Paradise PVC2 questions on Olivetti M19

[1ST1]

I love the M28 and M280, but it is a strange AT, using XT keyboard layout... But if M28 also uses that chipset, why you wrote new tool for M19 for switichng 4.77 to 8 Mhz? It should be there already for M28...

 

[Zare]

I'll have to discover more about M28. There is one on ebay in mint condition for about 300e. But I should have some documentation in my Olivetti archive too.

Why you wrote new tool for M19 for switichng 4.77 to 8 Mhz? It should be there already for M28...

Rather curious question; I didn't know until yesterday about M28 having FE2000A, ran on it accidentally by browsing Olivetti on ebay, those high-res pictures are from ebay item.
"The tool" I believe choas wrote and it's just a port byte out for value we discovered.

 

[Zare]

I don't care about M28 per se, but how it uses FE2000A might tell us what M19 can do.

I don't believe there is speed-switching in M28. The floppy set is same as M19, demo/test/keyb.utils.
The board has AT bus driver and AT clock generator. Like in 5170 you'd have to replace the crystal to bump up the freq.

AT crystal is *2 of CPU freq, on IBM 5170 it is a sole 12 MHz crystal on board.

On the retro web we have M280 board with 12 MHz 286 and a 24 MHz crystal next to FE2000A.

Obviously these M28/M280 boards have no 14.31 MHz crystal and still run FE2000A.

For me this makes 9.5 MHz M19 not a hypothesis but something tangible.

 

[Zare]

Tho Faraday documentation specifies certain oscillator frequencies, 14.31 MHz, 28.6 MHz, I think the chip itself works with frequency ranges. The numbers are written as they are, because of XT CPU controller context.

The control register value we use for 4.77 MHz should be seen as low speed, and 7.15 MHz value as high speed. Low speed, chip splits the clock by 3, high speed, chip splits the clock by 2 :

Olivetti M19 - 14.31 MHz crystal, runs 4.77 MHz with low speed (/3), 7.15 with high speed (/2)
Olivetti M28 - 12 MHz crystal, runs 6 MHz at high speed (/2)
Olivetti M280 - 24 MHz crystal, runs at 8 MHz at low speed (/3), 12 with high speed (/2), has user installable crystal option.

Math checks out. So by replacing master crystal in M19 we could manipulate the system frequency and try reaching fastest XT speeds.

Video card timing won't be a problem. There is no 14.31 MHz crystal on IBM PC AT mainboard, but if you slot CGA card in it it works. Faraday chip should take care of this.

 

[Zare]

Italian mag July 1986. Olivetti releases a range of new computers to follow M24/M21 in the "IBM compatible" market and the mag covers them.
These are M19, M22, M24 SP and M28.

In the typewriter market Olivetti will soon release two systems based on M19, ETV 260 and 500.

At early 1986 Olivetti had at least four machines in late development utilizing a Faraday chip to-be-released, the FE2000A.

That's a lot of machines to release with a new ASIC chip at the same time, and in my opinion the problem might've been what IBM avoided by using generic small chips throughout IBM PC; production, supply and logistics.

M19 is the only one that can act with old FE2000 at normal XT speed. In my opinion, there is no foul play here from Olivetti and M19 wasn't nerfed intentionally. Just production/supply issues.

 

[1ST1]

I have the original catalog / datasheet book with all of these computers, all options, monitors, printers, original software and so on. And I have all the computers on this picture except of the very rare M22 where only 6 prototypes have been made. I know mostly where 5 of these 6 prototypes are.

 

[Jan de Lange]

Just wanted to check this as it is too long ago, but indeed, the graphics library of Borland (Turbo Pascal, Turbo C++, Borland C++ for DOS) support the 600x400 graphics of the M19 as ATT400.

This is nice because I was planning to write some code for the M19.

 

[Jan de Lange]

Text in graphics mode: what a difference 200 extra lines make!

It was this post on Vogons on soft that supports the hi res mode of the Olivetti that made me check.

 

[Zare]

Gee I wonder who is the poster of that Vogons thread :D

Don't know how many of you have followed the talk about CGA here on vcfed, about the history and genesis and so on.

Olivetti M24 was so successful, it was the first machine with modified IBM standard to receive very wide support.

Up to this day, it is the most supported most used "non-IBM" PC graphics.

Other stuff that tried to enhance CGA with "full" memory of 32kB to get more colour failed. Plantronics, Quadram, even a bit younger than M24, but software support slim to none. Olivetti hit a nail on the head by just using a high resolution and no colour.

For example Windows 1.0 released in 1985 with M24 graphics driver, but no Plantronics 640x200x2bpp. Plantronics made sure that on programming side its easy to upgrade CGA code, in Olivetti's case that isn't exactly so. You can't just make additional 'assets' and double the graphics write code, the row mapping is different. Yet there is a M24 and no Plant driver.

That tells me Olivetti video system was more in market range of IBM video systems than any other of these 'alternate' cards like Plantronics which remained a niche.

It's not only about the specs of the video system but the machine running it, M24 in my opinion is the best personal computer of the first half of 80s. No contest.

Again, Windows brought Display Enhancement Board drivers for M24 in the years when it was known Olivetti isn't continuing with it, as they already moved to EGA. This expensive piece of hardware was not owned by many, its architecture was on the path to obsolesence, yet Microsoft made sure it was 100% supported.

In the end, there are many more supported software than listed on Vogons which we will discover as time goes by. Attached one of them.

 

[1ST1]

That tells me Olivetti video system was more in market range of IBM video systems than any other of these 'alternate' cards like Plantronics which remained a niche.

Olivetti was 2nd biggest producer of PC in the world at M24 times. But that's not all, it was popular in Europe, and as AT&T 6300 and Xerox 6060 popular in the US. I have german computer price lists of that era, and the cost of M24 was mostly equal as IBM 5150, 5160. Same price for better machine. Those who bought IBM instead bougt it only because of the smart IBM men in blue suit with tie.

It's not only about the specs of the video system but the machine running it, M24 in my opinion is the best personal computer of the first half of 80s. No contest.

But then M240 was better, more compatible, specially on low level floppy operations and easy upgradeable graphics.

Again, Windows brought Display Enhancement Board drivers for M24 in the years when it was known Olivetti isn't continuing with it, as they already moved to EGA. This expensive piece of hardware was not owned by many, its architecture was on the path to obsolesence, yet Microsoft made sure it was 100% supported.

Also GEM looks very nice on Go 329.

I think the 640x400 monochrome graphics mode of M24 was also the idea for Shiraz Shivji to implent such a mode in the Atari ST.

 

[Ghislain]

Hi, This is the last version of the M19 Video schematic

 

[Zare]

 

[Zare]

@Jan de Lange @choas75 Guys you run floppy drives and/or greaseweasel on M19? Could you, when it is convenient, test the floppy I/O speed while in normal and turbo mode?

 

[Ghislain]

I use Gotek

 

[Zare]

Ah yes you can test on the good Logabax too.
I wonder is there a DMA speed difference when switching frequency.

 

[choas75]

@Jan de Lange @choas75 Guys you run floppy drives and/or greaseweasel on M19? Could you, when it is convenient, test the floppy I/O speed while in normal and turbo mode?

Copied a 361723 bytes file using the copy command, not xcopy.
C drive is a CF on XT-IDE
A drive is a Gotek
B drive is a 360kB floppy drive
Z drive is an EtherDFS network drive

I waited for the B drive to stop spinning before starting a new measurent.
I did just three measurements each. Not 5 and removing the worst one.
I did an "echo . | time > start.txt" before the copy command (which sets the clock to 12:00:00am) and another one after.
Two decimals behind the dots probably gives a false feeling of accuracy. I didn't round, just copied the output.

Src/Dst	4.77Mhz	7.16Mhz
C/B	29.22s 30.33s 29.33s	29.05s 29.27s 28.33s
B/C	21.27s 21.22s 21.27s	20.38s 21.27s 21.27s
C/A	33.38s 33.38s 33.38s	32.33s 33.27s 32.22s
A/C	21.38s 21.33s 21.33s	21.16s 20.38s 21.00s
C/Z	22.38s 22.05s 22.38s	15.38s 15.38s 15.38s
Z/C	22.33s 23.00s 22.38s	16.05s 15.27s 15.27s

Not really impressive.
Only over the network makes some real difference.

 

[Zare]

This is the answer I was hoping to get. Thank you very much for the elaborate table and tests because you have covered everything.

The XT-IDE and NIC use PIO, floppy and Gotek use DMA.

Only in case of transfer between PIO devices, does the speed boost kick in, but bear in mind one side is EtherDFS which is a TSR and eats some CPU. You can try C/C that will give you the real PIO speed.

One should avoid DMA bound devices in XT arch, because DMA is slow and crippled, and DMA chips have their own clock input and rating. In XT arch it is highly uncommon to have above 5 MHz rated DMA controller, and to go to turbo speeds something needs to manage the waitstates on the controller to hold it at effective speed of 4.77 Mhz and sync that with the faster CPU bus. That is what wait state gen in Faraday chip does.

Which is a good thing because DMA is in check while modifying CPU speed.

 

[1ST1]

The result is as expected, because when running at "8" mhz the floppy drives wouldn't rotate faster, so FDC data rate stays the same. Surprising is the very little more speed for gotek versus mechanical floppy.

 

[choas75]

Surprising is the very little more speed for gotek versus mechanical floppy.

Yes, writing to Gotek is even slower than writing to a 5.25" 360kB floppy.

 

[Zare]

Yes, writing to Gotek is even slower than writing to a 5.25" 360kB floppy.

The result is as expected, because when running at "8" mhz the floppy drives wouldn't rotate faster, so FDC data rate stays the same. Surprising is the very little more speed for gotek versus mechanical floppy.

Maybe due to emulating floppy correctly, for the copy-protection schemes and such.