MSX Overclocking - Disconnect WAIT pin

by msd on 13-11-2005, 16:31
Topic: Hardware
Languages:

Marcel Delorme managed to increase the speed of his Philips NMS8245 with 10% by disconnecting one pin from the MSX-ENGINE. In MSX computers, there is an extra wait state in the M1 cycle of the Z80. For every memory read and write, the MSX-ENGINE sends a WAIT signal to the Z80. Curious whether or not his MSX could handle the extra speed, Marcel Delorme decided to disconnect the wait output (pin 41) from the MSX-ENGINE.

The result? An MSX2 that still boots! Using a program called CLKSPEED a 10% increase in performace can de measured. Another benchmark, txtspeed.com, also shows an increase of performance from 454 to 526 characters per second. Although it's not sure whether every MSX computer can handle the increased speed (for a long time), it's surprising this has - to our knowledge - never been tried before.

Comments (12)

By msd

Paragon (1510)

msd's picture

13-11-2005, 16:37

It's not overclocking.. it's only removing the wait Tongue

By HansO

Paladin (672)

HansO's picture

13-11-2005, 16:41

Pin 41 of the S3527 MSX engine. So only applicable to those MSX machines with that IC inside (Philips MSX 2 and later VG8020). If and how this might work with other engines such as the S9185 of the turboR engine is interesting to know.
Be carefull with that IC btw, its SMD technique. An external switch to enable/disable this feature is advisable.

This might make cassette I/O also less reliable.

By snout

Ascended (15187)

snout's picture

13-11-2005, 16:46

Aren't you 'overclocking' the memory in this case then? Either way, I hope/assume 'overclocking' is a generally accepted term for making a computer faster. Tongue

By msd

Paragon (1510)

msd's picture

13-11-2005, 16:51

The clock stays the same.. there is just no additional wait when read/writing ram. I suspect that at the time of the MSX1 slow rams very still very common (250nS) and once they added the wait they where stuck with it. We could always ask Nishi why they did it.

By POISONIC

Paladin (1012)

POISONIC's picture

14-11-2005, 13:10

i think when they developed the msx standard the faster ram was to expancive..... maybe they choose to lower the msx production cost.. they used slower ram and needed the wait state to solve the problem......

By msd

Paragon (1510)

msd's picture

14-11-2005, 13:28

Yes... but wait state is as far as I know not needed with ram/rom below 250nS

By Grauw

Ascended (10713)

Grauw's picture

15-11-2005, 00:17

Anyways, the MSX2 is technology that is about 4 years older than the MSX1. If they would have removed the wait state, it would break existing content. Backwards compatibility was more important than such a slight speed gain.

Makes sense to me.

Nice one, Marcel!

~Grauw

By msd

Paragon (1510)

msd's picture

15-11-2005, 08:12

Grauw: Don't you mean 4 years newer Tongue

By bacterion

Paragon (1326)

bacterion's picture

15-11-2005, 11:57

4 years?

By BiFi

Enlighted (4348)

BiFi's picture

19-11-2005, 12:18

the difference between MSX1 and MSX2 is about 2 years...

By NYYRIKKI

Enlighted (6033)

NYYRIKKI's picture

23-11-2005, 16:04

Well... 2 years, 4 years what ever... There has been 7MHz mod instructions, complex Z380 projects etc. but it took about 20 years to discover, that we can just disconnect that stupid wait signal to get more speed. Can't say we are too bright hobbyists. Tongue

By Fudeba

Expert (113)

Fudeba's picture

25-11-2005, 11:53

Ademir Carchano spoke about this thing years ago. In fact, once I talked with hit about this matter, and I understood that computers designed by Ademir Carchano (CIEL2+, Expert3 and ACE001) disable this wait when turbo is enabled (7MHz or 10MHz, depending on the machine). Once turbo mode require faster memory chips and break compatibility with those programs that rely on MSX original timings (3.57MHz with the extra Wait on M1 cycle) there is no reason to keep the extra wait in this mode (7MHz/10MHz). Tongue