I recently found a TI-84 Plus Silver Edition prototype!
As far as I know, this is the first one found although Patrick Verstrepen who was an engineering supervisor at TI at the time has 2 prototypes. Aside from the fact that they existed, very little was known about these prototypes because he never released any pictures of those units.
This is the only information I had about these units:


Presumably "TE" stands for "Teacher Edition", meaning it would be the ViewScreen variant.

I did know a little bit more about TI-84+ prototypes though because one cropped up on cncalc a few years ago and Adriweb got to see one when he visited the TI Dallas headquarters in 2011. If those were anything to go by, I would expect a TI-84 Plus SE prototype to also not have a name on the front and have the shiny silver variant of the front faceplate instead of the light gray variant because the early production models had that finish. Both of those assumptions turned out to be correct.



One of the TI-84+ prototypes mentioned above did not have the theta symbol silkscreened above the "3" key, but that doesn't seem to be the case for this one. The silkscreening does have one notable difference though which is the "," key above the "7" key which looks more like a backtick and less like a comma.



On the back, the serial number is "PTR 2093" which is more or less the expected format but I'm still unsure what this "TR" nomenclature means. It was used on 84+, 84+SE and 89T prototypes.
Edit: This serial format was also used on the unreleased TI-PLT-SU1.
According to Patrick's spreadsheet, it seems like the digits are not a serial number (likely something to do with the hardware) but there is no other numbers visible on my proto so I'm not sure what the "real" serial number would be.



Interestingly, the polarity of the batteries is inverted, instructing the user to put the positive end of the battery on the spring?? Additionally, the coin cell battery is different (389 or 390) and it was assembled in Taiwan which is not where early production TI-84+SE's were made.

Taking a look inside, we can see that the hardware is pretty close to production but the ASIC is different (a completely new version not present on any production model). The layout of the board is also different of course. It was manufactured in July 2003, which is approximately 8 months before the first production models around March 2004.



The USB circuit is there and looks fine but the early OS doesn't support it yet. This means that installing this OS on your TI-84+SE is irreversible without a silverlink. It also seems like one of the inductors for the USB port (L02K) was hand-soldered as indicated by the flux residue.
The PCB label on the bottom of the board is also mirrored which looks rather strange

It runs 83+SE OS 0.01, Base code 0.01s and Boot code 0.01. It was successfully and painstakingly dumped one flash page at a time by Adriweb. The OS update file generated from the ROM image can be found here in the TI-Planet archives.



The OS is very similar to the TI-83 Plus OS from that period which isn't very surprising but the clock stuff (which was new in the 84+ series) differs slightly from the final version. For one, the "Set Clock" option in the MODE menu is called "CLOCKSET" and it allows you to turn the RTC on and off as well as just setting the time. The isClockOn command also acts slightly differently when called on the homescreen.
There is also a typo in the archive full error message, which is rather odd since that wasn't there in the 83+ OS.



In the ROM image, there is the "GO MAVERICKS!!" easter egg which was discovered by BrandonW a few years ago when TI accidentally released OS 0.46 on their website instead of the latest version. This is probably in reference to the Dallas Mavericks, an NBA team. The easter egg is only found in prototypes and a couple early versions but is one of the more interesting ones as it consists of a PCL job which can be sent to a USB printer to print the following 4 things:



Steve Falaster was a software engineer at TI who evidently worked on the TI-84 Plus SE.
Right before the easter egg, there is also the phrase "Texas Instruments TI-83 Plus USB", which was a name used internally by TI for the 84+SE prior to its release (a TI-84+SE is essentially a TI-83 Plus SE with a USB port, an RTC and the S-RAM integrated in the ASIC). Later versions of the 84+SE had "TI-REF 83PLUSB/TA2" ASICs which is another reference to this internal name.

Critor on TI-Planet has also written news articles about this prototype here and here.

Today, I thought I'd talk about my TI-83 Premium CE prototype.




Like many of my protos, I found it on ebay.
This is one of the less interesting ones since it is essentially a production version.
The OS is 5.2.0.0035 and the Boot is 5.1.5.0014. Both of these are production versions and they aren't particularly early either (mid-2016). This matches up with the hardware though which is also from that period. The factory is L (Kinpo, Philippines), which is the factory that usually produces the CE... This is somewhat surprising since the only other known prototypes were from Kinpo, China but those were all pre-production.
On the back, the serial number reads "L-DVT-000270" followed by "DVT NOT FOT SALE" underneath.
I'm not sure what this would have been used for since this is more than a year after release but I guess they were still working on it at that point.



The ASIC is ET2017-01 which is the main difference with production models from mid-2016. That ASIC was only added around revision I in March 2017. This, along with the sticker on the back that reads "Tester SL" makes me think that it would have been an internal dev proto.
Its also possible that this would have been used for conferences like T3 or Didacta. For the 83PCE specifically, I know that they had production samples like this one at those events so that seems less likely.

Inside, the board (SG93/F/T 10-1) is the same revision as my metallic blue sample except that its populated of course... It still hasn't been spotted on any production calcs, but B1 is still missing from the docs and I wouldn't be surprised if this was just rev B1.

One of the older calculators in my collection is a TI-83 Plus Silver Edition ViewScreen Marketing Sample.



This is one of 2 known units, the other being in Patrick Verstrepen's collection.
Currently, there is a total of 4 known pre-production TI-83 Plus SEs:

• This Marketing Sample
  
• Patrick V's Marketing Sample
  
• Patrick V's early "TI-83 Plus 2"
  
• Chaosleo's Engineering Sample

Here is the sample on the left and a pre-A non-VS board on the right for reference:



Interestingly, for the TI-83 Plus SE, the production of ViewScreen units (Jun 2001) seems to have started a few months after the production of the clear non-ViewScreen units (Feb 2001).
This sample was produced in March 2001 which is in that time gap. The MB is actually the same as those early non-ViewScreen units (9867MB-40D) which is significantly different from the revision that went into the first production VS units (9TA867MB-42F). That being said, the LCD board is of course not the same since it has the VS connector itself. This particular revision of the LCD board along with the VS board itself (9867VISC-40D) never made it onto any production units.



Since this was produced after mass production had begun (albeit only for the non VS model), the software is unsurprisingly not very interesting. The boot code is 1.00 and the OS is 1.18 which are both common known versions.
When I initially found it in 2019, I posted about it here but I failed to notice the gap between the production dates of VS and non VS units. Although the information in that post is true, it does not represent the whole context.

I think that's... super cool

I just received a very rare TI-84 Pocket.fr prototype from andylithia (a calculator enthusiast who bought it second hand in China).
The serial number on the back is K-DVT-192, meaning it was manufactured by Kinpo in China, it is in the 2nd (DVT) phase of development and it is the 192nd. This is followed by the "DVT NOT FOR SALE" stamp which is on most protos from the mid 2000s or later.



On the front, there is 1 interesting difference with production models, which is that there is a spelling mistake on the "fenêtre" key, which is spelled "fenétre".



Software-wise, it runs Boot 1.03 and OS 2.55MP so no surprises there, this is the same as all the TI-84 pocket.fr.



Finally, opening it up reveals some more interesting quirks.



The PCB labels are the same as production versions (SG85-10-1 for the MB and SG85-20-1 for the KB) but interestingly, it uses a MX29LV800CTTC-70G Flash ROM instead of the A29L800ATV-70F used in mass production. Presumably this is just whatever happened to be cheaper at a given time because both are fully pin-compatible (hence why the rest of the PCB is the same).



The date code on the PCB is 4310, indicating it was manufactured in the 43rd week of 2010, which puts it only a couple months before the beginning of mass production.

What a coincidence! I just discovered that my TI-84 Pocket.fr was a prototype. I was going through my collection a little while ago, testing, cleaning etc and discovered that the back of my TI-84 pocket.fr had DVT NOT FOR SALE engraved in the back. I'd never noticed it before. I bought it a long time ago, and somehow missed it. A little googling led me to discover that it was a prototype.I found this thread around that time, I just hadn't gotten around to posting this yet.

Mine is K-DVT-180 so it's a tiny bit earlier than yours. It's pretty cool since it's a prototype of a pretty uncommon calculator.

I just received the earliest known TI-86 proto which I mentioned in this post
This was a very generous gift from Joerg Woerner who runs the Datamath Calculator Museum.



It runs version 0.2026 which is the oldest known version of the TI-86 "OS".
Just like my 1.4001, this one has a socketed ATMEL AT29C020 flash chip which allows the unit to be reprogrammable. It also contains the code to accept OS updates through the link port using the hidden menu (2nd MODE ALPHA F), which is not available on production TI-86 ROMS since they use OTP-EPROM.
That being said, there seems to be a small bug with the OS transfer code. When it inevitably fails (since I can't send any OS to it), an error is displayed on the screen but the calculator is still in the mode menu even though it has not redrawn it. This bug wasn't present on 1.4001.



Unlike the other known protos, TILP can communicate with it but cannot dump it because AsmPrgm is missing. I suppose since Asm( is present, I could have run the dumper manually but the built-in TILP dumper doesn't do that.
I quickly checked the catalog and noticed that AsmComp( is also missing. There is an extra token called Xmark. Aditionally, randM( and randNorm( are swapped in the catalog so I guess we can add "the alphabet" to the long list of things TI can't do properly


It doesn't contain the "Willy DO BE DO BE DO" easter egg so that must have been added somewhere between 0.2026 and 1.0.
In order to dump it, I removed the flash chip and put it in a PLCC32 to DIP32 socket adapter and slapped that into my EPROM programmer like I described in this post.

On the back, there's no serial number, just like the other known protos.


WabbitEmu doesn't load the rom correctly and detects it as a TI-83 so no emulation for now.
EDIT: Works fine with TilEm (thanks critor)

Internal pictures are available on Datamath

Very nice find! I wonder if this proto or ROM runs faster than the commercial TI-86 and ROMs in TI-BASIC?

Its probably very similar since I think the access time for that flash chip is the same as the OTP-EPROM chips used in production and it does the same kind of paging which is what makes the 86 slow in general. I'd be happy to give it a try though.

I recently found a lot of TI-Nspire prototypes from China that contained a TI-Nspire CX CAS "CRII EVT2.1".


This is the first such proto to come up and by the looks of it, TI did not make very many of them since the serial number is 096.
Although the "EVT" in the serial indicates it is in the 1st (Engineering) stage of development, the CRII tells us this is not the engineering of the original nspire CX from 2011 but a later revision. CRII stands for Cost Reduction II and was released around April 2014, however documentation of these years is lacking so I'm not sure exactly when TI switched from CR to CRII and then to CRIII.
This is confirmed by the date code on the board which is 1425 (25th week of 2014 or mid-june 2014)
Taking a look at the hardware, this proto was produced around the time revision P was being produced (CRII) but the hardware looks like revision T (CRIII) from a few months later (putting aside the duponts which have been soldered on manually)
I'm not exactly sure what those 4 dupont wires would have been used for since the pinout of that cable as far as I know has not been documented. I think its somewhat interesting that they decided to attach male dupont connectors at the end of those wires instead of just leaving them bare like they usually do.
Obviously the back case is wrecked or as tr1p1ea put it "little care has been taken with cutting it". I would have to disagree here and say that a lot of care was taken with cutting it for several reasons:

1. It makes it much easier to open the calculator and access the boards
   
2. It allows access to the main part of the board without needing to open it at all
   
3. The cut conveniently goes around the reset button so that was probably intentional (this would also explain why its not just cut straight across)
   
4. Keeping the bottom half of the case is necessary if you want to use a battery
   

CRIII involved the removal of the TI-6C053A1 chip and changing the li-ion battery. As far as I know, we don't know what that chip actually did but the "datasheet" (not really a datasheet at all...) lists it as "Communications Equipment" so my best guess is perhaps a microcontroller for the wifi network adapter. If that's right, this would be a rather expensive component so removing it would definitely be a significant cost reduction. Whatever it was, it was not integrated into another component since all the other major chips on the board remained unchanged. That being said, the rest of the circuit changed significantly (including a lot more discrete components) so perhaps they just implemented whatever that microcontroller did (or just the relevant portions) in discrete logic?
As for the li-ion battery which was also changed in CRIII, this unit was shipped to me without a battery so I don't know.
One thing's for sure, TI was working hard on the TI-Nspire CX hardware around that time, releasing a revision about every 2 months in 2014! I suspect they would have had to have engineering stages at the very least for each cost reduction since those generally involve pretty significant hardware changes.

Software-wise, it runs OS 3.9.1.34 which is an unreleased intermediate version that will need to be dumped.

EDIT: I've just managed to dump it using tmpOS, a little-known tool that needs to be run while the OS is being sent from one calculator to another. The program copies the OS (which is kept in the temp folder during transfer) to the ndless root directory. It needs to be run after the entire OS has been sent but before it gets rejected and erased. It is available for download here.

It contains Boot1 version 3.0.0.99 and Boot2 version 3.9.1.34 and Diagnostics version 3.9.1.37 which are all known production versions. No surprises here since these were updated much less frequently than the OS.

I recently received 2 interesting TI-Nspire Touchpad prototypes.



Unfortunately both of them are broken and do not turn on at all. The flash chips on monochrome nspires are FBGA packages so we can forget about clips as well.
Obviously, there is a JTAG (right) and an ARM programmer/debugger (left) breakout that sticks out of the case for easy access. This would allow the devs to transfer data directly to the flash chip, reset the calculator and most importantly, get debug traces from the ARM9 processor embedded in the asic.
One of them has a daughter board attached which reads: "NSC_Fixture_MTBCB_20120112". This makes me think these prototypes might have been used to develop the TI-Nspire CX since prototypes of that model were called NSC for "Nspire Color".



Taking a look inside (they both have the same hardware), the board indicates a production date around September 2010, which lines up with when early TI-Nspire CX prototypes were produced. A few years ago, BrandonW obtained some prototypes, one of which had the TI-Nspire CX theme but in grayscale on a monochrome LCD, so this would not be the first monochrome TI-Nspire CX prototype. I suspect if it was turned on, that's what I would see. The JTAG connector is also populated on the MB which is of course not the case for production models.



Interestingly, they both have the same serial number but one of them begins with a "U". That being said, the date code on the case is nearly a year after the production of the board so I'm not sure what's up with that.



The 2 keypads are not the same, one of them is a production S-0507 pre-A clickpad while the other is a Simpkey A1-P prototype with "MP1-TEST" engraved on the front. Weirdly, the label "A1-P" is the same as this keypad found by critor which has a completely different design.

A few days ago, I received an unusual TI-84 plus SE with a prototype sticker and no serial number.



This sticker was used from around 2002 to 2006 and can be seen on all kinds of prototypes from that period (TI-84+, TI-89T, TI-Phoenix 1, TI-Nspire 84+ Keypad, PLT-SU1 to name a few)
It was sold as "broken" but as with a lot of ebay listings, I half-expected it to work, since I often see calculators with bad contrast settings or no batteries being sold as "broken", but after taking a look for myself, I concluded that it actually was "broken".
As with all the protos I receive, I proceeded to open it and I was quite surprised with what I found...



The board wasn't populated! Of course this reminded me of my blue TI-84 plus CE sample, but unlike that sample, this board is also missing the top layers so there are no traces or silkscreening, its just bare fiberglass. In fact, you can see right through the board.
I am not sure what this unit would have been used for. I suppose it could have been a sort of display sample, only meant to be looked at and not used (perhaps when finalizing the case design), but the state of the board inside suggests that perhaps it was used as a sort of production test to inspect the raw PCB. Its also possible that it was a display sample which got stuffed with whatever they had lying around. I've got no way to know for sure, I don't have retrocognition
In any case, I took a look at the locations of the pads and compared it to my PTR-2093 prototype and a pre-A. Neither of them was a perfect match so this board must have been an intermediate version between the two, which allows us to set bounds on when it was made: between Jul 2003 and Mar 2004.

Recently, I got my hands on a new TI-84 Plus C SE prototype (again from china).



Initially when I received it, the screen was broken and it identified as 2.55 on tilp and ti-connect. Tilp's dumper did not work (there seems to be an issue with timings) but I was able to transfer files. This is the same condition as the last prototype CSE (K-MSB-R-429) which ran OS 84C-020 with boot ROY3. However, this new prototype runs 84C-012 with boot ROY2, a much older version.
I started off by replacing the LCD, which is fairly easy to do on the CSE. I didn't have any perfect ones on hand but I had a slightly damaged one which is good enough since the contents on screen are now readable.

Hardware-wise, the PCB label is SG86-10-2, which is the same as production models. This is not very surprising since the CSE is clearly not far removed from the 84+SE, so there must not have been a lot of hardware iterations in between. That being said, there are a few differences which stand out.
The ASIC is a TA3, which is what we would expect, but it also has the "ES" marking which denotes Engineering Samples.
The plastic battery connector is missing and there is just the 4 metal pins. Initially I thought perhaps it had somehow been broken off given the extent of the damage, but the battery that came with it is also missing the connector.
It was manufactured in the 19th week of 2012, which is about 7 months before the first production calcs were made.
It has no serial number.

Thankfully, I was able to send the relevant flash pages to appvars, transfer those over to my computer and rebuild a rom image from those. Its quite a laborious process to do this manually but its similar to what I've done before for this TI-82 Advanced (without backup file shenanigans since the CSE natively supports asm programs) so I was able to get it done over the course of 1 or 2 hours last night with some encouragement from tr1p1ea in sax .

Now equipped with a rom image, I proceeded to diff it with 84C-020 to see what changed between those 2 versions and the answer is quite a lot!

1. The biggest difference between 84C-012 and 84C-020 is the catalog help which doesn't exist. (Page EC is completely empty on 84C-012 and has 14kb of stuff on 84C-020. Pressing + in the catalog does nothing)
   
2. There is the "Connected" graph format token, which is a relic from 2.55 (it was later replaced with Thick/Thin)
   
3. Some stats menus were tweaked slightly
   
4. There is only a Dot token (like on monochromes), no Dot-Thick/Dot-Thin
   
5. The TI Logo sprite used by apps and stuff is missing
   
6. There are a lot more bugs, a lot of them I'm unable to replicate, but I've run into a handful of bugs I'd never seen before just by testing stuff (example)
   
7. Matrices are broken, initializing a 1x1 matrix throws a memory error consistently and junk gets displayed in the editor. (84C-020 is the first version to fix the angle() bug with matrices so perhaps they were working on that?)
   
8. All the interesting things from 84C-020 are there of course.

Very nice and great that you were able to labour through dumping the ROM - which looked tedious to say the least.

Very good that you can preserve this history .

Congratulations, a nice piece of history saved

I'm glad it works at all! Looks like it took quite a beating. Probably from the angry engineer who had to use it
Thanks for finding and documenting this!