STE took advantage of Moore's Law?

[Xebec]

I finally studied the motherboards of the STE vs the ST(fm) and noticed that the new STE chips are often combos of old ST chips (i.e. GST = MMU+GLUE), or old ST chips + new STE functionality (SHIFTER = old video shifter + new DMA sound).

 

Is it fair to assume these chips were also manufactured on a newer process than the original chips enabling more transistors for the same or lower cost?

 

I ask as I always found it interesting that commodore didn't take better advantage of Moore's law to reduce costs. The 1992 AGA chipset on the Amiga provided a brand new "Lisa" chip manufactured on a 1.5 micron process, but still included the original Agnus and Paula which were still being made on a very dated/costly 5.0 micron process.

 

P.S. It looks like Falcon may have continued the trend of a newer manufacturing process too.. COMBEL = MMU+GLUE+Blitter. , not sure what Falcon chip includes the STE DMA sound though..

[leech]

Wasn't the STE DMA sound 'emulated' via the DSP? (I could be completely wrong here, just a guess).

 

There were definitely a lot of people that were pissed off that Commodore didn't put forth effort into a newer sound chip for AGA. Though Paula was always one of the rather Unique things about the Amiga, it sounds fantastic. Kind of like how unique Pokey sounds.

 

My guess on the manufacturing costs? They probably just produced so many of the Agnus and Paula that they didn't sell in other systems, they figured they may as well stick 'em with Lisa. Kind of like I wonder if Atari had such a large stock of chips for the STfm they ended up bringing those back after the STe was released. Then again, from what I read more recently, it seems that the newer STfm were using the newer process and were made cheaper.

[Osgeld]

Mores law is not really a thing, and jamming a bunch of stuff in a LSI style chip does not necessarily mean cost savings (for example its usually cheaper to continue cranking out the same thing, even if it is old, rather than wait for a complete retool)

Edited July 4, 2017 by Osgeld

[ijor]

I finally studied the motherboards of the STE vs the ST(fm) and noticed that the new STE chips are often combos of old ST chips (i.e. GST = MMU+GLUE),...

Is it fair to assume these chips were also manufactured on a newer process than the original chips enabling more transistors for the same or lower cost?

Yes. This happened already even before the STE. There are several generations of the chipset. Each one based in a newer and cheaper process.

 

I ask as I always found it interesting that commodore didn't take better advantage of Moore's law to reduce costs. The 1992 AGA chipset on the Amiga provided a brand new "Lisa" chip manufactured on a 1.5 micron process, but still included the original Agnus and Paula which were still being made on a very dated/costly 5.0 micron process.

I'm not familiar with the Amiga chipset. But it is possible (likely) that those chips were NMOS. It is not always trivial, or easy at all to "port" from old school NMOS to a CMOS design, and yet mantain a 100% compatiblity. ST chips, OTOH, were CMOS since the beginning.

 

[ParanoidLittleMan]

Wasn't the STE DMA sound 'emulated' via the DSP? (I could be completely wrong here, just a guess).

DSP chip used in Falcon was not designed by Atari - it is universal purpose signal processing chip, done by TI, if I remember correct.

Falcon has STE DMA sound system too, although not 100% compatible. Lowest freq. mode is missing.

[leech]

DSP chip used in Falcon was not designed by Atari - it is universal purpose signal processing chip, done by TI, if I remember correct.

Falcon has STE DMA sound system too, although not 100% compatible. Lowest freq. mode is missing.

I always thought the DSP was a Motorola one. https://en.wikipedia.org/wiki/Motorola_56000

[Xebec]

Yes. This happened already even before the STE. There are several generations of the chipset. Each one based in a newer and cheaper process.

 

I'm not familiar with the Amiga chipset. But it is possible (likely) that those chips were NMOS. It is not always trivial, or easy at all to "port" from old school NMOS to a CMOS design, and yet mantain a 100% compatiblity. ST chips, OTOH, were CMOS since the beginning.

 

 

Commodore partially had the problem of owning the MOS fab.. I say problem because it initially gave them cost/time advantages for initial development and production, but later on it appears C= management didn't want to invest in newer chip technologies despite potential benefits.

 

The first *real* look at a new fab seemed to occur with their Hombre chipset which never made it to light of day unfortunately.. that was to be fabbed by HP on a modern process (600nm / 0.6micron), which means the transistors were on the order of 8x smaller than AGA Lisa, and 32x smaller than OCS/ECS -- quite the upgrade.

 

I'd be very curious what processes the ST chips were manufactured on over time and how many transistors they had.. I see the integration pattern (ST --> STE --> Falcon), but that's about all I can find on this topic.

[ijor]

I'd be very curious what processes the ST chips were manufactured on over time and how many transistors they had.. I see the integration pattern (ST --> STE --> Falcon), but that's about all I can find on this topic.

 

I'm not completely sure about the exact process and feature size. There were basically 3 generations of the ST chipset (before the STE). The two first ones were not full custom, they were gate arrays, somewhere around 3 microns. Obviously a consequence of the ST chipset being designed in such a short time. They are relatively small in terms of the number of transistors. In the order of 5000 transistors per chip.

 

There was no integration at the ST chipset. That started with the STE that basically combined GLUE and MMU. Later versions even combined Blitter as well.