I have never used the HP 48/49g series myself, so I cannot give a first hand report on the debate, but I can repeat what I've heard about them and quote some comparisons others have written. When some folks complained that the old version of this page was horribly out of date, I threw out a request for new material. A fellow who never gave a name obliged, and what you see here in the block quotes are excerpts from his posts, with some comments from me interspersed. (Of course the old page isn't all bad either, the stuff towards the bottom isn't that old and I think mostly still holds true.)

The HP calculator at issue here is the HP 49g+, an update to the HP 49g, which was generally praised as a step beyond the original HP 48 models yet never quite caught on. However, the HP 48gII is still around, as is the HP 39g+, which as far as I can tell is somewhere between the HP 48/49g series and the TI-83 Flash family in power. On the TI side is the TI-89 Flash family, the latest and greatest of which are the Voyage 200 and the TI-89 Titanium.

Naturally I'm not the only one attempting to do this. Alistair Borowski wrote a comparison of the TI-89 and the HP 49g+. Bhuvanesh Bhatt has performed benchmark testing to compare the two calculators. Though slightly dated (and therefore probably inaccurate in some places given current OS versions and whatnot), there's also the following: HP highlights the HP 49g's advantages over the TI-89 in this file. Steen Schmidt compared the TI-89 and HP 49g with a stopwatch . Tim Wessman wrote a comparison of the TI-89 and HP 49g, it's available as a pdf file or an html file.

For a starting synopsis, the general impression I've gotten is that it's reached the point where the two brands are functionally more alike than they are different, even though they often

• The hardware of the HP49g+ is generally better then the ti89, with 3 exceptions: The screen has a lower resolution, and the keys on early production runs are dreadful (hp has since redesigned the keyboard, and places like http://commerce.hpcalc.org say they ship the good one). I think this is a non-issue for new batches, but the buyer may get unlucky if they get old stock. There is also less flash memory, but the SD card slot negates this.

The rest of the hardware is a generation ahead of TI's offerings. With 3rs party tools, the CPU can be clocked between 12 and 203MHz, slowing down for simple games to save power, or speeding up for temporary calculations. There is flickerless 16 shade greyscale in hardware. Theres also an IR port (though this is probably just a gimmick).

• With the new CPU, the calculator speed seems to have crossed a psychological barrier - it now seems 'fast enough', even though most of the ROM is emulated. Of course programs coded in ARM asm will blow any other calculator out of the water. Both the TI and the HP plot roughly the same number of pixles per second. Since the HP has a lower resolution screen, I've heard it seems to plot faster.

• Software wise, it starts in algebraic mode by default. It is very similair to use as TI. EG to factor 'x^2+2x+1', you go red-shift 4 (the ALG menu, says ALG in red above the key). and a little box of algebraic comands pops up. You select factor. The text FACTOR() pops up, with the cursor in between the brackets. You type in 'x^2+2x+1' inside the brackets and press enter. It spits out (x+1)^2. To store that result as Q, press STO, then Q , then enter.

To plot y=x, press blue-shift then A (says y= above it in blue). It says, "no equ., press ADD" . Press ADD. It shows Y(x)=. Fill in X. press ENTER, then ERASE to srase the previous plot (if there was one), then DRAW.

• the CAS often asks to change modes for no reason at all - it seems to want to stay in RAD mode have the time.

As I mentioned, there are plenty of add on programs for each calculator. Of course the type of programs are somewhat skewed by the primary user base. Much of the TI offerings are appropriate for high school and college level math, while professional engineers have been using the HP 48 series for a much longer time and have put out many programs in that area. The TI calculators do have some robust engineering programs of their own though.

Just for the heck of it, I asked about how the calculators compare before adding any such extra software.

IF you mean 'as is', then the most obvious thing in favour of the HP is the equation writer. I consider equation writers as fantastic tools for long equations.

The other major thing I can think of that favours the HP is built in laplace transforms. These are very useful in some engineering work.

There is also the selection between algebraic and RPN mode, TI89 and TI89 style menus, but you know about those things already.

In TI's favour, is TI basic. Its much easier to write simple programs then userRPL. Also it does 'textbook integrals' much much faster then the HP does.

Also, the HP49 creates all these stupid variables when you graph something. These store the plot size etc, but often you don't care about them. They just clutter your current working directory. The TI89 doesn't. This is a bonus in favour of TI. Not a major issue but an annoyance.

If you are dealing with huge equations, then an equation writer makes things much easier. I also personally prefer RPN, but I know most people don't (most people have never heard of it).

In the old days, where RPN was the only option, then I would say it had a much steeper learning curve. These days, built in help and almost identical algebraic syntax by default make things much easier for new users.

From a programming perspective, HP doesn't set any intentional limits. Theres no need to get signings to make software for instance. On the other hand most of these limits have been bypassed long ago (copying flash apps is the exception). 99% of users won't care about this anyway, so big deal.

You may notice that the equation writer and RPN are two flagship favorite features, and this has been true since the HP 48g days. Once we allow add on programs again, both of those have been released by third parties for the TI-89 Flash family calculators. (You can find them in this list of programs.) Though I think some of those haven't been maintained and may not work under the more recent AMS versions.

Another way to compare the calculators is to see what their options and settings screens look like:

I can think of 3 main areas of interest:

1. When you press the MODE key, you get to select the basic functionality of the device. Eg you can choose between algebrac and RPN operating modes; you can turn off the beeper incase the user is doing an exam or something that requires silence. Also a rad/deg/grad setting. These are are pretty straighfoward and not a problem at all.

2. From the above menu, there is an option for CAS settings. Selecting it brings up a new menu.

From this menu, there are 8 yes/no options to choose, and also an option to set the independant variable. Its these settings that you've probably heard as confusing. There is 1 line of help for each setting. They are:

------

'Numeric Mode' - replace constants by values (I never use this, its a relic from the 48 series. I think its useless).

'Approx Mode' - Makes it no longer gives exact answers, only numbers and symbols. Sqrt(2) would give 1.41.... for instance.

'Complex mode' - allow complex answers.

'rigerous' - if checked, assumes sqrt(x)^2 = ABS(x), not x. Is checked by default.

'Verbose' - basically a useless setting for 99% of users. If checked, shows some text while the CAS is doing things.

'Step by Step' - if set, shows step by step solutions for some things. Great for matrices; generally crap for integration

'Incr Powers' - just a display setting for polynomials. Puts the powers increasing instead of decreasing order.

'Simplify Non-rational'. 'nuff said.

-------

3. The 'flags menu'. This contains a heap (30 or more) of yes/no settings. This is a superset of the previous menus. It lets you change exotic things, like the type of menu style, the modes of the built in compiler, etc. I see no reason for users to have to mess with it, except maybe once or twice to setup the mahine as they like it.

I don't think these settings are too hard to get to grips with. They all have at least 1 line of help explaining briefly what they do. There are shortcut keys to change most things quickly, eg toggle complex or exact mode, without having to go through menus.

The CAS will often prompt you to change modes when needed.

The problem is, (and this is the thing people complain about), the CAS is really only happy in radians mode. If you are in degrees mode, it will occasionaly prompt you to change into radian mode, even where it looks like it isn't necessary. This is an annoyance.

The flexibility does have benefits, but I would call the CAS mode changes a disadvantage. Parisse is very stubborn and refuses to fix this despite alot of complaints.

Naturally the TI-89 family has its share of settings. Exact/Approx/Auto for numerical output, for instance. Degrees and radians of course, though not gradians. Complex number display options, graph types, etc., the usual obligatory stuff. No step by step mode at all, much to the chagrin of students. If it sounds like the HP has a few more settings for somewhat more obscure things, that's probably not far from the truth and would be in line with the nature of the functions it has that the TI doesn't:

Of course, some things are easier with HP. There are more dedicated commands for less common tasks. There is a command that just does partial fraction decomposition for instance, and does it well, while I think the TI89 has no specific command.

Well, the extra commands are typically CAS commands - mostly not trivial to write. While something can be coded, whether it actually been has is a different story. In theory you could make a CAS for the TI83/4, but no-one has because its not worth the hassle.There _are_ good CAS libs for the TI's, but I think you would have to install a fair number of programs to get all the CAS features of the HPs.

Theres something to be said of having more commands built it - but lets face it, some are fairly advanced or exotic commands most people won't bother learning to use. Case in point: Go Convert->Trig. This brings up about 13 commands specifically for rewriting trig commands.

Want to linearize (a*x)*cos(b*x) for instance? Type it, then use TLIN to get cos(xa-xb)/2+cos(xa+xb)/2

While there is some on-calc help for each of these commands, including an example for each, it can be hard to find the right command unless you know what to look for. The CAT menu reports about 800 built in commands!

Only when you get to the more interesting mathematics does the HP's built in CAS begin to shine. If you spend a day filling the 2 calcs with programs from ticalc/hpcalc, I'm sure you probably could do the same things with both. In my opinion, the TI is a little easier to use from the start, the HP is a little more powerful.

So, what have we learned? I think it's just that the calculators are true to their roots. TI makes its calculators primarily for the classroom, so the TI-89 family delivers well on the sorts of things you would encounter in calculus class. HP sold graphing calculators to engineers back when the high cost of such a device made it not worth it to anybody else, so the presence of a variety of functions most would find obscure isn't surprising, while still getting the job done when it comes to the more mundane, everyday things.

I still like this wonderful analogy from a post by beezaur some time ago that points out how you fit the tool to the job: "Think of it in terms of cars: a hot rod vs. a HummVee. One is fun to drive, fast, and great for and going where you normally go. The other is a difficult beast in the city and only comes into its own in really rough terrain, but there it shines."

Now, the obligatory summary:

If I had to summerize my opinions on the debate, I would say:

Mathematics: You can pretty much do the same things on both calculators. Equation write is a good feature - built in on HP, can download some for the TI. Built in RPN option on the HP is good if you like it. The HP's CAS is annoying if you are not in RAD more. The TI's CAS is probably a little easier to use at first.

Hardware: I think theres no contest. HP wins by a mile. 75MHz CPU, SD card slot, IR port, beeper, the HP is better in almost every way - the exception being the screen resolution.

Hardware Quality: I think TI wins here. HP really screwed up the keyboard on early batches, and some people still aren't happy with the new keyboard. I think its useable but not fantastic (HP will replace early units with poor keyboards if you telephone them). The TI has a hard plastic cover while the HP has a soft black (leather?) cover. This is not as robust, but it sure looks classy.

Software: The tables have completely turned here. While the HP used to have great software written in hand optimized in asm, and the TI's used is written in C, the emulation layer destroys most of the advantage. The TI still wins some of the speed trials. HP is more flexible, and many things can be changed very easily. TI is probably more straitfoward to just 'pick up and use' by a completely new user.

Speed: The TI is generally faster for integration and other table based CAS routines. The HP is faster for general purpose things. UserRPL is faster then TI-basic, but much more cryptic. Programs written for ARM code are much much faster then any other calculator (there are not many at present, but things are changing now there is a C compiler)

Games: TI has many very good games. HP has no good ones at the moment (but it has 2 versions of tetris built in)

Development: the TI is fairly well documented by now. There is TIGCC and flash studio to write programs in. These are very good tools. You can write programs in tibasic, asm and C.

On the HP, you can write in user RPL (a built in langeage that shouldn't ever crash the calculator, but the syntax is very weird), HP-basic (really just an algerbraic version of user RPL), system RPL (a superset of user RPL thats faster, but can cause crashes), asm (arm and virtual saturn), and finally C (but the C libs are very small at this stage).

There is an on-calc system rpl / asm compiler built it but there are special keys needed to activate it to stop naive users causing crashes. Also HP doesn't intentionally place limits on what you can program. TI does, but these limits have mostly been bypassed.

Naturally 99% of users won't care about development, so its not a big factor.

Here's something else, part of a post by Steen Schmidt made to comp.sys.hp48 that brings us a more quantitative comparison of the TI-89 and HP 49g+:

Subject: Re: TI89 Titanium -- opinions?

Newsgroups: comp.sys.hp48

Lets' see...I'll just spend a moment digging through my last 300 hours of pitting the TI89 HW2 against the HP49G+;

I have compared almost every aspect of these two machines, ranging over object diversity, battery life, OS stability, algorithms, programming, plotting, I/O, 3rd party apps, customization, so on and so forth. And of course mathematical abilities, inspecting practicality/integration, diversity, usefulness....and of course performance.

I gather you'd like to stick with the latter, as that seems to be the only aspect of this you might put any weight to. I'll entertain with the very short story only (all claims are isolated, i.e. measurement overhead is deducted from the test results. All claims are general, in each case an opposite scenario can easily be constructed):

- The HP is roughly twice as fast for real numerical computations.

- The HP is 2-6 times as fast for complex numerical computations.

- The HP has better numeric to exact conversion.

- The HP is the most versatile regarding indefinite integration, and often the fastest too. Both
have quirks in this area.

- The HP is *much* faster at series expansion, and it is more versatile too (the TI really only
sporting Taylor expansion).

- The HP sports a host of modulo arithmetic which the TI has not. The same is true for a bunch of
special functions and integer and polynomial arithmetic. The latter is negated when taking Bhuv's
MathTools for the TI into account. A truly magnificent set of programs.

- The HP is more versatile regarding partial fraction expansion, mainly stemming from the choice
of complex or real mode.

- The HP is *much* more versatile when it comes to expression manipulation, in part helped by
the EQW. A couple of really high quality EQWs exist for the TI as well now.

- The HP is *much* more versatile when dealing with trig functions, while the TI is usually
faster. The versatility of the HP must prevail here though.

- The HP is 2-5 times as fast for symbolic matrix calculations, and 2-10 times as fast for numeric
array calculations.

- The HP is much better when dealing with units. Beyond description better.

- The TI is 2-4 times as fast for integer arithmetic.

- The TI is faster at differentiation, except when auto simplification plays a major role (typically
involving trig functions).

- The TI is *much* faster at numeric integration.

- The TI is *much* faster at limits, except when we're dealing with simple substitution.

- The TI is faster when doing symbolic and numeric sums. The Ti can also do symbolic closed
form products, while the HP cannot.

- The TI is much faster and more versatile at solving symbolic and numeric equations, except the
special case of univariate polynomials with strictly numeric coefficients.

- The TI is generally better at factoring polynomials.

- The TI is usually the master of simplification, substitution and evaluation, except again when its
auto simplification routines trip over each other. The HP is just generally half-slow, but rarely
ice-age slow (The only case that springs to mind is when dealing with RE, IM and ABS - forget
about it!).

- They both suck at integer factorization and primality tests (although the TI seems to implement
the most stable primality test by a hair).

- They are about equal dealing with vectors.

The TI has one fast parsing engine. It's truly amazing how quickly it can apply its heuristics to expressions - I'm currently trying to mimick some of these heuristics on the HP49G+, and the parsing is really a bottleneck (or I'm not clever enough).

One aspect of the TI, I find very comforting, is that it often didn't fail some complex computation, but instead kept going for hours and hours while the HP would give up after a very short time if it couldn't find an answer. Time limited algorithms are not good on a calculator in my opinion - they are necessary on a mainframe when the algorithm has determined that a solution within all probability will take millenia to compute, but on a calculator I fancy the user to press CANCEL rather than the HP not factoring something (not even a warning is issued, that the algorithm aborted because of time constraints).

One thing I like about the HP is the lack of auto simplification - or, more correctly, I'd very much like it to be there, but the user should be able to switch it on and off. This should be so on the TI as well.

The keyboard layout on the TI89 is bad, but the keys on the HP49G+ sucks so much I don't have words for it. The layout is perfect, but I miss keystrokes constantly - if I enter one simple expression, say, with 25 keystrokes or so, I see myself retyping maybe 4 or 5 entries (sometimes I have to press a key 3 or 4 times before it registers). When the HP is turned ON, my first keypress is often not registered. Delimiters is a problem area for me too - I have a hard time getting my current HP to register delimiter entry like ' ', " ", << >>, { }, [ ] and so on. Also the arrow keys, negate, inverse.....well, I'll be d***** - It's *all the keys*!

And without getting into too much detail, TI-Basic isn't really slower, generally speaking, than UserRPL. TI-Basic sucks at looping in most instances, and UserRPL really shines when the stack is utilized (which not that many people can do effectively really). TI-Basic is very limited, while programming the HP has almost no limits (that is when judging on-calc programming). The SDK & C development features are much better on the TI. By a large margin.

In the end, both brands have good calculators, each with dedicated followings. It depends on where your math skills stand and what you will use it for as to which you should get, since each has strengths in different areas. However, since the largest market for graphing calculators is students, the TI models do seem to be favored overall, regardless of any usability or power differences. Due to their marketing focus on math teachers, Texas Instrument defined the market for graphing calculators in American schools and are the standard there. The teachers are trained in, ask for, and teach with the TI-83 family and TI-89 family. While the required supplies lists that specify these calculators often say "or equivalent", it can be confusing to have a different calculator than the one the teacher is using to demonstrate, especially a different brand of calculator entirely. (Not to say it can't be done. Students dedicated enough to read the owner's manual and also understand it can use any calculator they want. I speak about the average student, who is not inclined to bother with the manual.) Even though HP's offerings are still the favorite old workhorse of many an engineer, TI is poised to all but completely take over as students using mainly TI models move out into industry.

Want to compare more before you decide which to get? HP made an HP49G emulator available, and you can get the manual on-line. As for TI, they've got the manuals on-line too, and the Flash ROM file is available on their site. Third party emulators such as Virtual TI are freely available. I'll let you put two and two together for this one since TI probably doesn't look too kindly on this sort of thing. At any rate, please don't hang onto an emulated TI for too long unless you buy a real version. Emulated calculators aren't very portable anyway.