ZX81 faster than the TS1000?

[ts_mike]

So, while I'm waiting for my ULA heatsink to come in (thank you so much for your help, BTW! ), I've been browsing around the net for some info, and I stumbled across this Timex Sinclair 1000 site which is now defunct, but being archived by another sinclair fan:

http://forum.tlienhard.com/TS1000/www.t ... sembly.htm

According to his benchmark testing, the 60Hz NTSC(US) TS1000 is significantly slower than the 50Hz PAL(UK) ZX81.

Is that claim true?

I've noticed that the emulators I've used have seemed much "snappier" than my old TS1000 ever was, but I always assumed that was due to emulation effects. Has the above claim ever been verified by anyone else? The theory makes sense--the CPU has to refresh the screen 10 more times a second and has less time in VBLANK to do actual processing (in slow mode, at least). I just never thought it might be a real effect. Thoughts?

[Andy Rea]

Yes it's true, in both machine the CPU runs at 3.25Mhz (3250000 cycles a second) it takes approx 40000 cycles to generate the display and 1700 for the Vsync/keyboard scan so....

50Hz = 41700 *50 = 2085000 , leaving approx 1165000 for user program per second

60Hz = 41700 *60 = 2502000, leaving approx 748000 for user program per second

so the 60Hz machine runs user code at approx 64% the speed of a 50Hz machine

All of the above are Rough estimates off my head....

Regards Andy

[RetroTechie]

Confirmed. On the ZX81 (and thus, TS1000) there's an ULA pin to select 50 or 60 Hz display refresh. If you change its voltage, the ZX81 generates fewer blank lines at top & bottom of the screen to achieve 60 Hz display refresh. This is done by reading a bit from an I/O port, which is used to set a system variable called MARGIN. This in turn is used by the display routines in ROM (check my ZX81 ULA project for details if needed).

BASIC and assembly software is run during display of those blank lines, so fewer of them means a smaller % of CPU time spent on executing software. The difference is definitely notable, even without a calculator or stopwatch at hand. In both cases the CPU speed is the same, just @ 60 Hz a smaller % of it is left for executing BASIC etc.

[sirmorris]

Yes. What they said.

[ts_mike]

Bummer... No wonder Sinclair struggled in the U.S. (I'll have another posting on that subject in a little bit)

[Bill H]

http://forum.tlienhard.com/TS1000/www.t ... sembly.htm

According to his benchmark testing, the 60Hz NTSC(US) TS1000 is significantly slower than the 50Hz PAL(UK) ZX81.

Is that claim true?

Very tyrue - I remember doing that benchmark and making the webpage

Bill H

[Bill H]

The most interesting part was the hacked ts1000 running at 50 frames being faster than the zx81. Maybe those in the UK should be using the ts1000 instead ?


Bill H

[RetroTechie]

Ran the BASIC test(s) in above link on my ZX81 clone, and results are consistent with ZX81 (50 Hz) and TS1500 / 16K (60 Hz), just 1 frame less for some tests. When comparing speeds, note that frames must be divided by different numbers to obtain wall clock time (and 50 / 60 Hz aren't the exact frame rates, either).

Overall conclusion: in SLOW mode, a ZX81 (or TS1000) @50 Hz runs programs about 1.5x faster than same machine @60 Hz. If you keep the frame rate in mind, that practically matches the different values for system variable MARGIN (31*60 vs. 55*50).

ZX81 and TS1000 are the same machine hardware-wise, so result for that TS1000 'hacked' to run @50 Hz is an odd one... Remaining variables that might have effect: ROM, and DFILE expanded or not (1K, 2K or 16K RAM). Perhaps that 'hacked' TS1000 had an older ROM version or something. For example a bugfixed ROM might affect the time needed to execute specific BASIC statements.

[Bill H]

ZX81 and TS1000 are the same machine hardware-wise, so result for that TS1000 'hacked' to run @50 Hz is an odd one... Remaining variables that might have effect: ROM, and DFILE expanded or not (1K, 2K or 16K RAM). Perhaps that 'hacked' TS1000 had an older ROM version or something. For example a bugfixed ROM might affect the time needed to execute specific BASIC statements.

I would have to pull the hacked ts1000 out of storage to find out but I think (and I could be wrong) there was only 1 rom version for the ts1000.

Bill H

[Crayon21]

I guess the commercial lied then

"power to learn at the speed of light" they said