Neo Geo games to Jaguar? Would be possible?

[sd32]

On 8/16/2024 at 6:02 AM, CyranoJ said:

I already have a tech demo with streaming backgrounds from ROM, full bitmap image, no tiles.

A 16bit 4096 x 255 16-bit bitmap packed down to 167k, and used just a few scanlines of GPU time to decompress during runtime.

 

 

Sounds impressive… but surely you can’t surpass Metal Slug “mighty” 30fps 🙃

[agradeneu]

31 minutes ago, sd32 said:

Sounds impressive… but surely you can’t surpass Metal Slug “mighty” 30fps 🙃

I think it runs 60. Despite Jaguar stopping and faking dead😅

[Kalani]

On 7/15/2024 at 12:24 AM, 42bs said:

I already wrote, that Gouraud shading is there. But that allone makes the Jaguar no 3D machine. I for example use Gouraud shading often to generate a gradient for the sky.

Many did 3D programs on the Jaguar. Including me. But the rendering needs a lot of work and takes many many cycles (and not talking of textured polygons).

It was designed to handle 3d graphics similar to 80s arcade and early 16 but home computers.  No one would of thought Sony would sell a 3d work station at a 400 to five hundred dollar retail loss. They off set it by producing some many chips in house. Doom and tempest 2000 for Jaguar on big screen and alone in the dark for 3do on big screen was amazing to ten year old me. 

[laymanpigeon]

It's far more plausible to port Space Hulk: Vengeance of the Blood Angels to Jaguar.

[DEATH]

Just to say that, on the Jaguar, the maximum (normal) ROM BUS bandwidth is 21MB/s, not 10MB/s.
The system clock is (approximately) 26.6Mhz, the minimum ROM access cycle is 5 (5.32Mhz or 188ns access time) and the maximum BUS size is 32 = 21MB/s
 

Oh, and in FAST mode (if it works...) the ROM BUS bandwidth is 53MB/s (32bit, 2 clock cycles, 13.3Mhz, 75ns)
 

On the COJAG (at least on some games) you can multiply these bandwidths by 2, the cartridge BUS being on 64bit.

[laymanpigeon]

I am just looking around FPM DRAM and not sure if I am misunderstanding or not. It seems to me ROM cartridge with 64 bit bus would have matched 64 bit bus of Jaguar's FPM DRAM thus is latter also clocked at 13.3MHz because only such makes sense for bandwidth of FPM DRAM to be meager 100MBps since if it matched clock speed of Tom and Jerry then it should have been 200MBps bandwidth while 486DX-33MHz with its 32bit bus has 100MBps bandwidth with FPM DRAM.

Edited September 6 by laymanpigeon

[+Biff Burgertime]

3 hours ago, laymanpigeon said:

I am just looking around FPM DRAM and not sure if I am misunderstanding or not. It seems to me ROM cartridge with 64 bit bus would have matched 64 bit bus of Jaguar's FPM DRAM thus is latter also clocked at 13.3MHz because only such makes sense for bandwidth of FPM DRAM to be meager 100MBps since if it matched clock speed of Tom and Jerry then it should have been 200MBps bandwidth while 486DX-33MHz with its 32bit bus has 100MBps bandwidth with FPM DRAM.

Sounds to me like you know exactly what you're talking about. You should have more confidence in your random numbers/acronyms!

[laymanpigeon]

1 hour ago, Biff Burgertime said:

Sounds to me like you know exactly what you're talking about. You should have more confidence in your random numbers/acronyms!

None of those are mine to begin with such as with FPM DRAM acronym and bandwidth of memory being 100 megabytes per second considering 486 is limited to 32 bit data bus width.

 

Jaguar with Tom having 64 bit DRAM controller in theory could have access of 212MBps bandwidth from four 16 bit FPM DRAM chips yet it is paltry 106MBps.

 

Though not out of possibility that FPM DRAM may have been at 25MHz clock in that example since here stated 25MHz is 100MBps and 33MHz is 133MBps which still would mean FPM DRAM on Jaguar is clocked at meager 13.3MHz that matches frequency of Motorola 68000 along half clock speed as either Tom and Jerry.

 

 

[42bs]

Someone should fix the specs page 🙂 

Quote

- 24-bit "True Color" display with 16,777,216 colors simultaneously (additional 8 bits of supplimental graphics data support possible).

or

Quote

- Blitter (processor #3)
- 64-bit RISC architecture

Or TOM

Quote

- Can read 64 bits of data in one instruction

 

Edited September 7 by 42bs

[+Stephen]

On 9/6/2024 at 7:44 PM, laymanpigeon said:

not sure if I am misunderstanding

Yes.  Definitely, positively, yes you are.

[laymanpigeon]

55 minutes ago, Stephen said:

Yes.  Definitely, positively, yes you are.

So go on... I am sure you can explain.

Edited September 8 by laymanpigeon

[42bs]

On 9/7/2024 at 7:29 AM, laymanpigeon said:

Jaguar with Tom having 64 bit DRAM controller in theory could have access of 212MBps bandwidth from four 16 bit FPM DRAM chips yet it is paltry 106MBps.

From THE docs:

Specifies the DRAM Speed. The page mode cycle time is always two clock cycles

 

And DRAM specs:

 

Edited September 8 by 42bs

[+CyranoJ]

28 minutes ago, 42bs said:

From ...

 

You are either correcting a dumb version of the chatbot from last year, or a person acting like it.  Either way.....

[laymanpigeon]

Or maybe consider that you two are in reality dumb.

 

EDO was rated for 40 MHz maximum clock rate, 64 bits of bus bandwidth, 320 MBps peak bandwidth and ran at 5 volts. It was tangibly faster than the older FPM DRAM that had only 25 MHz max clock rate and 200 MBps peak bandwidth. However, it was superseded by the faster SDRAM starting in 1996, after only two years of major use.

 

 

These graphic cards use FPM DRAM and have 64 bit memory bus width for maximum of 240 MB/s. 

https://videowww.vgamuseum.info/

Quote

 

Some of these graphic cards are from 1994 and that is not even a year after release of Jaguar.

 

Yet they have more than double bandwidth with same 64 bit wide memory bus as Jaguar that only has 106 megabytes per second bandwidth when these graphic cards have up to 240 megabytes per second bandwidth and I am suppose to believe you people that with 64 bits of Fast Page Mode DRAM only provides 106 MBps ?

 

Anyway interesting resources on various DRAM's:

 

https://www.cecs.uhttps://user.eng.umd.edu/~blj/talks/DRAM-Tutorial-isca2002.pdf

https://www.bunniestudios.com/bunnie/dramfaq/dram-workstation.pdf

https://thg.ru/mainboard/19981024/

ci.edu/~papers/mpr/MPR/ARTICLES/100605.pdf

https://www.ardent-tool.com/memory/pdf/sdramart.pdf

https://ctho.org/toread/forclass/18-741/2/SDRAM_Przybylski.pdf

 

 

Edited September 8 by laymanpigeon

[+CyranoJ]

20 minutes ago, laymanpigeon said:

Or maybe consider that you two are in reality dumb.

 

 

 

Show us your creations?

 

[42bs]

29 minutes ago, laymanpigeon said:

Some of these graphic cards are from 1994 and that is not even a year after release of Jaguar.

After googling GFX cards, google for Moore's law

 

Edited September 8 by 42bs

[DEATH]

Another comment just to say that if the Jaguar has a minimum access of 2 cycles to the DRAM in FPM mode it is because it was designed to run at 40Mhz (at least the original chipset). So 2 cycles at 40Mhz = 50ns and at the time the average access time in fast page mode for a 70ns access time DRAM was 45ns (50ns for a 80ns DRAM).

Also, it has always been stupid to measure the speed of asynchronous DRAMs in Mhz since their operation is based on propagation times only and therefore in "time" and not in clock cycles since the propagation time of a DRAM is invariable while the clock depends on the system on which they are installed..

[DEATH]

Oh I see some comments that.....
So, the maximum bandwidth of a FPM memory in fast page mode depends only on its minimum access time in Fast Page Mode, and on how the system on which it is located works of course. As said in my other comment the average time in Fast Page Mode was typically 45ns for a 70ns memory or 50ns for an 80ns memory. There were more (100ns for example) and there were less too (60ns or 50ns from my memory...) with Fast Page Mode times relative to these access times. Still from my memory for the fastest it had to go down to 40ns or 35ns. Which gives a theoretical maximum speed in Fast Page Mode at 35ns of 28.57Mb/s or 228.56MB/s on 64bit.
The Jaguar cannot reach this bandwidth because it uses FPM memory at 70ns or 80ns (so between 45ns and 50ns in Fast Page Mode) and its clock is at 26.6Mhz. Which is too fast to perform Fast Page accesses at a single cycle and must therefore do at least 2 cycles.
If the video cards mentioned are really capable of having a maximum bandwidth of 240MB/s on 64bit with FPM RAM and a 60Mhz clock, then this implies that they used FPM memory with a Fast Page Mode at maximum 33ns!! and 2 access cycles in Fast Page mode at 60Mhz. This may have existed (it's a long time ago), but this memory should not have been cheap!
 

When you don't know how a technology works, when you don't understand the numbers used, I think the best thing is to say nothing or at least to stay humble and not insult others. Otherwise it's the best way to look like an idiot.

Edited September 9 by DEATH

[DEATH]

NOTE, that they could also have used some sort of dual bank system to divide the frequency and access time by 2 and multiply the bandwidth by 2, but frankly I won't go looking to see what expensive technology they implemented to achieve this speed.

[laymanpigeon]

Merci @DEATH for your wonderful commentary on FPM DRAM and bandwidth of it !

 

I was trying to find information about it and you answered concisely, anyway I stumbled upon commentary on some forums that 60ns FPM DRAM was available in 1991 and same year EDO DRAM became available. Since EDO DRAM was evolution of FPM DRAM then they could work with each others memory controller though of course there would be no benefits for using in controllers designed for them respectively.

 

There was no clear interest about EDO DRAM with Atari's engineers even with Jaguar II when they were improving the hardware or at least redesigning the processors while maintaining full backward compatibility. 

 

To note is in 1992 development of SDRAM was completed with it entering production hence general availability in 1993. SDRAM in 1994 was used in 32X add-on and Saturn home console with former had two 256 kilobytes and latter 1 megabyte as SDRAM or SGRAM while Sony's PlayStation replaced VRAM with latter in 1995.

 

Well Nintendo went with RDRAM for 64 in effort to get highest performing hardware at lowest possible cost, other architecture decisions did not help that are worse off than Jaguar practically.

Edited September 9 by laymanpigeon

[DEATH]

Ahem... yeah...
Whatever.
Atari's strategy has always been (apart from 1 or 2 exceptions, like the TT for example) to offer the most powerful products at the lowest possible price. "Power without price".
At the time of the Jaguar, any technology other than FPM or similar (like nibble DRAM for example) was still way too expensive for the imposed costs and in any case the dates do not match. Superior products were not yet released.
EDO for example was only used from 1995, and even if SDRAM appeared a little before and ended up pull the rug from underneath the feet of BEDO (the successor of EDO which could compete with SDRAM), it was way too expensive for a machine like the Jaguar and was not ready when it was designed anyway.
In the Jaguar everything is a question of cost, especially compared to the initial version of the Chipset. A question of balance between cost and power more precisely. The frequency of its main clock and everything that comes from it (including DRAM access) is a question of balance between cost without "too much" impact on performance. If it had been able to keep the planned frequency of 40Mhz the Jaguar could have had a DRAM FPM bandwidth in 70ns or 80ns (access time which was also a question of cost) of 160MB/s in 64bit single channel. This is the maximum possible with all these criteria.
Note that TOM has 2 RAS/CAS channels that could possibly be used to make dual transfer channel and double the bandwidth. But to my knowledge I do not believe that the internal logic was planned for that, not to mention the complexity and the additional cost at the motherboard level.
In any case, the comparison with the products like graphics cards and consoles like 32x or saturn as mentioned is irrelevant and completely off topic. The prices are not comparable and the technology used was not ready at that time or was still overpriced. As a reminder, the Jaguar came out at the end of 1993 and its design dates from "well before".

Finally, do the math, no the Jaguar can't do better than 106MB/s with 70ns FPM DRAM, 26.6Mhz clock, 2 FPM access cycles, 64bit single channel. No machine with the same specs could do better. 26.6/2 * 64bit/8 = 106.4MB/s
If a supposed 486 can do ~100MB/s with a 32bit BUS, it's because there is something else behind it, such as a dual channel for the memory allowing it to perform a transfer each cycle in burst mode for example or a high speed FPM access time (40ns in Fast Page Mode). There is no magic.

[laymanpigeon]

17 hours ago, DEATH said:

Atari's strategy has always been (apart from 1 or 2 exceptions, like the TT for example) to offer the most powerful products at the lowest possible price. "Power without price".

Makes sense from perspective of marketing and economics to pursue such strategy yet from engineering standpoint it can be described as unrealistic in consideration of technology available compared to expectations of upper management to make it reality somehow when such is beyond realistic, a fantasy.

 

Similar happened with Nintendo 64 when price point was also 250USD in 1996 that is lower than Jaguar in 1993 when accounting for inflation. RDRAM was chosen because of much higher bandwidth per pin than any other DRAM and CPU had to go through GPU to access system memory that added to already very high latency thus in end similar situation to Jaguar with bandwidth starvation.

17 hours ago, DEATH said:

At the time of the Jaguar, any technology other than FPM or similar (like nibble DRAM for example) was still way too expensive for the imposed costs and in any case the dates do not match. Superior products were not yet released.

You are wrong as it does match, very much so.

17 hours ago, DEATH said:

EDO for example was only used from 1995, and even if SDRAM appeared a little before and ended up pull the rug from underneath the feet of BEDO (the successor of EDO which could compete with SDRAM), it was way too expensive for a machine like the Jaguar and was not ready when it was designed anyway.

Development of Pentium P5 processor and Socket 4 motherboard were finished in 1992 and launched in March of 1993 do support EDO DRAM. Fujitsu manufactured 2 megabits 60ns and 70ns EDO model 81C4256A EDO chips in 1992 and same year 16 megabits SIMM were sold by Shin Etsu Handotai company containing 8 of 81C4256A-60 chips. Samsung was selling 16 megabits SIMM model KM48SL2000 in 1993.

17 hours ago, DEATH said:

In the Jaguar everything is a question of cost, especially compared to the initial version of the Chipset. A question of balance between cost and power more precisely. The frequency of its main clock and everything that comes from it (including DRAM access) is a question of balance between cost without "too much" impact on performance. If it had been able to keep the planned frequency of 40Mhz the Jaguar could have had a DRAM FPM bandwidth in 70ns or 80ns (access time which was also a question of cost) of 160MB/s in 64bit single channel. This is the maximum possible with all these criteria.
Note that TOM has 2 RAS/CAS channels that could possibly be used to make dual transfer channel and double the bandwidth. But to my knowledge I do not believe that the internal logic was planned for that, not to mention the complexity and the additional cost at the motherboard level.

There is contradiction with design of Jaguar since it could have operated without 68000 yet was added thus added cost as too for motherboard that became more complex as result in order to support it. On top of that it was 16MHz 68HC000 that cost same as 16MHz 68EC020 while only 5USD less than 68020 in 1st quarter of 1993.

 

Removing 68HC000 would save up 15USD. Jerry having own 64 bit bus and 16 megabits would be ideal. Doubling memory bus to 128 bits and adding four more 4 megabits FPM DRAM chips would certainly improve performance more than increase in cost of console. If we go by averages that would be 44~48USD in 1992~1993. 20 percent increase in price of console for at least 25 percent improved performance overall.

17 hours ago, DEATH said:

In any case, the comparison with the products like graphics cards and consoles like 32x or saturn as mentioned is irrelevant and completely off topic.

It is not completely off topic unless you are not aware at all that SNK has ported and released Neo Geo games on Sega Saturn that Jaguar has competed against as was too 32X add-on that with Genesis the total cost is 250USD. Same as Jaguar.

17 hours ago, DEATH said:

The prices are not comparable and the technology used was not ready at that time or was still overpriced. As a reminder, the Jaguar came out at the end of 1993 and its design dates from "well before".

Design of EDO came well before development of Jaguar was done and was produced in 1992 while EDO is evolution of FPM thus compatible with memory controller units of each other. Of course using one in other has no benefit, though it would not require radical redesign of DRAM controller unlike if it was SDR/SDRAM.

17 hours ago, DEATH said:

 

 

[42bs]

While googling chip prices you should also check how much more complicated a128bit bus is compared to a 64bit one.

 

And then try to find out what options for RAM and CPU and ASIC where available in 1990. Only those are relevant for the design of the Jaguar. Not 1991 or later.

 

No designer relies on possible future chips.

[+Sauron]

2 hours ago, 42bs said:

While googling chip prices you should also check how much more complicated a128bit bus is compared to a 64bit one.

 

And then try to find out what options for RAM and CPU and ASIC where available in 1990. Only those are relevant for the design of the Jaguar. Not 1991 or later.

 

No designer relies on possible future chips.

No point in arguing with him, he's just yet another smartass throwing numberwang around without having the slightest idea of what it takes to put together a hardware product from scratch, like most of the other idiots who come on here and do the same thing. No matter how much you try to explain, they'll never understand that putting together a hardware product involves more than just adding up the total cost of the parts.

 

[laymanpigeon]

Hello there, can you go somewhere else projecting yourself instead of jumping to own desired conclusions (?) about me thus inadvertently lie about me. There are so many of people like you that can do that along throw insults because that is only thing people like you are capable off. No decent culture, no civility and let alone not doing to others you would not want not upon yourself. I guess you for example took worst traits of your parents. I pity your siblings if you have any.

Edited September 10 by laymanpigeon