Novag Aquamarine RISC II 26.6 MHz

Novag Aquamarine RISC II 26.6 MHz

		Hersteller	Novag
		Markteinführung	1998
		CElo	2032
		Programmierer	Kittinger, David


Prozessor	Takt	RAM	ROM
Hitachi H8/3258 (rom B84) Hitachi H8/3214 (rom A06)	26,6 MHz (internal 13,3 MHz) 16 MHz	1 KB	32 KB
Rechentiefe	BT-2450	BT-2630	Colditz
16 Halbzüge	-	1958 / 1972	-

Bibliothek	13.000 Halbzüge
Display	2x 4-stellige 7-Segmentanzeige
Spielstufen	56

Zugeingabe	Drucksensoren
Zugausgabe	16 Rand-LEDs

Einführungspreis

Prozessortyp	8 Bit, Singlechip
Stromversorgung	Netz = 9V/0,3A DC (Plus außen!) / Batterie = 6 x AA
Maße	23,5 x 28,5 x 2 cm

Verwandt	Novag Zircon II, Novag Jade II, Chess Wizard IQ V

Sonstiges
wurde ausschließlich in Spanien verkauft

Infos

Introduced by Novag in 1998, this chess computer was apparently distributed only in Spain.

It has a program very similar to the built in previous Novag chess computers, but with a simpler plastic housing. What is most striking about this model is the ostentation of the "supposed" processor speed (26.6 MHz), being indicated in large letters on the box, and even printed on the chess computer case.

For Novag "MHz" possibly was just a marketing strategy: the more the better. In fact, for a long time, Novag printed in the boxes and manuals the frequency of the quartz crystal, instead of the actual system clock. So far nothing new.

But it was later marketed with a different MCU and quartz crystal, probably because the previous one reached EOL (End-of-life). In this new version ("2nd version" from now on) the quartz crystal is not 26.6 MHz but 16 MHz, and it plays better than the original one (it performs calculations 20% faster). Despite this, the box continued to indicate 26.6 MHz.

Some, but not all, of the 2nd version units were sold with "Siglo XXI" (XXI Century) printed on the computer.

Review

Some of these computers, internally examined, reported the following data:

Version	MCU	H8 Family	Quartz Crystal	VCC
1st version	HD6433258B84F	H8/325	26.6 MHz	5 V
2nd version	HD6433214A06F	H8/3214	16 MHz	5 V

Data reported by Hitachi for that microcontroller family:

H8 Family	ROM	RAM	Recomm. Max. Clock at 5V
H8/325	32 KB	1 KB	10 MHz
H8/3214	32 KB	1 KB	16 MHz

Block Diagram of Clock Signal Generator for this two microcontrollers:

Clock signal in H8/325 is divided by 2 automatically. Therefore, in computers equipped with this microcontroller, the system clock frequency (Ø) is 1/2 of the frequency of the quartz crystal. With H8/3214 division by 2 is optional (selected by software). Therefore, the system clock (Ø) may be equivalent to the quartz crystal, or 1/2.

Whereof we can obtain this table:

Version	H8 Family	Quartz Crystal	Divider	System Clock (Ø)
1st version	H8/325	26.6 MHz	/2	13.3 MHz
2nd version	H8/3214	16 MHz	/1	16 MHz

Overclocking? In the "original" Aquamarine maximum clock frequency recommended by manufacturer is exceeded, but this is possible. The manufacturer (Hitachi) sets 10 MHz as the maximum (recommended) frequency because they certifies that, under these conditions, 100% of the manufactured chips will work fine in a wide range of ambient temperatures. But it's possible that integrator (Novag in this case) decide to select those chips that work correctly even exceeding the recommended clock frequency by a certain percentage. The increase in the "original" Aquamarine is 33%

In any case, these values are confirmed when performing various comparative test between the two computers.

Testing

When the same problems are proposed on both computers, logging the time required for each problem, it's confirmed that 2nd version (also "Siglo XXI") always solve approximately 20.3% faster.

Illustrative Example:

After performing Reset (ACL) on both computers, we select G5 Level (Mate in 5)

The next position is established, and solving time is counted:

Computer	Solving Time
Aquamarine RISC II 26.6MHz (1st version)	15min 12sec
Aquamarine RISC II 26.6MHz (2nd version & "Siglo XXI")	12min 37sec

2nd version is 20.4% faster

In Blitz and Mate Search levels the same result is always obtained: 2nd version performs calculations about 20.3% faster. But this advantage is not noticeable in long games (with more than 3 minutes per move, for example) because both seem to have the same program, and eventually both make the same decisions with exactly the same valuation per move.