Jump to content
IGNORED

Atari Jaguar Creator Interview


pacman000

Recommended Posts

Found this on Slashdot: https://games.slashdot.org/story/02/09/17/146252/interview-with-atari-jaguar-creator-john-mathieson

 

They didn't have the full article, only a link, and the link was dead. Luckily the Wayback Machine had it: https://web.archive.org/web/20021013115028/http://www.chez.com/toxicmag/real/articles/mathieson.htm

 

(Scroll down a bit unless you know French; the intro is in French.)

Edited by pacman000
  • Like 3
Link to comment
Share on other sites

11) What about Jaguar II? Can you tell us if some prototypes exist or if it is a myth?

The replacement for Tom was fabricated and worked, it was called Oberon.

- The blitter was upgraded to a full triangle engine, and

- on-chip texture caches were added.

- It did anti-aliased texturing at full bus speed (i.e. same fill rate as Gouraud shading on Tom) and looked gorgeous.

- The companion chip, called Titania, which was Jerry plus another GPU with cache to act as CPU was never fabricated.

Damn, that sounds cool...

 

12) Taking into account that a part of Jaguar II was created (Oberon), do you know if any "real" uncompleted (without Titania) prototype of Jaguar II was created ?

There were 2 or 3 boards build with Oberon and Jerry.

Is there any known ownership/whereabouts of at least one of those ?

 

13) If you know someone who owns a "real" Jaguar II uncompleted prototype, could you give me his address ? ( I will steal the Jaguar II for my collection of Atari machines! :-) )

I suspect they were scrapped when Atari went down. I don't know any more than that. Sorry.

Hope that was actually not true. But, hey....

Link to comment
Share on other sites

Anyway it makes the same mistake of the old Jaguar, when you draw with the blitter you have to draw each primitive one by one (at least now you can draw triangles), instead of using a list of primitives (like the OP or PSX).

Incorrect.

 

Actually more than an order of magnitude incorrect.

 

The following test scene has 190 triangles, but has a grand total of 5,272 scanlines. Meaning, you have to spin up the Blitter over five thousand times (thus totally killing any available performance on simple handholding of the Blitter). Instead of just 190 calls (as is the case with Jag2).

 

5,272 / 190 = ~28 (hence more than an order of magnitude difference).

 

post-19882-0-42374100-1510449929.gif

 

- Interestingly, and I tested that, even current Blitter has enough bandwidth to display same amount of pixels at 60 fps in 190 calls.

- But it's impossible to expect that it's going to be able to do that with over 5,000 calls per one frame.

- Obviously, there will be a threshold where it's faster to use a SW rasterizing approach instead of Blitter, so to get most out of jag, a complex hybrid engine has to be written that utilizes its power properly

 

And in all honesty, if somebody has a problem with writing a simple brute-force loop of calling Blitter once per triangle, he has no business attempting 3D on jag. This is not an OpenGL Lego-style environment on PC...

  • Like 1
Link to comment
Share on other sites

That Jaguar 2 seems pretty awesome. I'd love to know what it could do. Technically, does it look like it really provided any new capability, other than simply being able to process graphics a little faster?

In a 3D environment, if you compare how many cycles are left on GPU after it's done hand-holding the Blitter (scanline by scanline), a very conservative estimate is 10x faster.

That is, if it was clocked the same and had the same instruction throughput. Which it doesn't:

Jag1's GPU: 26 MIPS

Jag2's GPU: 127 MIPS

127/26 = 4.7

4.7 * 10 = 47x

 

You may have heard that MIPS don't mean much, but in a software rasterizing 3D environment, MIPS is everything.

 

My current 3D engine is totally limited by how many MIPS jag does, even in 768x200. If I had 5x more MIPS, it would literally run 5x faster, save for very minor bandwidth slowdown (less than 5%, so not really worth talking about, let alone in jag2's environment).

 

 

A first-generation Jag2 engine would, on a simplest and most primitive possible engine design smashed together in 2 weeks do at the very least this:

-10x more polygons than jag1

- textures

- dynamic vertex lighting

- 60 fps

- HiRes

 

Realize, that instead of 52 MIPS on jag1 (of which you have maybe 5 MIPS available, once you are done handholding Blitter), you have 182 MIPS + Free texturing (to the limit of Blitter's bandwidth).

 

This is a game changer. Unfortunately, we all know the business realities of those times, so even if it got out, it stood no business chance against marketing budget of Sony anyway...

  • Like 1
Link to comment
Share on other sites

The following test scene has 190 triangles, but has a grand total of 5,272 scanlines. Meaning, you have to spin up the Blitter over five thousand times (thus totally killing any available performance on simple handholding of the Blitter). Instead of just 190 calls (as is the case with Jag2).

 

With my method you'll have to write the triangle list and then just 1 call to draw all the scene...

  • Like 2
Link to comment
Share on other sites

 

With my method you'll have to write the triangle list and then just 1 call to draw all the scene...

Yes, but that laziness will eventually result in quite a few other visual issues down the road. I'm not spending another page explaining what, as you'd just again come back with another one-liner outside of the reality.

 

 

The sheer ability to just supply vertices to the Blitter and not having to care is a huge game changer in itself.

 

Anybody can draw the loop to submit the vertices to the Blitter in assembler. But proper scanline rasterizing and efficient clipping ? In just 4KB ? That drastically limits the target coder audience that will be able to achieve this succesfully.

 

 

With jag2, any Joe Schmo from the street, could just pick up the dev environment, and start creating the 3D games in a weekend. Yes, there would be a lot of shovelware, like with the other platforms that allowed the same (Saturn, PS1, 32X, ...). But you need shovelware so that best titles can rise up (and for few other reasons too).

Link to comment
Share on other sites

I'm not talking about laziness, or how to write less code, I'm just telling how to make the hardware more efficient.

 

Write a list of commands, tell the blitter to draw it, while the blitter is drawing you can calculate the next frame, is much better than having to draw each primitive one by one and don't process the next frame until all drawing is finished.

  • Like 3
Link to comment
Share on other sites

I'm not talking about laziness, or how to write less code, I'm just telling how to make the hardware more efficient.

If you're not paying for the HW design, you have no business demanding what and how it should do.

 

 

... is much better than having to draw each primitive one by one and don't process the next frame until all drawing is finished.

Oh, I see - so you can't imagine processing next frame before current frame is finished :) ?

 

If your idea of 3D engine is waiting for the Blitter to finish drawing triangles, there's really no significant difference between the following two:

1. Draw Poly by Poly, wait after each poly

2. Draw whole batch of polys, wait after the batch

 

The difference between the two is not what makes Jag2 more efficient. Not by a long shot.

 

All it would do is lower the barriers of entry to substandard coders, nothing else.

 

 

Jaguar2 was an amazing and powerful design at the time it happened.

Link to comment
Share on other sites

If you're not paying for the HW design, you have no business demanding what and how it should do.

Yeah, how stupid of him to merely suggest something that both makes obvious sense and was supported on competing hardware.

 

All it would do is lower the barriers of entry to substandard coders, nothing else.

Sure. FIFO queues, DMA, dedicated memory, multithreading support... all that stuff only exists because coders are lazy, and hardware people love wasting transistors.

Real programmers have no use for features that make the code both easier to write and more efficient, because they have infinite time and resources.

Edited by Zerosquare
  • Like 3
Link to comment
Share on other sites

Yeah, how stupid of him to merely suggest something that both makes obvious sense and was supported on competing hardware.

Jaguar 2 was a huge step in the right direction and a great technological upgrade compared to Jag1, so it makes me sick that all the major improvements are being ignored, but one inconsequential feature is blown out of proportion (just so somebody feels better by kicking into it, while it's lying on the floor), despite making a more than an order of magnitude progress since jag1 .

 

 

But yeah, some people will just complain and bitch, no matter what. Must be in their nature (or perhaps including nurture)...

Link to comment
Share on other sites

If your idea of 3D engine is waiting for the Blitter to finish drawing triangles, there's really no significant difference between the following two:

1. Draw Poly by Poly, wait after each poly

2. Draw whole batch of polys, wait after the batch

 

With the second option you can do AI, Collision detection, T&L of the next frame while the blitter is drawing the whole batch. With the first option if you want to process the next frame while drawing you have to use interrupts, and IIRC the mmul and mac instructions can't be used with the interrupts.

 

Jag2 it's much better than Jag, but worse than Dreamcast (if finished it would be released between psx and dreamcast) because it lack of perspective correct texture mapping, 3d pipeline with floating point data, triangle rasterizer...

  • Like 1
Link to comment
Share on other sites

Found this on Slashdot: https://games.slashdot.org/story/02/09/17/146252/interview-with-atari-jaguar-creator-john-mathieson

 

They didn't have the full article, only a link, and the link was dead. Luckily the Wayback Machine had it: https://web.archive.org/web/20021013115028/http://www.chez.com/toxicmag/real/articles/mathieson.htm

 

(Scroll down a bit unless you know French; the intro is in French.)

 

 

Et n’oubliez pas : « Do the Math » ! :-)

  • Like 1
Link to comment
Share on other sites

World Tour Racing coder, Lee Brigs didn't seem that impressed by Jaguar II specs:

 

JEO] What do you know about the Jag2? Do you have a copy of the specs?

 

[LB] I never had a personal copy of the papers, all I know is from memory.

I will say that it was disappointing. The papers I saw were dated mid 95

and described a machine intended for release in the Christmas 96 market.

  • Like 1
Link to comment
Share on other sites

World Tour Racing coder, Lee Brigs didn't seem that impressed by Jaguar II specs:

 

JEO] What do you know about the Jag2? Do you have a copy of the specs?

 

[LB] I never had a personal copy of the papers, all I know is from memory.

I will say that it was disappointing. The papers I saw were dated mid 95

and described a machine intended for release in the Christmas 96 market.

I'd like to read that interview; where can I find it?

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...