Whatever happend to 65T32(Terbium)?

[fachat]

Disclaimer: I have no direct information from WDC, basically only this forum.

However, frankly, I think the terbium is dead. Not enough requests for it. And maybe lacking some crucial features (like multiplication) to be a viable competitor to the ARM. And to compete with the ATMega, there should have been more controller-like packages.

Unfortunately it has kept many people just looking at it like a deer in the headlights. Only now there have been attempts to go beyond 8/16 bits, be it 65org32, or my 65k core

André

[GARTHWILSON]

fachat wrote:

However, frankly, I think the terbium is dead. Not enough requests for it. And maybe lacking some crucial features (like multiplication) to be a viable competitor to the ARM.

I just found the following in an email from Bill Mensch from seven years ago (!) while looking for something else: "The Terbium microprocessor is moving forward and we now have a data sheet and rough draft of a developer guide with sample code and comparisons with the ARM7 and other processors including a DSP processor from TI."  Since it has been so long, I would also have to say the Terbium is dead too.  News to the contrary would be a pleasant surprise.

[BigDumbDinosaur]

fachat wrote:

...I think the terbium is dead. Not enough requests for it. And maybe lacking some crucial features (like multiplication) to be a viable competitor to the ARM. And to compete with the ATMega, there should have been more controller-like packages.

I'd have to agree with André. If WDC truly wanted to get this product out there they would already have done so. The reality is the 65T32's need to preserve 8 and 16 bit mode backward compatibility is a downside and probably complicated the design to where Bill Mensch—who will be 67 this year and hence reaching the tail end of his illustrious career—decided to scrap it. WDC is doing fine with the current MPUs and Mr. Mensch is making a comfortable living from his work, so why expend design time and effort on something that will probably not see any adoption in the embedded market.

[GARTHWILSON]

Quote:

The reality is the 65T32's need to preserve 8 and 16 bit mode backward compatibility is a downside

That would be a big performance limiter; but from past emails and from things that used to be on WDC's website, I don't think the Terbium was going to retain 8-bit backward compatibility like the 65832 did.

[BitWise]

I suspect the 65C832 died because it was beaten into production by other 32-bit chips with far better performance.

Given that most people use C for 16 and 32 bit applications (rather than assembler) the backwards compatible features of the chip aren't that important to most target users. Once they've taken the hit of re-coding into C they have a more portable code base they can more easily port to any brand of microcontroller.

It would have been an interesting upgrade to a BBC, Commodore or Apple.

I've been meaning to have a go coding one in VHDL but I don't have the time right now.

[BigDumbDinosaur]

GARTHWILSON wrote:

Quote:

The reality is the 65T32's need to preserve 8 and 16 bit mode backward compatibility is a downside

That would be a big performance limiter; but from past emails and from things that used to be on WDC's website, I don't think the Terbium was going to retain 8-bit backward compatibility like the 65832 did.

The 65T32's draft data sheet listed "E8" and "E16" compatibility modes that respectively emulated the 65C02 and 65C816. A new instruction, XFE, would have been used for switching into E16 mode. The problem was the entire opcode matrix was already spoken for and there wasn't any room for XFE.

[GARTHWILSON]

Then all the more reason to do our own here and streamline its performance by not requiring backward compatibility to get in the way.

[BitWise]

BigDumbDinosaur wrote:

GARTHWILSON wrote:

Quote:

The reality is the 65T32's need to preserve 8 and 16 bit mode backward compatibility is a downside

That would be a big performance limiter; but from past emails and from things that used to be on WDC's website, I don't think the Terbium was going to retain 8-bit backward compatibility like the 65832 did.

The 65T32's draft data sheet listed "E8" and "E16" compatibility modes that respectively emulated the 65C02 and 65C816. A new instruction, XFE, would have been used for switching into E16 mode. The problem was the entire opcode matrix was already spoken for and there wasn't any room for XFE.

XFE might have been implemented as a co-processor instruction, with the WDM opcode or by having XCE adjust E16 and E8 when in 32 or 16 bit mode.

[White Flame]

Mensch's Wikipedia article says this, dunno if the year is also tied to the upcoming decisions or not.

Wikipedia wrote:

As of 2012, Bill Mensch is still involved with design engineering at WDC, in addition to his work as CEO. He has written the upcoming Terbium processor family's data sheets and will be making the major RTL design decisions associated with that processor architecture.

[BigDumbDinosaur]

White Flame wrote:

Mensch's Wikipedia article says this, dunno if the year is also tied to the upcoming decisions or not.

Wikipedia wrote:

As of 2012, Bill Mensch is still involved with design engineering at WDC, in addition to his work as CEO. He has written the upcoming Terbium processor family's data sheets and will be making the major RTL design decisions associated with that processor architecture.

That statement about the Terbium has been in the Wikipedia article for as long as I can remember. I edited the date to 2012 not too long ago to keep the article reasonably current. In any case, we all know how authoritative and well-written Wikipedia is, especially when the Internet monkeys get involved.

[Alienthe]

Were there any datasheets or other publications issued that can be found on the net? I have searched but didn't find anything.

Even if Mensch had not decided never to make Terbium it would nevertheless be interesting and most likely instructional to see how he envisaged a 32-bit design, considering the sheer elegance of the 6502 design. In my experience, if something is right it is often also elegant, looks right and feels right. And vice versa. And perhaps it could inspire SWEET32.

In aircraft design you have one such example: SR-71. And it flies every bit as fast as you can imagine when you see it on the ground.

[BigEd]

This is all I've ever found:
viewtopic.php?p=18872#p18872

(The other frequently linked file, 65xxx.txt, can now be found here but it says little about Terbium)

[White Flame]

Alienthe wrote:

In aircraft design you have one such example: SR-71. And it flies every bit as fast as you can imagine when you see it on the ground.

The thing leaks JP-7 all over the place until its skin gets hot enough at speed to expand and seal. I wouldn't quite call that elegant.

I've noticed in many things that if you want to increase performance, you've got to increase complexity. When simple elegance emerges, though, it's a wonderful but rare thing.

[BigDumbDinosaur]

White Flame wrote:

Alienthe wrote:

In aircraft design you have one such example: SR-71. And it flies every bit as fast as you can imagine when you see it on the ground.

The thing leaks JP-7 all over the place until its skin gets hot enough at speed to expand and seal. I wouldn't quite call that elegant.

The environmentalist whackos would have a field day suing Lockheed and everyone around them over some jet fuel drips. Never mind that during the height of the cold war, especially when Brezhnev was running things in the old Soviet Union, the existence of technology like the Blackbird was definitely a game-changer. However, it's probably good that the Blackbird was retired when it was.

Whether a Blackbird ever flies again, or not, it will always be a monument to the visionary thinking of Kelly Johnson and his crew, just as the 6502 will always be a monument to the brainpower that was embodied in Chuck Peddle and Bill Mensch.

Quote:

I've noticed in many things that if you want to increase performance, you've got to increase complexity. When simple elegance emerges, though, it's a wonderful but rare thing.

Performance almost always comes with a price in complexity. As things happen faster and the time available for them to happen decreases, timing issues become unavoidable. The original 6502 design started out very simple, so performance scaling was not nearly as difficult as it was with later and more complicated designs. A good comparison is the MC68000, which was initially much more complicated and never scaled as well as the 6502 design.

Consider that the W65C02S and W65C816S can be operated—with a reasonably good circuit design and fast logic—at 20 MHz, a 20-fold increase over the progenitor microprocessor—plus the 16 bit '816 can often accomplish the same work as the 8 bit models with fewer instructions. I have done some testing on my POC unit to see just how much usage of the '816's 16 bit registers, where practicable, can help in that regard. In the case of 32 bit integer arithmetic (which is much used in accessing mass storage), the performance improvement worked out to about 30 percent in addition and subtraction, and a whopping 55-60 percent in multiplication and long division. This being the case, the '816 in doing math has scaled the performance of the original 6502 by a factor of 30 or more. Had the original 68000 scaled to the same extent it would be the equivalent of operating it at 120 MHz—the fastest 68000 was clocked at 20 MHz.

The scalability of the 6502 family demonstrates that simplicity is fundamental to performance. However, an inevitable tradeoff involving performance is that any significant increase in one area will be detrimental to another. Although knowledge about it is uncertain, the genesis of the 6502 design probably sprang from the desire to observe KISS, rather than set a new record for MIPS. That the 6502 design has scaled as well as it has is as much a testament to KISS as to its designers' clear thinking and the setting of realistic goals.

[Alienthe]

White Flame wrote:

Alienthe wrote:

In aircraft design you have one such example: SR-71. And it flies every bit as fast as you can imagine when you see it on the ground.

The thing leaks JP-7 all over the place until its skin gets hot enough at speed to expand and seal. I wouldn't quite call that elegant.

Oh but surely it is. It solves the problem where it is a problem, not on the ground where it is not a problem. This is an example of thinking outside the box, putting it on fire and kicking it into the corner. Kelly Johnson excelled in this.

Quote:

I've noticed in many things that if you want to increase performance, you've got to increase complexity. When simple elegance emerges, though, it's a wonderful but rare thing.

Agreed. Which is why I would like to see the Terbium become real.

Back in the day it was said that ARM was inspired by the philosophy of 6502, and only now is ARM being extended into 64 bit. Clearly there is still room for the skilled designer.