Gorf

Im still waiting on someone to update one of the emulators to allow added RAM as well as POKEY simultainiously to the carts. Once I see that happen I acn finish UFO and Warbirds(working title).

Not dead (at least in my corner), just on a hiatus while I deal with RealLife issues. Hey PMP.....hope all things are smoothing out for you. Im still going to do those 7800 games I started. I too have crap to deal with so I feel you man! Good to see you posting. Gorf

More likely a blitter trick.

Yeah, too bad games didn't use voxel environments very often, I think Atari Karts might use height mapping in some of the tracks (for hills), though a lot of the tracks look mostly flat, still a really cool rendering method that looks great for a variety of environments. (and more impressive than most contemporary polygon rendered games -certain environments are better suited for polygons -but even then a combination of voxels and polygons could have been better) I dont think thre are ANY polygons or voxels happening in Atari Karts. Those are probably just clever uses of the OPL objects.

Sigh, poor emulation. So much hate in the world. Why ya gotta be a hata? I dont hate emulators. I hlped write MAME and MESS(the Bally Midway Astrocade hardware specifically). Im a diehard hardware fan. Emulators are neat tricks for the most part but they can never replace the real thing. I do use an emu for O^2, Astrocade and 2600/7800 developement though.

I think it would be awesome to see what it was capable of. the one chip was never finished, so its really impossible to emulate as it was only in prototype stages,many things would have certainly been changed. From what I've heard there is more than one 'chip' not emulated. I believe the Combi chip and the T-mapper caches never made it to the proto. There was never an RGPU or a Jerry II DSP either. Essentially you have a Jag I with a powerful Blitter(worlds more) and a GPU used with the old Jerry. The protos would never achive the 900,000 Textured, lighted, shaded, filtered polies or even the 2,500,000 g-shded polies with the current prototypes. Also the 68k was still used instead of the 020. And then the clock was only at 33 mhz. So the Jag II proto's were nothing more than a Jag I with a monster blitter.

Can I quote you on that? I'd rather see Jag II. The only use I'd have for an emu is development anyway. The 8 bit emu's are ok. But there would be nothing to run on it.

Im quite sure the Jag can do Power drift in a 256 color depth....which from the vids I saw are about all it is. However, use of a voxel engine like in Phase Zero would blow away anything in PD anyway. Why use scaled sprites when you can have a much more realistic looking game?

I have a fair amount of experience. You mean at the power adapter or I know the 2600 has a little black chip that supplies 5 volts to the board but does the 7800 still have that? I learned about that chip when my 2600 died 2yrs ago by researching each component inside the unit and testing each item till I got to that one which was the problem. Where this 7800 still works its more difficult to pinpoint the issue. I'll grab my multi meter from my parents in the morning and see what my power adapter is putting out. thanks for your help Though the V reg is a possible candidate, it might just be one of the components are slighty out of value, probably somewhere in the video output circuit.

As Far as the RF goes, if your 7800 is like mine the cable is seperate from the unit...ie you plug it into an RCA on the unit, then to the RF modulator. Try changing that out with anew cable. I've used a monster video cable with good results as well.

Save yourself some time. Use the 68k C compiler to get the system setup(and the incuded assembly startup lib files) but learn the J-RISC processor assembly. That is where the machines power is. If you are only worried about simple 2D games, the 68k C compiler is plenty efficient and will make your life much simpler. But if you want to pu the pedal to the metal, Learn the J-RISC's as they are only 64 instructions and rather simple. Why folks are so afraid of these processors, I'll never understand. If I can learn it on my own, anyone can...really.

Ask Scott Legrand who has a PHD or two and several patents pending. It does not use it much at all from what Doug and Scott tell me. They also told me they could do even better if they started over knowing what they know now.

Yes, the blitter had single command triangles and the RGPU was geared toward C code and 3D math. The JagII could render 900,000 textured, shaded, lighted, filtered(bi-linear) polygons per second with just the RGPU and Tmapper chip in the blitter. It was also capable of 2,500,000 non-textured polies per frame. This was the original specs. The proto boards were not this strong and only ran at 33mhz. The original specs were supposed to be 66mhz. Jag II proto's did not have the RGPU (part of Jerry II), COMBI or the tmapper chip which was to have 8k fifo read and 8k lifo write private cache and 256k internal cache workram area. The proto blitter had 8k basic texture rom and work 8k RAM. the DSP had 96k cache. the GPU still only has 4k in the protos. According to Downinx the blitter II circuit alone was way bigger than the original Tom and Jerry in transistors combined. That means it could blow the PS1, N64 and the Saturn easily away in poly count per frame without even using the GPU and DSP. That is a 15,000 polygons per frame not even counting AI and game logic which were also done in local rams making the parallel ability amazing.Contrast this with Jag I at about 25,000 G-shaded polies per second and 450 polies per frame. PS1 was 180,000 textured, shaded, lighted, per second and only one MIPS processor.(3000 polies per frame) Saturn 75,000(guess) textured, shaded, lighted, and two SH-2's processor.(1334 polies per frame) N64 was 100,000 textured, shaded, lighted, filtered, and only one MIPS processor.(1667 polies per frame) So clearly the Jag II would have handily handed their asses to them other consoles without even using the other two RISC chips to help out. But then the unified bus would have caused a few bottlenecks but the locals were bigger on the RGPU and the AI was done on it. So that would have greatly helped the bus contentions by leaps and bounds.Theses figures for all machines are based typically around 50 pixel polygons.

This thread is truly superflous...however.... Just release Jaguar II at 400 mhz. Oh, and @ Ti guy the blitter was plenty of 3D capable. It was the system bus design and that accursed 68k that was the biggest detriment to the system. In fact the Blitter should be what GFX card manu's should be thinking today. Well they do ise blitter but they are dedicated and hardwired to polygons only. Even with todays GFX cards, points lines and circles are very hard on their performances and eat up a lot of BW. In fact the JAguar blitter can outdo even some of todays poweful graphics cards using voxel technology. Todays cards are just not geared for this and voxels are much better at representing 3D than polygons are. See Phase Zero for the Jaguar. Im not sure 3DO had any 3D hardware per say either...I could be wrong about that though.

The blitter and GPU can handle it alone. Why add the extra burden of the OPL? Use that for sprites in the foreground.

The 32x can't keep up with the Jaguar hardware wise. It came after the Jag and died before it. Even the supposedly awsome poly demo did very little to impress. Still never surpassed even some of the least greatest Jag games. Let's see the 32x do BattleSphere, Phase Zero or your excellent RPG project. Can you say CHOKE? sure you can.

Jag I fixed would be plenty fine. Thought the Jag II blitter is a monster. However the emulation point is moot. We've not even got the JAguar I emulated(thankfully) so why bother with a Jaguar II. I'd much rather see the real hardware than a silly imperfect emulator. Stella is probably the only emulator that best represents the machine it is trying to emulate, by leaps and bounds.

That is the problem. We got too many Genny like ports that did nothing to add color and more levels. More game like the Morphs, IS's and I-Wars with less 68k code would have worked fine, like the winning formula of Tempest 2k and more Atari classic updates would have gone a long way in boosting the Jaguar's market share. I know many Folks including myself were standing on the promises the Tramiels made regarding this and they did not deliver.

the host processor would have little to do with streaming and no doubt would be handled by the DSP.

See BattleSphere. And it is a detriment to the system because of it.

The obvious solution would be having the Tramiels support Atari with personal funds until the Jaguar was ready for release, or at least help things a bit more, maybe not a perfect launch, but more coordinated than the half assed '93 one. Howver, a very interesting and more ambitious option was suggested by atarian a while back: If they'd done that, and it worked, there should have been sufficient funding to really make a proper debut for the jaguar, with bugs fixed (and maybe other improvements like blitter buffering, and get rid of the DSP's 6-cycle delay), the added 020, and maybe even a CD drive. (especially if that would have further increased investor intrest) Have a suitable launch lineup of some dozen games (at least 1/2 heavy hitters like T2K, Wolf 3D, and AvP or Doom, along with some simpler upgraded ports like Raiden, and other exclusives like an improved Cybermorh, and reasonably complete Crecent Galaxy) With that kind of fundign early enough on, maybe flare could have implemented cache for the RISCs and toss out the added host CPU entirely to save cost. In any case, better funding and accelerated development should have facilitated a good RISC compiler, particularly with the bug-free chips. Kskunk mentioned C coding wasn't considered a key feature by flare when designing the Jaguar, but perhaps they chould have shifted rioretys faster or done some other changes with the additional funding available. (later in development, once it became more clear how important C was becoming, particualrly with the 3DO) Maybe the same could have been done with texture mapping support as well, or at least shifted a bit for that once it's importance was realized. (but hardware polygon rasterization might have been better to focus on) If adding the cache and dropping the host wasn't practical, perhaps another option would be possible. Back to the ARM again. The ARM-60 used in the 3DO wa a very small chip 9at 35,000 transistors), much smaller than an 020, and even smaller than the Jag's 68000, it also should have substancially higher performance than an 020, but it lacks a cache, so would have bus contention problems. (but still way better than the 68k, and would still increase jerry to 32-bit width) Perhaps atari could have licence the ARM-6 core (the ARM 60 being a bare core) and incorporated it into Jerry, maybe adding a 1-2 kB scratchpad to work in as well, or allow it to use the DSP's scratchpad. (even with 2k of SRAM and the ARM core added to jerry, it should still be smaller than TOM). However, ideally you want a chip with cache, and the ARM 600 and slightly cut-down 610 are far larger, and more expensive chips becouse of this, with ~245,000 transistors more from the added cache. Of that 245,000 transistors, 196,608 comprise the SRAm, with the remaider being cache logic, so it should be possible to design a custom chip with the ARM-6 core and coresponding cache logic, but much smaller cache. With a small, 256 byte cache (like in an 020), such a chip should be 95,680 transistors. (about 1/2 that of the 020, and added to Jerry, still more than 50,000 transistors less than Tom, of course, simply adding those onto Jerry isn't simple, it would need to be incorporated while designing the ASIC in the first place, to efficiently accomadate it onto the die) Going by this line of reasoning, you could add 1 kB of cache and still not be as large as Tom, with the ARM portion adding 132,544 transistors. (and as a standalone chip, still a good bit smaller than an 020, and even poduced on 1 nm should use less silicon than the the 600,000 T .5 nm Jerry) If such a chip was really not practical to include on Jerry, it could either be provided by ARM on the standard 1 nm process, or possibly licenced and produced along with Tom and Jerry (by Toshiba iirc) on the .5 nm process, making it a very small chip indeed. (for the time) Or licence the full 610 and have it produced along with Tom+Jerry on .5 nm. (at which point it would be only 2x the size of a 1 nm ARM-60, and only a little bigger than a 68000 on 1nm) When making such a custom hip, cut down fromt eh 610's cahe, maybe the cache logic could be stripped down as well to a basic direct mapped instruction cache, making the chip even smaller. (though that would probably be rather cumbersome, so only cutting out the cache RAm would seem a lot more practical, maybe 1 kB split into 2 512 byte sections) The 020 would be many times more effective in cost and implementation than all of this.

Well they had over 80 titles from their own vault. They did not use them. They had a deal with Time-Warner/Atari for some. They did not take advantage of it. So I like you agree that Atari was their own worst enemy in this. No developer support. Not letting the technichians designing the system to do a few simple things to make the machine MUCH better. Too busy wasting money on lawyers suing the other VG companies. Speaking of bottom line, let's face it...their heart was not really in this. It was in their bottom line. If they spent a little more money on either an 020 or instead, keep the 68k but make it impossible for it to run in the game loop after boot and add the necessary registers needed in the blitter for some double buffering of the blit regs, true dual port write backs to eliminate stalls, and dont tie the DSP to the host, allowing it to be a true 32 bit processor externally, Doom would have ran at twice the resolution and never drop below 60 FPS....either way. These hardware changes (which this topic is about anyway) were a giant leap over the 68k as a host....also you would have had in game music in DOOM and suffered no slowdown. After all , the music fit on cart and it actually plays with a recompile and the right flag set to allow it. Ah hindsight is a beautiful thing, aint it?

Umm, OK, I would have just thought the sound signals would be mixed after being converted to analog audio. YEs butyou still nedd the DSP to take those sound channels, mix them and then send them out to the I2S channels. Yes like Doom or Wolf 3D for instance, adding new levels and even entire new missions.

You still need the DSP to mix it with itself...that does not happen automatically. Right the CD was still an over all success and if the Jaguar had a more powerfull porccessing chain( ie the 020) the tools would have been better from day one, running AI an game logic on that intead of the 68k or splitting up the rendering with the GPU. Either way the 020 is win win. The JAg CD did better video than most of the other systesm off CD....the Jag II was a pipe dream after the first few years of disaster with the Jag. and games could ahev been developed to use CD and cart and just the cart work alone(with lesser content) if you did not have the CD. This would have given users a reason to upgrade to an addon later. Yes some of the FMVstuff was interesting but never game sellers. You'd not have made a lot of money on FMV only with no game. No it wouls have had the same bottlenecks without a true dual port unity cache between them, which would require more silicon to allow for. This is the cahce that would allow the units to talk to each other without hitting the man bus, requiring a small private bus to communicate. The blitter would fill the cahce from either side.. a simple 2k or 4k would be plenty for this. Even in main you still have bus contentions. You'd have had much less with an 020 or a unity cache without a host. The powerline would not have affored this alone. Oh and btw even without the private bus for the dual port ram, TRUE dual port ram was not cheap in 92 by anymeans.

The DSP is hitting the bus at 6 cycles per access, thats why. As far as developers, over 300 signed with Atari before it was even manufactured. GAmes like Area 51 would have been and then you could always release that after the fact on CD, as sega did and Sega did not end up splitting their market as a result. The cache is the key to the performance since the Jaguar is a unified bus. There for the 020 would have stayed off the bus most of the time while still handling all the game logic and AI. The only thing the GPU would be doing is render, render, render. The only the the DSP would be doing is sound, networking and input. The cheapest ARM did not run very fast and those that did would be much more expensive than an 020. The 68k was used as Atari was able to load up on them as they were very cheap as they were also at that point on their way to obsolesence. Motorola was moer than happy to dump them off cheaply to Atari who was one of their bigger customer bases.