an article about z80 affecting MSX computers was published

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

By litwr

Rookie (30)

litwr's picture

14-12-2018, 12:52

Thanks again. It will be interesting to have Z280 timings. It looks like Zilog was reluctant to publish this data for all its processors since Z8000. However this text testifies rather that Z280 was much slower than R800 with the same clock.

By Grauw

Ascended (8516)

Grauw's picture

14-12-2018, 14:17

litwr wrote:

However this text testifies rather that Z280 was much slower than R800 with the same clock.

Thanks for that article, a nice read.

But doesn’t the R800 have a 28.6 MHz clock? The article says the Z280 is about 1.5 times faster than a regular Z80 at the same clock speed, so if clocked at 28.6 MHz it would be about 12x faster… The R800 is about 6x faster than the Z80, taking into account that it is limited by the bus speed (let’s say the internal memory bus is 7.16 MHz), it could match up.

litwr wrote:

It will be interesting to have Z280 timings. It looks like Zilog was reluctant to publish this data for all its processors since Z8000.

The Z280 technical manual describes the timing in appendix E.

By litwr

Rookie (30)

litwr's picture

15-12-2018, 12:27

I am really perplexed... R800 uses about 7 MHz frequency in TurboR but its speed is like Z80 at 28 MHz (or even 42 MHz as you say). The article claims that Z280 is only 50% faster than the z80 at the same frequency. So it is a clear difference in numbers, 4-6 vs 1.5 times. I even more perplexed by the fact that z280 timings are published. Why don't just compare them with the corresponding for R800?
A man supposedly with Zilog background did an interesting post. So maybe all Z280 were just too buggy to a real use.

By Grauw

Ascended (8516)

Grauw's picture

15-12-2018, 15:46

Note that the “4-6 times as fast” I mentioned is the R800 at whatever clock frequency it runs at relative to a 3.58 MHz Z80. If we pose that the R800 internally runs at 14.3 MHz, then per-cycle it is ~1.5x the speed of a Z80.

So the question becomes, what clock speed does the R800 run at. Why do you think R800 runs at 7.16 MHz internally? I know that’s what people say, but it’s fed with a 28.6 MHz crystal. Why else provide such a high clock frequency, if the processor doesn’t run at it?

R800 timings are expressed in 7.16 MHz cycles, true, but it could very well be that lower frequency only applies to the bus (memory access) speed which quantises the timings. This is my suspicion.

The Z280 documentation says (§9.2):

Quote:

The frequency of the processor clock is one-half of the frequency of the external clock source or crystal. The processor clock can be further subdivided by a factor of 1, 2 or 4 to provide the bus timing clock, as specified by the contents of the Bus Timing and Initialization register (see Chapter 3). The bus timing clock is output by the MPU for use by the rest of the system.

I’m posing the theory that the R800 works in the same way, clocked at 28.6 MHz, internally running at 14.3 MHz, with a bus clock divider of 2 (7.16 MHz) for the internal memory, and a bus clock divider of 4 (3.58 MHz) for the external memory and I/O.

By litwr

Rookie (30)

litwr's picture

15-12-2018, 22:05

Thus you say that the information that R800 uses less cycles for instructions is incorrect, don't you? You say that it is an effect of higher internal frequency? I can't agree because we have R800 instruction timings which show that LR A,B executes for one cycle at R800 and 4 cycles at Z80, but ADC HL,BC takes 2 cycles at R800 and 15 at Z80. If you theory was true ADC HL,BC would take about 4 cycles at R800, not 2. IMHO 28 MHz could be used for video. 8-bit Commodores can use 14 or 16 MHz crystals when CPU works at less than 1 MHz.

By Grauw

Ascended (8516)

Grauw's picture

16-12-2018, 01:49

litwr wrote:

I can't agree because we have R800 instruction timings which show that LR A,B executes for one cycle at R800 and 4 cycles at Z80, but ADC HL,BC takes 2 cycles at R800 and 15 at Z80. If you theory was true ADC HL,BC would take about 4 cycles at R800, not 2.

I'm not saying R800 timings are identical to Z80 and only clocked faster. The Z80 has a 4-bit ALU, the R800 has a 16-bit ALU. It can execute 16-bit adds in one cycle where the Z80 needs several. So that argument says nothing about whether my theory is true or no.

If you want to compare then compare to the Z280 timings, which has a 16-bit ALU too. Though that would still only say anything about the similarity between those processors, not about how fast the R800 is internally clocked… for that we’d need to find instructions which have different timings depending on what precedes it, which could give evidence for it.

litwr wrote:

IMHO 28 MHz could be used for video. 8-bit Commodores can use 14 or 16 MHz crystals when CPU works at less than 1 MHz.

Check the turboR ST / GT service manuals, the clock is not divided before it enters the R800, nor used for anything else. Several MSX2 systems also clock the VDP at 21.5 MHz, but the frequency is divided by 6 to the normal 3.58 MHz before it's fed into the Z80. 28.6 MHz is not a useful frequency for the VDP, it needs 21.5 MHz and has a separate crystal in the turboR.

By gdx

Prophet (3091)

gdx's picture

16-12-2018, 08:19

It is not rare that the clock frequency is divided internally to get a well square frequency signal. It is the case for the PSG.

MSX turbo R Technical Hand book page 22 wrote:

従来のMSXクロック周波数は3.58メガヘルツで、turbo R のクロック周波数は7.16メガヘルツ。

The standard MSX clock frequency is 3.58 MHz, and it is 7.16 MHz for the turbo R.

Other documents say the same thing. In addition they say the statements are faster because R800 has a RISC architecture, and cycles are explained. There is no doubt to have. The real clock frequency of R800 is 7.16 MHz, Same for the BUS clock frequency.

By erpirao

Paladin (962)

erpirao's picture

16-12-2018, 12:38

bluemsx+msxturboR emulation:
using clkspeed.com
effective clockspeed: 17897700hz
standar (optional) : 3579545hz
relative performance: 499.9%

login.com
cpu: z80280
Mhz: 32.6184

By litwr

Rookie (30)

litwr's picture

16-12-2018, 13:02

@Grauw You have almost convinced me that R800 is just simplified Z800 but the other MSX guys say that it can't be true...

By mars2000you

Enlighted (5515)

mars2000you's picture

16-12-2018, 13:24

.

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