an article about z80 affecting MSX computers was published

Page 1/7
| 2 | 3 | 4 | 5 | 6

Par litwr

Resident (56)

Portrait de litwr

01-09-2018, 09:44

It is here. I will be glad to get more information about z80, corrections are also appreciated. This article is a part of CPU reviews - https://litwr.livejournal.com/.
And BTW what is the best web site dedicated to z80?
Thank you

!login ou Inscrivez-vous pour poster

Par ali6x944

Master (198)

Portrait de ali6x944

01-09-2018, 12:33

without a duote Z80.info

Par PingPong

Enlighted (4136)

Portrait de PingPong

01-09-2018, 14:08

about the z80 there are a lot of erroneus assertions.
for example the article say about the z80 not running at effective speed on some machines due to video memory contention. It does appear as a problem related with z80. Effectively it is not specifically related to any processor.
For example on the c64 the video memory contention limited the 6502 clock to be at 1Mhz at most.
And the 6502 suffered from VIC-II cycle steal as happened on z80 on the ZX spectrum with ULA or the Amstad CPC.

In the c64/6502 the problem is so heavy that designers, in order to achieve more speed on c128, used a 6502 like processor @ 2Mhz but they were forced to another video architecture.

Par litwr

Resident (56)

Portrait de litwr

01-09-2018, 23:23

@ali6x944 Thanks but are there any forums dedicated to z80?

@pingpong

Quote:

It should also be noted that some of the ticks in most systems with z80 or 6502 are taken from the processor to the schemes for generating the video signal, for example, because of this, the popular Amstrad CPC/PCW computers have the effective processor frequency of 3.2 MHz, not full 4.

I can even add that some 6502 systems give up to 50% of their clocks to the video.

Par RetroTechie

Paragon (1563)

Portrait de RetroTechie

06-09-2018, 22:14

PingPong wrote:

And the 6502 suffered from VIC-II cycle steal as happened on z80 on the ZX spectrum with ULA or the Amstad CPC.

ZX Spectrum: only 16K version. 48K ZX can run code @ full speed if it's located outside the RAM block used for the display.

litwr wrote:

@ali6x944 Thanks but are there any forums dedicated to z80?

Don't think so. People are attached to specific computer systems, not so much the CPU's in there. But for example the (English) ZX81 forum has several users that know all ins & out of the Z80 CPU (possibly myself included ;-).

As silly as it sounds: I'd say the Z80 CPU has had a considerable influence on my life (at least the hobby part). So when I die and get cremated, feel free to toss a few early 80's manufactured, original Zilog NMOS Z80's in there 'cause I think I deserve to take a few of them with me lol...

Par PingPong

Enlighted (4136)

Portrait de PingPong

06-09-2018, 23:07

RetroTechie wrote:
PingPong wrote:

And the 6502 suffered from VIC-II cycle steal as happened on z80 on the ZX spectrum with ULA or the Amstad CPC.

ZX Spectrum: only 16K version. 48K ZX can run code @ full speed if it's located outside the RAM block used for the display.

We know, due to a trick in bus design.
Anyway my point was to make clear that the shared memory problem does not have nothing to do to a specific cpu, z80 or 6502, as suggested by the article.
msx. no need to wait at all due to vram because it is not connected to cpu bus.
(sadly wait on m1 cycle and vram access slow)

Par HansO

Paladin (672)

Portrait de HansO

07-09-2018, 10:23

Quote:

And the 6502 suffered from VIC-II cycle steal as happened on z80 on the ZX spectrum with ULA or the Amstad CPC.

Quote:

I can even add that some 6502 systems give up to 50% of their clocks to the video.

The 6502's clock is, like the 6800, a bit special and allows this without any penalty or speed down.
50% of a clock cycle on a 6502 no memory is accessed, and other peripherals (or another 6502!) can use that to access memory. The VIC IC in Commodore's is an example of this and allows any part of RAM to be accessed by the CPU as videoram. So no suffering on the C64 or VIC-20!
It also does not limit overclocking the CPU frequency as long as the other IC's can keep up with access speed of RAM. If the CPU clock also is used for the video timing, then of course overclocking is not possible.

Par PingPong

Enlighted (4136)

Portrait de PingPong

07-09-2018, 20:15

HansO wrote:

videoram. So no suffering on the C64 or VIC-20!

Please study a bit more in depth the architecture of C64.
Checkout the "Bad line" topic for c64. in the worst case the cpu is halted for an entire scanline that is 64us!
Plus the VIC-II is used on c64 not in VIC 20. Different bandwidth requirements.

Par litwr

Resident (56)

Portrait de litwr

08-09-2018, 10:55

PingPong wrote:

Anyway my point was to make clear that the shared memory problem does not have nothing to do to a specific cpu, z80 or 6502, as suggested by the article.
msx. no need to wait at all due to vram because it is not connected to cpu bus.
(sadly wait on m1 cycle and vram access slow)

Let repeat again: where did you find this suggestion?
It is odd but ZX Spectrum is faster than later and more advanced MSX or Amstrad CPC/PCW.

Par RetroTechie

Paragon (1563)

Portrait de RetroTechie

09-09-2018, 08:21

MSX & ZX use same Z80 clock speed. ZX Spectrum has a simple architecture, and allows direct cpu access to VRAM. But cpu may be halted sometimes when screen output gets priority. MSX has a more complex / flexible / expandable memory layout. Which is likely the reason the M1 wait was introduced, to allow for that and slow ROMs. Also it moves VRAM out of the Z80 address space, which has pros & cons.

So if an application depends on frequent updates to VRAM (say, software sprites etc), then ZX Spectum (esp. 48K) is the faster machine. If otoh, your application flexes the VDP's muscle (like with big end bosses in Konami MSX1 games), then ZX is left in the dust.

In hindsight I think M1 wait on MSX was a design mistake. Sad And slot mechanism too complex for its purpose, making life of programmers harder than necessary. But that's history now, and we have a whole library of software that depends on a specific cpu speed (including M1 wait).

Par hit9918

Prophet (2932)

Portrait de hit9918

09-09-2018, 19:22

Quote:

In hindsight I think M1 wait on MSX was a design mistake

without M1 wait the memory acess is 1.5 times more harsh. given a certain memory spec, a z80 with M1 wait is FASTER! can clock it 1.5 times higher!

Page 1/7
| 2 | 3 | 4 | 5 | 6