In a previous news, the first performance test of the new TI-84 Plus C Silver Edition made it appear as an extremely slow device.
But this test was comparing it with a TI-83 Plus that has a different CPU clock and which is not weakened by the MathPrint mode. When we compare things, we must try to only change one parameter at a time.

Later on, we confirmed this sluggish behaviour by only looking at the menu display time.

However the fist 84C ASM game, by Kerm Martian showed very correct performances.


All this seems to be a little blurry and weirdly contradictory, so let's test the performances of text output function in TI-Basic on the text screen :

We'll be thus testing the performances of:

• TI-84 Plus C Silver Edition, MathPrint mode
  
• TI-84 Plus C Silver Edition, Classic mode
  
• TI-84 Plus Silver Edition, MathPrint mode
  
• TI-84 Plus Silver Edition, Classic mode
  
• TI-83 Plus.fr (2008, Classic mode)

Let's start with a simple loop displaying the integers from 1 to 500, with the Disp instruction.

Here are the results:

• 1st) TI-84 Plus Silver Edition Classic Mode (0min33s)
  
• 2nd) TI-83 Plus.fr (0min35s)
  
• 3rd) TI-84 Plus Silver Edition MathPrint Mode (1min53s)
  
• 4th) TI-84 Plus C Silver Edition Classic Mode (2min14s)
  
• 5th) TI-84 Plus C Silver Edition MathPrint Mode (2min33s)

So:

• On this program, the Classic TI-84 Plus, despite its 2x faster 15MHz CPU, has similar performances to the TI-83 Plus with its little 6MHz CPU.
  
• When enabling the MathPrint mode, as already known since the 2.55MP OS release, the TI-84 Plus becomes 2.5x slower but the TI-84 Plus C only slows down by 14%.
  
• In Classic mode, the TI-84 Plus C is 4x slower than the 84 Plus.
  
• But in MathPrint mode, the 84 Plus C is only slowed down by 35% more than the 84 Plus. Would the slowness issues in MathPrint mode be fixed ? Or is it the overall slowness of that beast that's "hiding" it ?

Anyway, the 84 Plus C, whether in Classic or MathPrint mode, still is at the end of the rankings. The vertical scrolling needs a full-screen refresh, which is now slow for 16-bit colors for the little z80 CPU that does that in about 1 second for 2 new lines...

But even if it comes last, we could in theory have expected something like 16x slower than a 84+, since it's going from 1 to 16-bit colors, which is far from being the case.



Let's go on and try to find these optimizations with a 2nd test, but using the Output instruction. We now have no more vertical scrolling.

Here are the results:

• 1st) TI-84 Plus Silver Edition Classic Mode (3s)
  
• 2nd) TI-83 Plus.fr (5-6s)
  
• 3rd) TI-84 Plus C Silver Edition Classic & MathPrint Mode (7-8s)
  
• 4th) TI-84 Plus Silver Edition MathPrint Mode (16s)

So:

• On this program, the 84 Plus in Classic mode still gets first.
  
• The TI-83 Plus is almost twice as slow, which is coherent.
  
• With MathPrint enabled, the TI-84 Plus is also 2.5x slower, but the TI-84 Plus C seems not to be slowed down at all !
  
• The TI-84 Plus C is then 2.5x slower than the TI-84 Plus in Classic mode, but 2x faster than the TI-84 Plus in MathPrint!

This confirms the fact that optimizations were done, otherwise the 84 Plus C would still have been last.
More precisely, KermMartian from Cemetech discovered that contrary to the other calculators, theTI-84 Plus C does not necessarily refresh all the screen, but can limit the refresh to a specific area of the screen. An optimization perfectly fitted for our test !



In conclusion, the TI-84 Plus C is not, in my opinion, the catastrophe that some websites try to present.

It is clear that it isn't fitted for games needing a full-screen refresh, like games using horizontal/vertical scrolling, emulators, or Doom-likes.

But on other types of games, it is perfectly capable of doing even better than its predecessors and moreover in color, especially since most users stay with the latest OS with MathPrint enabled (by default), which is the now-optimized configuration !



Video for the test:




Source:
http://tiplanet.org/forum/viewtopic.php?t=11309&lang=en

Critor, very nice summary, thank you for sharing it with us. I agree that some people seem to be blowing the shortcomings of the device out of proportions. I'm surprised that TI did not optimize some of the graphics more fully. For example, pasting program names from the program menu makes it redraw the whole name for each letter it pastes. The same thing is true while typing out math equations.

And sometimes, we get the exactly opposite behaviour: the calculator forgets to refresh some parts of the screen.

Which menu(s) or tools have you seen that bug in? I've been trying to keep a record of all the bugs I find so that I can pass them along to my contacts.

Yeah I think the issue is that some people focus only on the full screen updating slow speed, rather than other stuff. Videos we have seen so far makes it clear that the calculator is fast to display smaller things, just not for full screen animations or updates.

I wonder how a 2008 TI-83 Plus.Fr (which is a TI-83 Plus, unlike the 2013 one) manages Disp speeds on-par with a TI-84 Plus in classic mode, though?

Presumably it's nearer to being an I/O bound operation, rather than a CPU bound one, and iirc, don't the 83+s actually have better LCD controllers than the 84+s?

Because as Elfprince13 says, on any of the older calculators, the display is I/O bound, not CPU bound. Both the 6MHz and 15MHz z80s can push 96x64 pixels at a pretty comfortable speed while leaving power left over for computation. That's why all of our games have worked for so long. That same 15MHz CPU cannot push (320*240*2)/(96*64/ = 200 times as many bytes per screen-ful without the obvious 200-fold decrease in refresh rate.

Ah, that might explain it. I was sure that my 83+ had slower speed than that 83+ Fr model, though. I will need to try again when I find my batteries. Maybe the French model or the more recent TI-83 Plus got a better LCD than older ones.

Also Disp is about twice faster on my TI-81 than my TI-83 Plus. I can't check on my FX-7000G, because the only Disp command on it pauses execution.