz
ausgeliefert, ausser beim MC68060RC50, dort ist sie 30 MHz.
Bei den Transferraten sind jeweils die ben”tigten Bustakte angegeben.

Prozessor

Als Prozessoren lassen sich der MC68040 oder der MC68060 einsetzen.
Beide Prozessoren haben als Prozessorclock den doppelten Bustakt.
*******

Here is a summary of what I understand:

If the Hades has a MC68060RC50, then the bus frequency is 30MHz.
Else, it is 32MHz.

The Processor frequency is the double of the bus frequency:
so the Hades is clocked at either 60 or 64MHz depending on its processor.

Am I right???

Thanks a lot.

Guillaume.

Link: http://www.medusacomputer.com/program/Hadessh.zip

the text is taken from the file DOKU0101.TXT.

[Dennis Schulmeister]

On Sun, 29 Nov 2015 16:57:21 +0100

Post by Arachide
Am I right???

Yes. The text also says that the processor can either be MC68040 or
MC68060.

Greetings from Germany. :-)

Dennis

[Arachide]

Post by Dennis Schulmeister
On Sun, 29 Nov 2015 16:57:21 +0100

Post by Arachide
Am I right???

Yes. The text also says that the processor can either be MC68040 or
MC68060.
Greetings from Germany. :-)
Dennis

Many thanks !
Some are still thinking that the Hades is clocked at 120MHz...

Best regards,

Guillaume.

[Francois LE COAT]

Hi,

Post by Arachide

Post by Dennis Schulmeister

Post by Arachide
Am I right???

Yes. The text also says that the processor can either be MC68040 or
MC68060.
Greetings from Germany. :-)

Many thanks !
Some are still thinking that the Hades is clocked at 120MHz...

MC68060 processor of my Hades060 is internally clocked at 120MHz.
That means that the Kronos benchmarks gives 120MIPS with 60MHz
external clock of the CPU, because the MC68060 is "superscalar".

Maybe you're making confusion about people you're speaking with.

ATARIstically yours =)

[Michael Schwingen]

Post by Francois LE COAT
MC68060 processor of my Hades060 is internally clocked at 120MHz.

No, it is not. The MC68060 is superscalar (for integer, not floating-point),
meaning it executes up to two integer instructions per clock cycle (in
certain conditions). That means given the right instructions, it runs
(nearly) *as fast as* a theoretical singlescalar CPU which runs at 120MHz,
but the internal logic of the MC68060 is nevertheless only clocked at 60MHz.
Also, with worst-case code, it falls back to singlescalar operation (see
section 10.1 in the datasheet).

The datasheet is quite clear in stating that internal logic is clocked at
"clk" pin speed, not 2*"clk".

cu
Michael

[Francois LE COAT]

Hi,

Post by Michael Schwingen

Post by Francois LE COAT
MC68060 processor of my Hades060 is internally clocked at 120MHz.

No, it is not. The MC68060 is superscalar (for integer, not floating-point),
meaning it executes up to two integer instructions per clock cycle (in
certain conditions). That means given the right instructions, it runs
(nearly) *as fast as* a theoretical singlescalar CPU which runs at 120MHz,
but the internal logic of the MC68060 is nevertheless only clocked at 60MHz.
Also, with worst-case code, it falls back to singlescalar operation (see
section 10.1 in the datasheet).
The datasheet is quite clear in stating that internal logic is clocked at
"clk" pin speed, not 2*"clk".

I won't argue. It's written 120MHz on the face side of the computer.
This has a meaning, because with the same external clock frequency,
60MHz and with a MC68040 CPU, it should be written 60MHz on the Hades040

Well, but I have a Hades060 machine since about twenty years now. I made
benchmarks, and it is fast as it was announced on its face side !

I'm not arguing about the Motorola's documentation, but real hardware.

ATARIstically yours,

[Michael Schwingen]

Post by Francois LE COAT
I won't argue. It's written 120MHz on the face side of the computer.

That is then plainly wrong.

Post by Francois LE COAT
This has a meaning, because with the same external clock frequency,
60MHz and with a MC68040 CPU, it should be written 60MHz on the Hades040
Well, but I have a Hades060 machine since about twenty years now. I made
benchmarks, and it is fast as it was announced on its face side !

It may be run as fast as a 68040 clocked at 120MHz, bit it *is not clocked*
internally at 120MHz - that's the thing about being superscalar.

There is an important difference: a superscalar CPU will only get that
performance some of the time - it is not always 2* speed, it depends on the
mix of instructions - only some can be dispatched to the second execution
pipeline.

Post by Francois LE COAT
I'm not arguing about the Motorola's documentation, but real hardware.

This *is* the documentation for the real hardware - you won't get it any
more exact than from the manufacturer of the CPU.

cu
Michael

[Francois LE COAT]

Hi,

Post by Michael Schwingen

Post by Francois LE COAT
I won't argue. It's written 120MHz on the face side of the computer.

That is then plainly wrong.

There's cabled green LEDs displaying "120" for 120MHz on the face side.
It was shipped like this by Medusa Systems, and the advertisements for
the computer said the Hades060 was clocked at 120MHz:

<Loading Image...>

Do you intend to say that Fredi Aschwanden was a liar shipping very
expensive hardware, with false characteristics ? I don't think so.

Post by Michael Schwingen

Post by Francois LE COAT
This has a meaning, because with the same external clock frequency,
60MHz and with a MC68040 CPU, it should be written 60MHz on the Hades040
Well, but I have a Hades060 machine since about twenty years now. I made
benchmarks, and it is fast as it was announced on its face side !

It may be run as fast as a 68040 clocked at 120MHz, bit it *is not clocked*
internally at 120MHz - that's the thing about being superscalar.

Yes, I know. MC68060 has a double pipeline handling two consecutive
instructions at the same time, what is doubling the execution rate.

Post by Michael Schwingen
There is an important difference: a superscalar CPU will only get that
performance some of the time - it is not always 2* speed, it depends on the
mix of instructions - only some can be dispatched to the second execution
pipeline.

Two instructions are executed in parallel at the same time with the
double pipeline, considering that these instructions are independent.
The full speed can only be obtained in this case, otherwise the
pipeline is broken. This is the responsibility of the compiler to
dispatch independent consecutive instructions to obtain full speed.

Post by Michael Schwingen

Post by Francois LE COAT
I'm not arguing about the Motorola's documentation, but real hardware.

This *is* the documentation for the real hardware - you won't get it any
more exact than from the manufacturer of the CPU.

The documentation is free, but the Hades060 cost me 3000Euros in 1996.
That was the first Hades ever sold in France, that launched the market.

ATARIstically yours,

[Michael Schwingen]

Post by Francois LE COAT
<http://download.abandonware.org/magazines/ST%20Magazine/stmagazine_numero121/st%20magazine%20-%20N121%20-%20novembre%201997%20-%20page083.jpg>
Do you intend to say that Fredi Aschwanden was a liar shipping very
expensive hardware, with false characteristics ? I don't think so.

Call it what you like, 120MHz is plain wrong.

Post by Francois LE COAT
Yes, I know. MC68060 has a double pipeline handling two consecutive
instructions at the same time, what is doubling the execution rate.

*up to* two, not always two. Look in the manual ...

Post by Francois LE COAT
Two instructions are executed in parallel at the same time with the
double pipeline, considering that these instructions are independent.
The full speed can only be obtained in this case, otherwise the
pipeline is broken. This is the responsibility of the compiler to
dispatch independent consecutive instructions to obtain full speed.

You can't control legacy code, and even the best compiler can't group
instructions whose results depend on each other.

I'm not arguing about the speed: the 68060 is a nice CPU (I still have my
hand-soldered Milan 060). The MC68060 may perform like 120MHz (with some
code), but it is most definitely not *clocked* at 120MHz internally.

cu
Michael

[Arachide]

Post by Michael Schwingen

Post by Francois LE COAT
<http://download.abandonware.org/magazines/ST%20Magazine/stmagazine_numero121/st%20magazine%20-%20N121%20-%20novembre%201997%20-%20page083.jpg>
Do you intend to say that Fredi Aschwanden was a liar shipping very
expensive hardware, with false characteristics ? I don't think so.

Call it what you like, 120MHz is plain wrong.

We had the same debate on the french group fr.comp.sys.atari.
Whatever our arguments, François never changed his mind.

The best argument is the one directly from Medusa Systems that wrote to me:

********
Hello

I have checked.
Bus clk 32MHz
CPU clk 64MHz x 2 integer unit = 128MHz integer performence
********

(the bus is 30 or 32 MHz according to some CPU versions)

So the 120 (or 128MHz) is only a calculation of a supposed performance
in the best case. This is not a clock nor a mysterious internal frequency.

But François goes on with "I have a 120MHz CPU".

Guillaume.

[calimero]

Post by Arachide

Post by Michael Schwingen

Post by Francois LE COAT
<http://download.abandonware.org/magazines/ST%20Magazine/stmagazine_numero121/st%20magazine%20-%20N121%20-%20novembre%201997%20-%20page083.jpg>
Do you intend to say that Fredi Aschwanden was a liar shipping very
expensive hardware, with false characteristics ? I don't think so.

Call it what you like, 120MHz is plain wrong.

We had the same debate on the french group fr.comp.sys.atari.
Whatever our arguments, François never changed his mind.
********
Hello
I have checked.
Bus clk 32MHz
CPU clk 64MHz x 2 integer unit = 128MHz integer performence
********
(the bus is 30 or 32 MHz according to some CPU versions)
So the 120 (or 128MHz) is only a calculation of a supposed performance
in the best case. This is not a clock nor a mysterious internal frequency.
But François goes on with "I have a 120MHz CPU".

nice debate :) let me add some more fuel :)

please take a look at wikipedia Talk Page about 80486 - there was claim from Apple that 68040 in Macintosh Performa are clocked at 80MHz! LINK: https://en.wikipedia.org/wiki/Talk:Intel_80486#040-based_Macintosh

You will also find explanation for such claim in Rodolphe Czuba article linked at Wikipedia Talk page...

[Arachide]

Post by calimero
nice debate :) let me add some more fuel :)
please take a look at wikipedia Talk Page about 80486 -
there was claim from Apple that 68040 in Macintosh Performa are
https://en.wikipedia.org/wiki/Talk:Intel_80486#040-based_Macintosh

Very interesting!

According to the Freescale PDF documentation about de MC68040, there are
two cloks:

BCLK: the clock for the bus operations.
PCLK; the clock for the processor operations

And it is said tha PCLK is twice BCLK, but in the sense that the
designer of the mother board must provide two inputs clocks with one
running at twice the speed of the other.

The docs then explains that the processor can be clocked at 25, 33 or 40
MHz for PCLK. So BCLK must be 12,5 or 16,5 or 20MHz.

There is no clock doubling.

And the overclocking is hard: I have a 33MHz CPU in my Quadra and an
accelerator that runs fine until 39MHz but then the system hangs at
40..! 33 to 39 is +18%. Not more.

Guillaume.

[Michael Schwingen]

Post by Arachide
According to the Freescale PDF documentation about de MC68040, there are
BCLK: the clock for the bus operations.
PCLK; the clock for the processor operations
And it is said tha PCLK is twice BCLK, but in the sense that the
designer of the mother board must provide two inputs clocks with one
running at twice the speed of the other.
The docs then explains that the processor can be clocked at 25, 33 or 40
MHz for PCLK. So BCLK must be 12,5 or 16,5 or 20MHz.

Exactly. Basically, that means the bus can run at either the same as the
internal clock or half the internal clock. On the Milan, we used the
half-speed option (not sure what the hades did), since that nicely fits with
60ns EDO memory - running the bus faster than the memory yields no
improvement. When using SDRAM (which came available later), running the bus
at 60MHz works fine (but I doubt the performance gain will be high, due to
the sufficiently large caches of the '060).

Post by Arachide
And the overclocking is hard: I have a 33MHz CPU in my Quadra and an
accelerator that runs fine until 39MHz but then the system hangs at
40..! 33 to 39 is +18%. Not more.

Yup - I have the impression the later chips ('040 and '060) are closer to
the specified limits than the earlier ones. I was able to run a 68HC000-16
stable at 36MHz (with zero-waitstate fast RAM). That beast was
significantly faster on real work (compiling using PureC) than a 33MHz 68030
with only ST-RAM and external caches ...

cu
Michael

[Francois LE COAT]

Hi,

Post by Arachide

Post by Michael Schwingen

Post by Francois LE COAT
<http://download.abandonware.org/magazines/ST%20Magazine/stmagazine_numero121/st%20magazine%20-%20N121%20-%20novembre%201997%20-%20page083.jpg>
Do you intend to say that Fredi Aschwanden was a liar shipping very
expensive hardware, with false characteristics ? I don't think so.

Call it what you like, 120MHz is plain wrong.

We had the same debate on the french group fr.comp.sys.atari.
Whatever our arguments, François never changed his mind.
********
Hello
I have checked.
Bus clk 32MHz
CPU clk 64MHz x 2 integer unit = 128MHz integer performence
********
(the bus is 30 or 32 MHz according to some CPU versions)
So the 120 (or 128MHz) is only a calculation of a supposed performance
in the best case. This is not a clock nor a mysterious internal frequency.
But François goes on with "I have a 120MHz CPU".

It is written 120MHz in the advert, and my Hades060 displays "120" for
120MHz on the face side. I can't rewrite the history like it was in
1996, when I bought this machine. If you tell me that the Hades060 is
clocked at 60MHz, I can admit this. But it wasn't announced like this
20 years ago. It wasn't essentially a simple mistake. I think the
Hades060 was competing contemporary Pentium PCs. You can't say your
Falcon CT060 competes any other machine twenty years after. Jealous !

I don't regret anything buying my Hades060, in 1996 when I got it :-)

ATARIstically yours =)

[Miro Kropáček]

Post by Francois LE COAT

Post by Arachide
But François goes on with "I have a 120MHz CPU".

It is written 120MHz in the advert, and my Hades060 displays "120" for
120MHz on the face side.

It must be true then! I guess you believe your TV in similar way. ;-)

Post by Francois LE COAT
But it wasn't announced like this 20 years ago. It wasn't essentially a simple mistake.

The sad truth is that Atari market, like any other market, was filled with not-so-truthful claims. I remember the AB040 advertised with the BCLK clock (claiming that the AB040 is capable of running at 64 MHz), a proof that it's total BS was part of CT2 documents (which was, in retrospective, also advertised with full of lies, "easy installation" anyone? ;-)).

[Tonton Th]

Post by Francois LE COAT

Post by Arachide

Post by Dennis Schulmeister
Yes. The text also says that the processor can either be MC68040 or
MC68060.

Many thanks !
Some are still thinking that the Hades is clocked at 120MHz...

MC68060 processor of my Hades060 is internally clocked at 120MHz.

Can you give a link to the datasheet who say "120 MHz" ?