Unfortunately, I must point out that the first program is measuring MathPrint's slowdown, not the TI-84+CSE. Try turning it from MathPrint to Classic mode, and you'll see the difference. I did some speed tests on my TI-84+CSE with Output and friends with results here: http://cemete.ch/p200193 . I tested the speed of graphscreen fills earlier in the thread, as well. Great effort, though, and nice test methodology!

Actually, I experience *nearly* no difference regardless of MP or CL mode. I uploaded the better of the videos that I shot for the news article, but I ran multiple tests. Here is an even poorer quality classic mode test of the same program:

http://youtu.be/CqLHsDInKn

Scratch that last link, I pasted it wrong.

http://youtu.be/CqLHsDInKnI

Ah, that one I can see. My bow towards your accurate tests remains; your criticisms are certainly valid.

That speed (for the 1 to 500 Disp test) is rather depressing. I do hope TI will consider some optimizations to help the display I/O bottleneck.

Fun fact: this is about the speed that the old TI-99/4A would print lines in its built in TI BASIC, IIRC. :-)

Travis, I just can't help myself:

oldcomputers.net/pics/ticosbyad.jpg

Although I wasn't able to see that view, I see your point. Since there are so many more pixels on the TI-84+CSE, and the CPU is not appreciably faster, the more pixels you need to change, the slower the TI-84+CSE is going to be than its brethren. Cheers again on quite a good video, self-castigation aside. ;)

I agree. I'm still very eager to see what differences exist between the current models floating around and the ultimate retail models. Additionally, I look forward to seeing the usual host (and of course, the new faces) of talented programmers finding ingenious ways of making ASM and BASIC performing at maximum capability :)

Seconded; I'm confident that this will help renew programmers' interest in finding new and amazing ways to make limited hardware do their bidding. I know that many among us enjoy a good challenge, which is exactly what this device presents. You have a more optimistic view than me about what's going to change between this model and the retail model, but I hope you're right.

Being heavily reliant on computations shouldn't get any slower, it's graphics I/O that will almost certainly be the new bottleneck.

I wonder if programmers can turn pixels a certian color...

BASIC programmers definitely will be able to do this on the graph screen, although I believe that the homescreen does not allow for this. ASM programmers will likely be able to bend time, space, and reality, assuming that there is enough available RAM.

This won’t be a challenge for most of the prospective programmers who have their hands on it already. I mean, Kerm Martian already got a nice picture of his face on one of these, how hard will the homescreen be?
Of course, I’m pretty sure he’s part Time Lord :P

I was posting on Slashdot earlier, calculating the number of pixels on the screen, and the speed of outputting bytes to an IO port on a Z80. I came to the calculation that it would be between 3 and 4 frames per second to redraw the entire screen.
From here, I can see that scrolling the screen is a complete redraw, and it looks like it's running between 3 and 4 frames per second. Looks like I totally called it.

So this means that any games running on this thing would need to use dirty rectangles in order to minimize drawing time, and scrolling would be very slow.

I’m afraid for what this may mean. Hopefully retail versions will be a much bigger improvement?

I think this is the classic "built in system routines are unoptimized" syndrome we've seen on all TIs.

Asm programmers will write their own much faster routines I am certain.

I really wish they switched over to the ez80 processor (like they were supposed to do originally)-everything would be so much faster and (presumably) without compatibility issues.