Back to Atari with a German Mega 4: help with two add-on boards

[1200XL M.U.L.E.]

I've been out of the Atari ecosystem for about 18 months as I worked through a lot of personal health issues. My mind and attentions were scattered. It looks like quite a lot of good things and topics happened here on Atari Age! 

 

Today I received a Mega 4 from Germany that I bought on eBay. It has two add-on board and I am looking for more information on them. Here is picture of the overall system.

 

 

The first board is hard to see so I circled it and marked it as "1". Here is a close-up of it.

 

 

After lots of Google searches, I think this is an accelerator board. Specifically, this is a HBS640. Is that correct? 

 

It is connected to two switches in the back.

 

 

Is there any documentation on this board? What do the two switches do?

 

The second board looks pretty straightforward, a Floating Point Coprocessor Card of some sort.

 

 

I assume there is nothing mystical or magical about this board except that it simply exists. Oh, and you can see the two switches connected to the accelerator board sticking out of the case.

 

The machine has a 220VAC power supply and I am in the US with 120VAC. I can't turn on this machine until I change the power supply. The seller posted a picture of this machine seemingly working with a monochrome display.

 

Any help would be greatly appreciated! Thank you!

[Chri O.]

Nice looking mega 4 with what seems to be at least 16 MHz CPU upgrade. 🙂

[1200XL M.U.L.E.]

I found some documentation on the accelerator board but it is in German. Does anyone know if an English exists somewhere?

 

https://archive.org/details/hbs-640-68000-cpu-mit-16-mhz/mode/2up

 

Looks like the accelerator board is a HBS640 of sorts.

[1200XL M.U.L.E.]

After swapping in a 120VAC power supply for US power, I turned on the machine and ran SYSINFO.PRG. Here's what I saw.

 

 

Wow, a super-fast Mega 4! The machine does feel very snappy and quick compared to my 1040 STe. I need to run some tests to measure the increase in performance and to ensure the machine runs stable at this higher speed. My first thought was to run YAART.PRG but I have no idea what to run for benchmarking. Any thoughts?

 

The two switches in the back toggle the clock and the cache. 

 

I found a configuration program on a site called CD-Text files.

 

http://cd.textfiles.com/atarilibrary/atari_cd03/UNPACKED/LAST_MIN/HBS640/

 

I'm not clear on what this program does. It seems to control the board ... but isn't that what the switches in the back are for? Here is what the application looks like.

 

 

I haven't experimented much with this yet. Ran out of time last night.

 

Does anyone have any experience with this program?

 

Thanks!

[+DarkLord]

Did some searching... From:

 

https://tinyurl.com/annpu4et

 

HBS640/T28 and HBS640/T36

The two turbo cards HBS640/28 & HBS640/36 from Heyer&Neumann represent a technical masterpiece. According to the data sheet from Motorola, the MC68000 CPU was only produced for a maximum of 16 MHz and yet the developers at Heyer&Neumann managed to get up to 36 Mhz out of this processor without endangering the stability of the system.

hardware data
HBS640/28

    68000 16-bit processor
    68881 FPU (can be retrofitted)
    64 KB data and instruction cache
    28MHz clock frequency
    Fallback function to 8 Mhz

HBS640/36

    68000 16-bit processor
    68881 FPU (can be retrofitted)
    64 KB data and instruction cache
    36MHz clock frequency
    Fallback function to 8 Mhz
    Cache RAMs are faster than in the HBS640/28 version.

extensions

    The HBS640/xx can be retrofitted with a numerical coprocessor MC 68881 via 2 post field sockets.
    The integration of the MC 68881 is compatible with the Atari coprocessor card for the Mega ST, so that all programming languages (e.g. GFA-BASIC, OMIKRON.BASIC, Megamax C) with coprocessor support work.

installation

    The original processor is replaced by the HBS640/xx and the blitter should be removed or at least switched off.

    Computers that have an IMP chipset cannot use the HBS640/36 without hardware modification "IMPerator".

system

    The HBS640/xx can be built into all Atari computers that have a 68000 CPU.

performance data
QIndex 2.1 8MHz 28MHz 36MHz
CPU memory 100 219 281
CPU register 100 355 457
CPU divide 100 355 457
CPU shift 100 361 466
average
Quick Index 100 322 415

 

And a couple of files:

 

HBS640.ZIP

 

hbs640.tgz

 

Don't know how much help this will be but... 

 

[Zogging Hell]

16 hours ago, DarkLord said:

HBS640/28

    68000 16-bit processor
    68881 FPU (can be retrofitted)
    64 KB data and instruction cache
    28MHz clock frequency
    Fallback function to 8 Mhz

HBS640/36

    68000 16-bit processor
    68881 FPU (can be retrofitted)
    64 KB data and instruction cache
    36MHz clock frequency
    Fallback function to 8 Mhz
    Cache RAMs are faster than in the HBS640/28 version.

Wow, nice upgrade that! Wouldn't mind one myself I do remember these being advertised (I think System Solutions sold them in the UK), didn't know they had FPU support and cache as well.

[1200XL M.U.L.E.]

I ran a program called Gembench. Here are its results.

 

 

I was able to run this using a mouse connected to my NetUSBee.

 

I was really surprised at how fast and how slow this machine is at the same time.

 

I am also surprised the program says no FPU is present despite an Atari Corp branded board being installed in this computer. Do I need to run some driver to enable this board?

 

My mouse dies at the end of the benchmark and I'm not sure why. The red LED on the underside of the mouse is no longer illuminated. I am forced to reset the machine. Any ideas what could be causing this?

 

Thanks!

 

EDIT: Actually, the mouse LED does not turn off each time. Two times in a row does not make a case, I suppose. I think the program simply hangs.

Edited September 16, 2022 by 1200XL M.U.L.E.
Update to crashing behaviour

[+DarkLord]

Those values look a bit wonky to me...

 

[1200XL M.U.L.E.]

22 minutes ago, DarkLord said:

Those values look a bit wonky to me...

 

I agree, mostly.

 

I remember reading somewhere, maybe the Atari-Forum site, that running the NetUSBee drivers invoke a noticeable performance hit. The hit comes from the drivers constantly polling the NetUSBee device and the USB ports. I can see that adversely affecting the numbers. This Mega 4 machine also does not have a Blitter. That could also slow down some of the graphics. I also wouldn't be surprised if the RAM and ROM devices themselves are faster on the STe by nature. 

 

I ran the same program on my 1040 STe and got results matching the reference.

 

 

 

[+DarkLord]

To the OP, I distinctly remember reading some HBS640 stuff at either Atari Forum or Exxos's Forums.

 

(or maybe both)

 

You might want to post there and see if someone there has one or knows more about it. Just a random

thought... 

 

[1200XL M.U.L.E.]

I ran the CoreMark benchmark program. Here is the link to the Github repo.

 

https://github.com/czietz/coremark

 

First, I ran it on my 1040 STe.

 

 

Execution time was 15.5 seconds for 30 iterations. My result of 1.9354 comes close to their measurement of 1.92.

 

Second, I ran it my Mega 4 with the NetUSBee drivers.

 

 

Execution time was slightly longer at 18.5 seconds but the benchmark got 110 iterations in. The result here is 5.9459. That's an improvement of 3.07x!

 

I wanted to run this benchmark without any of the NetUSBee drivers. Since I won't have a mouse available to me, I put the program in the AUTO folder of my boot disk. Here's what I got.

 

 

Same number of iterations but slightly less time. Thay boosted the score to 6.0723, which is about 2.1% better.

 

From this test, I think we begin to say the NetUSBee drivers have a minimal impact on performance. However, I understand this is a very CPU-focused benchmark. There could be other system wide effects from the driver.

 

All in good fun.  

[Paul Westphal]

For comparision, this is my Mega2 with ICD Adspeed / 4mb ram / Eiffel 3 :

[1200XL M.U.L.E.]

@Paul Westphal Very cool! Thanks for posting this!

 

I noticed you have a Blitter in your system. Mine doesn't and that probably hurts my performance metric.

 

[1200XL M.U.L.E.]

A couple of interesting updates to my exploration of the performance gains achievable with the HBS 640!

 

You may have noticed that I am running my Mega 4 with a monochrome display (i.e., VGA monitor). I have a color SC1224 monitor connected to my 1040 ST^E but it is very inconvenient to move the connection to my Mega 4. A few days ago, I decided to move the connection temporarily just to see the Mega 4 run in color. Cables were dangling every which way. It only had to last for a few minutes.

 

I was shocked to hear what sounded like a stuck key! The sound was probably there the whole I have been playing with my Mega 4. Since my monochrome display doesn't have a built-in speaker, I would never have heard it. 

 

The first thought that jumped into my head was the Mega 4 is wasting precious clock cycles, interrupts, and CPU power to do this. Would this have any effect on the benchmark scores I am seeing? 

 

The next day I received an Eiffel kit from @snarkdluG (thank you, thank you!) and got it working with an IBM Model M keyboard + cheapie PS/2 mouse. The stuck key clicking sound was gone and the first thing I did was run GEM Bench.

 

 

All the numbers are a bit higher than my previous run using my NetUSBee. (should link to a message above).

 

I also reran Coremark and saw a jump compared to running it without a keyboard connected.

 

 

I'm very happy with this jump in performance and the increased stability of the Eiffel. The NetUSBee is probably very stable by itself and maybe with other accelerators. It just doesn't seem to play well with my Mega and its accelerator. 🤷‍♂️

 

I haven't focused any attention in this thread on the FPU board inside my Mega 4 and it is probably time I do! Yesterday I found a seller on eBay who posted what looks like a complete Atari FPU kit. The kit seems to include a floppy disk! Here's a crop from a picture in their auction listing.

 

 

The auction # is 115538365001.

 

That disk looks pretty official. Does anyone have a copy of the contents? Is it anything useful like drivers or diagnostics?

 

Thanks!

[+slx]

Those FPU cards come up every now and then. I was tempted to buy one but after some research did not find any real use for it as it only works with some specialised software.

[+DarkLord]

22 hours ago, 1200XL M.U.L.E. said:

I'm very happy with this jump in performance and the increased stability of the Eiffel. The NetUSBee is probably very stable by itself and maybe with other accelerators. It just doesn't seem to play well with my Mega and its accelerator. 🤷‍♂️

 

 

It's been documented by several people (myself included) that the NETUSBee does not work with Pak 68/3 accelerator boards,

like I have in my STacy.

 

Pity that... 

 

[1200XL M.U.L.E.]

I put together a sort of triptych of my Mega 4 benchmark results .

 

 

The picture is 2400x600 so it may appear small embedded in the post. Click on it to see it better.

 

I benchmarked my Mega 4 without any acceleration or NVDI5 installed, followed by enabling hardware acceleration alone, and concluded with hardware acceleration plus NVDI5 loaded.

 

I am still interested in getting a copy of the Atari SFP004 disk. Today I found another auction on eBay for this card. The pictures show the seller running GEMBench and some utilities. Their version of GEMBench seems to recognize the FPU card. Mine won't. The utilities seem to test the FPU card. I asked the seller if he could make a copy of the disk and archive it or post it somewhere. Hopefully they respond in the positive!   The auction is here :

 

eBay #155192152540

 

Some images showing the disk and maybe the utilities on the disk.

 

 

 

 

I'd really like to run these utilities on my Mega 4. If my card is dead then I want to remove it.

 

Does anyone have a copy of the disk?

 

Thanks!