There was a nice discusion about the speed variation between z80 ti8x calcs. I will post the log for reference:
Code:
Here is a snip on testing and gathering data:
Code:
Photo of the calc, better than datamath!
If you have tested your calc's speed, stay tuned here as we may need to do some polling and data collection.
If you have anything to add that you feel is important, post here
Code:
[15:43] <KermM> that was on the 83+ calcs with an RC tank oscillator
[15:43] <AHelper0> KermM, ah. But that is inside the asic now, right?
[15:43] <KermM> it has a crystal oscillator
[15:43] <AHelper0> o_O
[15:43] <KermM> with a controllable divider
[15:43] <Runer112> what
[15:43] <AHelper0> oh nm
[15:43] <Runer112> but the CPU speed isn't controlled by a crystal timre
[15:43] <KermM> The 83+SE/84+/84+SE
[15:43] <AHelper0> forget about me here
[15:44] <Runer112> is it?
[15:44] <KermM> not a crystal timer, a crystal oscillator
[15:44] <Runer112> erm
[15:44] <IkariTari> which also drives the crystal timers
[15:44] <Runer112> what's the difference
[15:44] <KermM> correct
[15:44] <AHelper0> the osc. drives both
[15:44] <KermM> "crystal timer" is a construct that TI made up. It's an accurate timer that counts clock ticks
[15:44] <AHelper0> independantly
[15:44] <AHelper0> right?
[15:44] <KermM> the ticks come from the crystal
[15:44] <Runer112> huh
[15:44] <KermM> the crystal also is used for the clock rate
[15:44] <KermM> AHelper0, yes
[15:44] <Runer112> but there are modes that count accurate second intervals
[15:44] <KermM> correct
[15:44] <AHelper0> The Prizm does the same
[15:44] <Runer112> whereas the CPU ticks aren't on accurate second intervals
[15:45] <KermM> using the accurately-timed ticks from the crystal
[15:45] <KermM> on the TI-84+ they are, Runer112
[15:45] <AHelper0> the RTS clock drives the internal timers, the RTs unit and the CPU
[15:45] <Runer112> erm
[15:45] <AHelper0> *RTC
[15:45] <AHelper0> and *RTC
[15:45] <Runer112> then why do some calcs run at like 14.5MHz
[15:45] <Runer112> and others run at 17MHz
[15:45] <KermM> 
really?
[15:45] <Runer112> yes
[15:45] <AHelper0> TI sucks?
[15:46] <KermM> that makes no sense
[15:46] * AHelper0 runs
[15:46] <KermM> this bears further investigation.
[15:46] <Runer112> they all count seconds properly, but they run at different speeds
[15:46] <AHelper0> Were they from the same factory and date?
[15:46] <Runer112> like mine is the slowest 84+ on the face of the planet
[15:46] <Runer112> I measured it at 14.6MHz
[15:46] <saxjax> (C) [aeTIos] <_<
[15:46] <KermM> this is very odd. That would suggest that they either use the wackiest divider ever
[15:46] <Runer112> and I know some other people have really fast ones
[15:47] <KermM> or are actually using an RC tank for the clock speed, which makes NO sense
[15:47] <AHelper0> Probably me
[15:47] <KermM> Runer112, do you know if it gets slower on a lower voltage?
[15:47] <Runer112> it does slightly
[15:47] <Runer112> by I think a couple KHz
[15:47] <KermM> Then it's an RC tank. WTF, TI?
[15:47] <AHelper0> Do you have a post-2006 model? one that has 3 ram pages?
[15:47] <Runer112> no
[15:47] <AHelper0> hmm
[15:47] <KermM> Do you have a program that uses the RTC to calculate the clock rate?
[15:48] <Runer112> yes
[15:48] <KermM> awesome
[15:48] <KermM> this bears significant investigation
[15:48] <AHelper0> yay
[15:48] <Runer112> I'm actually a little surprised that you guys didn't know about this =o
[15:48] <AHelper0> I knew about it
[15:49] <Runer112> oh
[15:49] <AHelper0> I assumed that TI just doesn't know where to buy parts from :-P
[15:49] <KermM> Runer112: I think someone should start a topic about it.
[15:49] <Runer112> truspeed, where are you
[15:49] <KermM> This has some bearing on whether we can actually overclock the TI-84+/SE, which I always assumed we couldn't
[15:49] <Runer112> oh
[15:49] <AHelper0> I will if you wantHere is a snip on testing and gathering data:
Code:
[15:51] <AHelper0> What data do we need, though?
[15:51] <Runer112> well clock speed obviously
[15:51] <KermM> Clock speed of calcs
[15:51] <Runer112> but I always wondered about stuff like LCD delay
[15:51] <Runer112> LCD refresh rate
[15:51] <AHelper0> model, hardware version, etc?
[15:51] <Runer112> yeah
[15:51] <KermM> Ideally the degradation of clock speed over battery life
[15:51] <saxjax> (C) [aeTIos] Hmm
[15:51] <KermM> I'll pump in some controlled voltages to my TI-84+ and see what happens
[15:51] <AHelper0> Wait
[15:52] <KermM> Waiting.
[15:52] <AHelper0> if the CPU runs at any slower, then you can hook up an osicl. to the IO port and see if changing voltages makes a slight difference in timing :-)
[15:52] <KermM> True
[15:52] <AHelper0> (given that you have a program to pump stuff out on the IO port)
[15:52] <KermM> That would be a good way to measure
[15:52] <Runer112> or use the fixed-speed crystal timers
[15:52] <AHelper0> Not that that would be the test everytone must do :-P
[15:53] <KermM> wait 10000 cycles, toggle the port, wait 10000 cycles...
[15:53] <Runer112> which is what I know thepenguin77 has used in the past
[15:53] <Runer112> going to try his TRUSPEED right now actually
[15:53] * KermM is sad that people don't disseminate their hardware work
[15:53] * KermM glares in example at Simon
[15:53] <Runer112> jeez my calc is getting slower
[15:53] <Runer112> 14.59MHz
[15:53] <AHelper0> Use the USB to test?
[15:53] <AHelper0> hmm... GlassOS time!
[15:54] <Runer112> just use the fixed-speed crystal timers lol
[15:54] <Runer112> doesn't require external hardware
[15:55] <Runer112> but yeah, voltage definitely matters
[15:55] <Runer112> because if I run thepenguin77's CPU speed testing program over and over
[15:55] <Runer112> the results drop by a few KHzPhoto of the calc, better than datamath!
If you have tested your calc's speed, stay tuned here as we may need to do some polling and data collection.
If you have anything to add that you feel is important, post here
