How powerful was the cancelled Atari Panther compared to the Atari ST/Amiga?

[phoenixdownita]

8 hours ago, OldSchoolRetroGamer said:

Hey after 18 pages of madness and bickering back and forth I was wondering about something. How powerful was the cancelled Atari Panther compared to the Atari ST/Amiga? ?

I'm sure it would have beat them, if it didn't there would have been no reason to its existence.

 

Wrt 3D of that era I believe we forget how painful it was, I recently got the CD32 bug back so I bought a TF328 (pretty much doubles the speed of the system thanks to fast RAM) and then a TF330 (68030 @50Mhz and even faster RAM, making it about 7x faster than a stock CD32 when not dealing with the coprocessors) and some of those CD32/A1200 3D games move around better but honestly it wasn't until 68040 or actually 68060 that the Amigas could try to play Quake (as it is FP heavy).

 

For reference Doom on an an accelerated A1200 (68030 @50Mhz) seems to be around 11FPS 

 
Quake on an A1200 with a 68060 @66Mhz delivers about 14 FPS:

(the author of the video has it in the comments:
anouk33
1 year ago
14fps. timedemo1)
 

Yes the Amigas eventually were held back by their custom chipset (AGA/ECS/OCS they were slow) but it helps putting things in perspective. RTG graphics changed that but it got expensive, and too late.

So the Panther even with a 68020 would not have been a monster just because of it, but it should have been capable of beating A500/AtariST and likely the stock A1200 (just don't make the mistake to have a unified bus or in the case of the CD32/A1200 to ship with 0 dedicated mem to CPU).

 

On a CD32 to see the effects of fastmem and/or 030@50Mhz check this:

everything is much smoother but even an 030@50Mhz is not cutting the mustard in 3D, one really needs a much faster gfx subsystem and some 3D offloading to the hardware.

 

 

In that regard what was achieved by the Jag is actually very good, its Doom port is "silently fantastic" imho, much better than the above. I believe a Jag with just Tom and a more traditional audio subsystem (take something like the Sega Genesis) would have run circles around Amiga/AtariST (we have evidence of that in the Acorn Archimedes already).

Note a Sega 32X already runs in circles around those (the dual SH-2 setup being no slouch).

[Lost Dragon]

On 9/28/2020 at 9:48 PM, Leeroy ST said:

 

Do you know of the Panthers 2D capabilities? Or any quotes from developers on them? I know Cybermorph was a 3D port later released for the Jaguar which would have been impressive for the time, but in a hypothetical where the Panther launches in 1991, could it compete with Gen and SNES in 2D games?

All the Panther information I have had, I shared. 

 

If people still want to spread the myth Cybermorph started life on Panther and are happy to ignore multiple sources from both ATD and Atari Corp pointing out it was built from the ground up for the Jaguar, all the hours invested in bringing credible information to the community has been an utter waste of time. 

 

I'm done playing social media experiments. 

 

 

Now the GTW book is out, the Lynx Panther Strider II interview Frank Gasking carried out is up, that's it. 

 

I'm out. 

[Lost Dragon]

17 hours ago, carlsson said:

I thought that Lost Dragon debunked that myth about 15 pages ago? (September 4th to be exact) The early footage and screenshots are from the Jaguar. There is no solid proof that even a prototype version of Cybermorph ran/was developed on the Panther. However I realize some urban legends never will die.

I did. 

 

Fred Gill and Brian Pollock of ATD confirmed it was Jaguar only. 

 

Various Atari Corp sources confirmed it to not only myself, but Shinto. 

 

The likes of Shinto, myself, Frank Gasking etc have presented our evidence, named our sources, carried out due diligence into claims of titles like Cybermorph and Daemonsgate, Humans and Raiden from Imagitec Design on Panther. 

 

If people still have doubts then i suggest they carry out their own research and see how they get on. 

 

We are done. 

[Lost Dragon]

And the original plan was to launch the Panther and Lynx at the same time. 

 

See Bob Gleadow interview which has been posted in Panther thread or listen to Bill Rehbock interview and you'll see that never happened as Atari lacked the resources to launch 2 consoles at same time. 

 

The Atari UK P. R manager talk of there only being a 9-12 month window with Panther before Jaguar would of been ready to launch is also in the same thread. 

 

It's all there and as a poster said a few pages back, we are being played here. 

 

Time to put a few threads etc on ignore. 

[Leeroy ST]

9 hours ago, Lost Dragon said:

And the original plan was to launch the Panther and Lynx at the same time. 

 

See Bob Gleadow interview which has been posted in Panther thread or listen to Bill Rehbock interview and you'll see that never happened as Atari lacked the resources to launch 2 consoles at same time. 

 

The Atari UK P. R manager talk of there only being a 9-12 month window with Panther before Jaguar would of been ready to launch is also in the same thread. 

 

It's all there and as a poster said a few pages back, we are being played here. 

 

Time to put a few threads etc on ignore. 

You do realize that no one is saying you are wrong about Cybermorph. You completely glossed over the question about the Panthers 2D capabilities, a different subject because of you believing that the conversation was only about cybermorph because of a rushed misunderstanding.

 

Again, do you have any other (other than Strider 2) UK quotes about the Panthers 2D capabilities from devs? I know you posted things about some having issues but not specifying, and some saying it "blew SNES out the water" but did anyone actually break anything down about the strengths and limitations of Panther 2D? (not talking about cybermorph.) This is a subject not really touched on even in the Panther thread, outside limited exceptions.

Edited September 30, 2020 by Leeroy ST

[phoenixdownita]

On 9/29/2020 at 8:19 AM, mr_me said:

https://web.archive.org/web/20031207084119/http://www.atari-explorer.com/Panther-Spec.htm

 

Here are the panther specifications.  Similar to the 7800 it is object/sprite based.  The key is how many objects can it display on a scanline without overloading the system.  There's some comments about this in the panther topic in the prototypes section.

I think this one got lost in translations and thanks for the link (linking directly to the inner GIF here)

 

[carlsson]

It can be noted that Jeff Minter made a reference to never being able to put 65535 sprites on the screen, but the brief spec only promises 2000 sprites. Based on what I've read, for specific tech demos it performed well, but not sure how it translated to actual game situations. Many systems still tended to be stuck in the idea of a tile or bitmap based background layer, and then independently movable objects on top of it. Like mr_me noted, Flare seem to have shared the same philosophy as GCC and designed it to be mostly or fully sprite oriented, meaning all your background tiles need to be individual sprites with all it involves when it comes to e.g. scrolling the screen and ordering a series of sprites to move.

 

Regarding the resolution, wasn't 320x200 a bit in the low end even with a palette of 256K colours?  I know the Genesis only does 320x224 (from a palette of 512 colours) but it appears the SNES in some situations can be coaxed to an interlaced mode of 512x478 and has a decent palette of 32768 colours. I don't have stats on how many Amiga and Atari ST games operate in 320x200 vs 640x200. Obviously the higher resolution, the more strain on the graphics processor to move pixels and they may have gone for fps rather than resolution here by sticking with the seemingly lower resolution.

[mr_me]

The 320x200 resolution is plenty.  In the prototype thread there's references to two developers complaining about problems with only a few objects/sprites per scanline.  I assume that's in addition to all the objects needed to render a background.  The pixel depth is definable per object, between two and eight bits.  Maybe reducing the pixel depth helps.

 

Not sure why it would be limited to 32 colours per scanline.  Each object can reference a different palette  Perhaps it's 32 colours per object/sprite which is a colour depth of five bits.

[zzip]

6 hours ago, carlsson said:

Regarding the resolution, wasn't 320x200 a bit in the low end even with a palette of 256K colours?  I know the Genesis only does 320x224 (from a palette of 512 colours) but it appears the SNES in some situations can be coaxed to an interlaced mode of 512x478 and has a decent palette of 32768 colours. I don't have stats on how many Amiga and Atari ST games operate in 320x200 vs 640x200. Obviously the higher resolution, the more strain on the graphics processor to move pixels and they may have gone for fps rather than resolution here by sticking with the seemingly lower resolution.

I was thinking the same, the system should be a little more forward-looking

 

Although on the low-resolution TVs of the day,  320x200 doesn't look so bad, with enough colors on screen it could look realistic

 

Very few ST games use 640x200.  Colors are too limited and pixel aspect is weird.   Not sure about Amiga, since it could do more colors in that mode.

[Leeroy ST]

7 hours ago, carlsson said:

Flare seem to have shared the same philosophy as GCC and designed it to be mostly or fully sprite oriented, meaning all your background tiles need to be individual sprites with all it involves when it comes to e.g. scrolling the screen and ordering a series of sprites to move.

 

That does work well for Isometric and Pseudo3d 2d titles which may be what they were going for considering the early 3D capabilities would only really work as a novelty at first.

 

3 hours ago, mr_me said:

The 320x200 resolution is plenty.  In the prototype thread there's references to two developers complaining about problems with only a few objects/sprites per scanline.  I assume that's in addition to all the objects needed to render a background.  The pixel depth is definable per object, between two and eight bits.  Maybe reducing the pixel depth helps.

 

Not sure why it would be limited to 32 colours per scanline.  Each object can reference a different palette  Perhaps it's 32 colours per object/sprite which is a colour depth of five bits.

Some other documentation I can find online implies it's sprites and not scanline. Wouldn't make much sense for 32 colors per scanline.

[carlsson]

I suppose depending on how the graphics processor draws the sprites, it might have a limitation on colours per line even if those belong to different sprites.

 

As long as we don't have a working prototype to investigate, we can only speculate what it means, if the brief specs were glossing over technical details or simply misunderstandings.

[zzip]

3 minutes ago, carlsson said:

I suppose depending on how the graphics processor draws the sprites, it might have a limitation on colours per line even if those belong to different sprites.

 

Yeah swapping palettes per line is a common software technique to get more colors.  It may be doing that in hardware.    The other thing is changing palettes saves screen memory

 

320x200 32 colors / line = 5 bit planes or about 40K.    Otherwise putting 7860 on screen would require at least 13 bit planes  or 104K.   That spec sheet says this thing was only to have 32K RAM, which doesn't seem like enough.  (unless we are using 2600 programming techniques)

 

But on the other hand 200 lines @ 32 colors = 6400 colors,    which is 1460 short of the 7860 it says it supports.    Maybe those are extra colors provided by sprites and/or overscan?

[carlsson]

Ooh... an entirely sprite based console that is chasing the raster beam at 16 MHz. Yeah, if that was the case I really could see how it would be deemed even more cumbersome to program than the Jaguar and a big flop once it was released.

[zzip]

24 minutes ago, carlsson said:

Ooh... an entirely sprite based console that is chasing the raster beam at 16 MHz. Yeah, if that was the case I really could see how it would be deemed even more cumbersome to program than the Jaguar and a big flop once it was released.

Yeah exactly..  the thing must have had more than 32K RAM  this was the early 90s, right?

[carlsson]

Curiously that spec mentions sound RAM separately, but not video RAM. While many systems have shared memory between CPU and video system, 32K SRAM sounds very little. Even the Lynx has 64K DRAM.

[carlsson]

Did anyone link to this page before?

https://valentinomiazzo.github.io/Atari-Panther-Part-2/

https://valentinomiazzo.github.io/Atari-Panther-Part-3/

 

Assuming the author knows what he writes about, that is interesting. The 2000 sprites quickly became 480 sprites.

Edited October 1, 2020 by carlsson

[zzip]

14 minutes ago, carlsson said:

Curiously that spec mentions sound RAM separately, but not video RAM. While many systems have shared memory between CPU and video system, 32K SRAM sounds very little. Even the Lynx has 64K DRAM.

And the Jaguar had 2MB, which was comparable to home PCs of it's era.  It would seem the Panther should have at least 128-256K just for video-  for page-flipping, scrolling and what not, plus some system memory for heap.

 

No wonder they couldn't get it to work ?

Edited October 1, 2020 by zzip

[carlsson]

It seems you feed the object processor line by line, filling up the SRAM. The page says one of its strengths is to be able to read directly from ROM compared to e.g. SNES which needs to read from graphics RAM, which probably is good for 2D platform games, shoot 'em ups, puzzle games etc but I think once it comes to 3D, even pseudo-3D, you want to calculate frames along the way where the RAM would be a shortage. It  makes me wonder if Atari had a price point and Flare came up with a spec that was possible to fit within the price, and then a more advanced one that broke the budget but had more to offer.

[zzip]

15 minutes ago, carlsson said:

It seems you feed the object processor line by line, filling up the SRAM. The page says one of its strengths is to be able to read directly from ROM compared to e.g. SNES which needs to read from graphics RAM, which probably is good for 2D platform games, shoot 'em ups, puzzle games etc but I think once it comes to 3D, even pseudo-3D, you want to calculate frames along the way where the RAM would be a shortage. It  makes me wonder if Atari had a price point and Flare came up with a spec that was possible to fit within the price, and then a more advanced one that broke the budget but had more to offer.

It was also supposed to support a CD-rom.   Without much memory to buffer data, the drive would need to seek constantly, providing a miserable experience

[carlsson]

One would hope that the CD-ROM expansion contained its own buffer RAM, and that the object processor was able to read it just like it was cartridge ROM.

[Leeroy ST]

3 minutes ago, carlsson said:

It seems you feed the object processor line by line, filling up the SRAM. The page says one of its strengths is to be able to read directly from ROM compared to e.g. SNES which needs to read from graphics RAM, which probably is good for 2D platform games, shoot 'em ups, puzzle games etc but I think once it comes to 3D, even pseudo-3D, you want to calculate frames along the way where the RAM would be a shortage. It  makes me wonder if Atari had a price point and Flare came up with a spec that was possible to fit within the price, and then a more advanced one that broke the budget but had more to offer.

Some reports I've heard over the years is one of the reasons the Jaguar was gaining so much internal interest was that it's demos had surpassed was their Falcon computers demos could do. Which may explain why as they shifted entirely to the Jaguar (and Falcon sales weren't where they needed it to) they quickly cut the Falcon as well.

 

While that doesn't really give to much detail it does imply that the Panther may have possibly been weaker than the Falcon.

 

May not have been too far off the mark, looking at this Quake demo engine made by someone

 

Of course it struggles with it obviously, but the fact it can do what's displayed is pretty impressive. Of course the Jaguar is stronger than the Falcon but if the Falcon can do this that does make it seem reasonable to assume the Panther would have to be quite a bit weaker.

 

Like I said, it's circumstancial evidence but it wouldn't surprise me if Jaguar sparked rapid interest due to its capabilities. This video was a later demo for the system so at the time of 1992-1993 the Jaguar would have likely smashed the floor with whatever demos were available at the time.

 

 

[carlsson]

Well, the Falcon also has a 16 MHz CPU, but a full 68030 instead of an overclocked 68000. It also has the famous 32 MHz 56001 DSP chip. I'm sure there are plenty of comparisons of the Falcon and Jaguar hardware. However the Falcon was quite pricey, if my memory serves 2-3 as much as e.g. an Amiga 1200 which has more limited specs.

 

For sure if the ST/STE ever had been consolidized, if it was affordable the DSP would've found its way into it too. However I believe it would have made a rather expensive console for its specs though of course volumes bring down the unit price.

 

Although there is not a huge lot of 1st hand or factually verifiable information about the Panther, the more I read about it and put it in context with the competition, the more it makes sense to have canned the project.

[Leeroy ST]

18 minutes ago, carlsson said:

Well, the Falcon also has a 16 MHz CPU, but a full 68030 instead of an overclocked 68000. It also has the famous 32 MHz 56001 DSP chip. I'm sure there are plenty of comparisons of the Falcon and Jaguar hardware. However the Falcon was quite pricey, if my memory serves 2-3 as much as e.g. an Amiga 1200 which has more limited specs.

 

For sure if the ST/STE ever had been consolidized, if it was affordable the DSP would've found its way into it too. However I believe it would have made a rather expensive console for its specs though of course volumes bring down the unit price.

 

Although there is not a huge lot of 1st hand or factually verifiable information about the Panther, the more I read about it and put it in context with the competition, the more it makes sense to have canned the project.

 

If the Jaguar is stronger than the Falcon and the Panther is weaker than the Falcon this put in an an odd range of specs.

 

Considering what we ended up having with the Jaguar today (and it being upped by the 3DO) the question is how limited was the Panther? Because the more digging goes on the weaker the system seems and while there are comments about the console "blowing out" the Gen/SNES I wonder how true it actually is.

[Punisher5.0]

1 hour ago, carlsson said:

Curiously that spec mentions sound RAM separately, but not video RAM. While many systems have shared memory between CPU and video system, 32K SRAM sounds very little. Even the Lynx has 64K DRAM.

I was thinking the same thing. Id imagine it would be a huge bottleneck with the proposed capabilities.

[carlsson]

Perhaps the "blowing out" related to specific tech demos. I don't know how the 256K palette works, or if the object processor could dynamically apply any of the colours in the palette onto the data it read. Possibly if reading from ROM without needing to shift data into RAM was beneficiary, you could do very cool demo effects, colour cycling etc that you could only dream about on the SNES but unless you're a die hard Jeff Minter fan (meant in a positive way, I like his games but not everyone do), this console might not be for you.

 

Also I saw that the dev boards have extra RAM that acts like a soft ROM. It makes me wonder if you can access it like it was genuine RAM, and if the devs who got early access to the system tested it out by using features not intended on the production model, before getting to make actual games. Then again, if those devs reported back to Atari what they liked and didn't, it would have been a reasonable last minute design update to double or quadruple the amount of RAM if that is what would make it fly. (Any reference to "Where did you learn to fly" is completely unintentional)