MSX Turbo-R : Unexpressed potential?

Page 3/3
1 | 2 |

By PingPong

Prophet (3793)

PingPong's picture

24-10-2021, 08:51

To be more clear, it is not that the v9938 was poorly designed. It was the TMS that was full of stupid design flaws. Yamaha had to deal with this heritage flaws to maintain compatibility. The v9990 is clear and well designed because yamaha did not have to bother with all previous garbage that have its roots in the crappy tms vdp.

Plus the amiga has a different architecture, but this is not the point. You can left the tiles approach using a full bitmap organization, and this is a more flexible approach, but you need some sort of graphic acceleration (the blitter)

By gdx

Enlighted (4811)

gdx's picture

24-10-2021, 11:20

PingPong wrote:

To be more clear, it is not that the v9938 was poorly designed. It was the TMS that was full of stupid design flaws. Yamaha had to deal with this heritage flaws to maintain compatibility.

I exaggerate a little, it is not entirely poorly designed but Yamaha should have done better. The commands at least three time faster and use the port #9A to quickly read some status bits as CE, EO, FH, TR, HR and VR for example. All this has nothing to do with the maintaining compatibility.

By ARTRAG

Enlighted (6567)

ARTRAG's picture

24-10-2021, 11:19

Actually I see it a a problem of balance between CPU speed and video bandwidth/command speed
If I/O and vdp commands are slow, the CPU is forced to spend its time waiting
On msx1, thanks to tile modes, you can update the screen in one frame (almost) without problems
On upper models, in bitmap mode, there is no simple way to update the whole screen in one frame but on msx2 and 2+ the z80 can push data at full speed. On TR you waste the CPU power waiting.
In short, the v9958 is too slow to work with the R800 and the lack of DMA or other ways to disengage the CPU from vdp I/O makes things worst.

By ARTRAG

Enlighted (6567)

ARTRAG's picture

24-10-2021, 11:27

I recall also that the vdp bandwidth in screen 5 and 6 could have been twice faster if only the V9938/58 would have used two interleaved ram backs in all its modes, as it happens in screen 7 8 and upper.
This would have required to drop configurations with only 64KB of vram because based on a single ram chip...

By PingPong

Prophet (3793)

PingPong's picture

24-10-2021, 11:43

the original design used 99 to read status register, they maintained the protocol maybe under the pressure of "one port" for status register.
About the command three times faster, the vdp is already clocked @21Mhz. Was a great speed for the era.
Due to timing constraints it was not so easy to speed up things, but a little more would have been possible.

Imho the thing that is worse (due to compatibility) is the sprite 2 subsystem. it should have been completely different from the tms one.
the idea of having a SAT scanline based for sprite attributes is a true source of troubles, because you need to manage the early clock (for example) at scanline level. Plus you have to move 512 bytes of vram if you do sprite plane rotation.

What is more stupid is the MAGIC Y value, a true non sense even on tms and absolutely crappy on V9938 with the ability to scroll the screen via vertical register. Required us to do a check to see if y is the magic value and change it in order to avoid sprite disappearing. They should have removed this at all in msx2 modes.

By gdx

Enlighted (4811)

gdx's picture

24-10-2021, 11:51

Absolutely, 64K mode was stupid. I think Japaneses have chosen Yamaha out of chauvinism.

By PingPong

Prophet (3793)

PingPong's picture

24-10-2021, 16:44

gdx wrote:

Absolutely, 64K mode was stupid. I think Japaneses have chosen Yamaha out of chauvinism.

AFAIK there is no msx2 with 64K VRam

By mars2000you

Enlighted (6014)

mars2000you's picture

24-10-2021, 16:51

PingPong wrote:
gdx wrote:

Absolutely, 64K mode was stupid. I think Japaneses have chosen Yamaha out of chauvinism.

AFAIK there is no msx2 with 64K VRam

Ehmmm? You should take a look here:

https://www.msx.org/wiki/VRAM

By ARTRAG

Enlighted (6567)

ARTRAG's picture

24-10-2021, 16:55

I'm not aiming only to the 64KBt limit but to the fact that interleaving two memory chips in all modes would have doubled the VRAM access speed in screen 56. This is done only in screen 7 and 8, but would have been possible also in lower modes dropping the support to 64KB configurations.

By journey

Hero (537)

journey's picture

25-10-2021, 11:06

I understand very little on a technical level ... but I think I understand that a V9990/V9978 more than on a hypothetical MSX3, would have been much more useful if it had been mounted directly on the MSX Turbo-R...

Page 3/3
1 | 2 |